Pioglitazone for prevention or delay of type 2 diabetes mellitus and its associated complications in people at risk for the development of type 2 diabetes mellitus

Emil Ørskov Ipsen; Kasper S Madsen; Yuan Chi; Ulrik Pedersen-Bjergaard; Bernd Richter; Maria-Inti Metzendorf; Bianca Hemmingsen

doi:10.1002/14651858.CD013516.pub2

Pioglitazona para la prevención o el retraso de la diabetes mellitus tipo 2 y las complicaciones asociadas en personas con mayor riesgo de desarrollar diabetes mellitus tipo 2

Declaraciones de intereses de los autores

Versión publicada: 19 noviembre 2020 Historial de versiones

https://doi.org/10.1002/14651858.CD013516.pub2

Contraer todo Desplegar todo

Resumen

disponible en

Antecedentes

El término prediabetes se utiliza para describir a una población con un elevado riesgo de desarrollar diabetes mellitus tipo 2 (DMT2). Con las proyecciones de un aumento de la incidencia de la DMT2, la prevención o el retraso de la enfermedad y sus complicaciones es primordial. Actualmente se desconoce si la pioglitazona es beneficiosa para el tratamiento de las personas con mayor riesgo de desarrollar la DMT2.

Objetivos

Evaluar los efectos de la pioglitazona para la prevención o el retraso de la DMT2 y las complicaciones asociadas en las personas con riesgo de desarrollar DMT2.

Métodos de búsqueda

Se hicieron búsquedas en CENTRAL, MEDLINE, bases de datos chinas, el portal de búsqueda ICTRP y ClinicalTrials.gov. No se aplicaron restricciones de idioma. Además, se investigaron las listas de referencia de todos los estudios y revisiones incluidos. Se intentó establecer contacto con todos los autores de los estudios. La fecha de la última búsqueda en las bases de datos fue noviembre de 2019 (marzo de 2020 para las bases de datos chinas).

Criterios de selección

Se incluyeron los ensayos controlados aleatorizados (ECA) con una duración mínima de 24 semanas, y participantes diagnosticados con hiperglucemia intermedia sin enfermedades concomitantes, que compararon la pioglitazona como monoterapia o como parte de un tratamiento dual con otros fármacos hipoglucemiantes, intervenciones de modificación de la conducta, placebo o ninguna intervención.

Obtención y análisis de los datos

Dos autores de la revisión, de forma independiente, revisaron los resúmenes, leyeron los artículos de texto completo y las entradas, evaluaron el riesgo de sesgo y extrajeron los datos. Se utilizó un modelo de efectos aleatorios para realizar el metanálisis, y se calcularon las razones de riesgos (RR) para los desenlaces dicotómicos y las diferencias de medias (DM) para los desenlaces continuos, con los intervalos de confianza (IC) del 95% para las estimaciones del efecto. La certeza de la evidencia se evaluó con GRADE.

Resultados principales

Se incluyeron 27 estudios con un total de 4186 participantes asignados al azar. El tamaño de los estudios individuales varió entre 43 y 605 participantes y la duración varió entre seis y 36 meses. Se consideró que ninguno de los estudios incluidos tuvo bajo riesgo de sesgo en todos los dominios de "Riesgo de sesgo". La mayoría de los estudios identificaron a las personas con mayor riesgo de presentar DMT2 por la alteración de la glucosa en ayunas o la alteración de la tolerancia a la glucosa (IGT), o ambas.

Las principales medidas de desenlace fueron la mortalidad por todas las causas, la incidencia de DMT2, los eventos adversos graves (EAG), la mortalidad cardiovascular, el infarto de miocardio o el accidente cerebrovascular no mortal (IM/ACNM), la calidad de vida relacionada con la salud (CdVRS) y los efectos socioeconómicos. Las siguientes comparaciones en su mayoría sólo informaron de una fracción del principal conjunto de desenlaces.

Tres estudios compararon la pioglitazona con la metformina. No informaron de la mortalidad por todas las causas y cardiovascular, ni del IM/ACNM, la CdVRS o los efectos socioeconómicos. La incidencia de DMT2 fue de 9/168 participantes de los grupos de pioglitazona versus 9/163 participantes de los grupos de metformina (RR 0,98; IC del 95%: 0,40 a 2,38; p = 0,96; tres estudios, 331 participantes; evidencia de certeza baja). En dos estudios (201 participantes; evidencia de certeza baja) no se informó sobre los EAG. Un estudio comparó la pioglitazona con la acarbosa. La incidencia de DMT2 fue de 1/50 participantes del grupo de pioglitazona versus 2/46 participantes del grupo de acarbosa (evidencia de certeza muy baja). Ningún participante presentó EAG (evidencia de certeza muy baja). Un estudio comparó la pioglitazona con la repaglinida. La incidencia de DMT2 fue de 2/48 participantes del grupo de pioglitazona versus 1/48 participantes del grupo de repaglinida (evidencia de certeza baja). Ningún participante presentó EAG (evidencia de certeza baja).

Un estudio comparó la pioglitazona con una consulta personalizada de dieta y ejercicios. No se informó de la mortalidad por todas las causas y cardiovascular, ni del IM/ACNM, la CdVRS o los efectos socioeconómicos. La incidencia de DMT2 fue de 2/48 participantes del grupo de pioglitazona versus 5/48 participantes del grupo de consulta de dieta y ejercicios (evidencia de certeza baja). Ningún participante presentó EAG (evidencia de certeza baja).

Seis estudios compararon la pioglitazona con el placebo. Ningún estudio informó sobre la CdVRS ni los efectos socioeconómicos. La mortalidad por todas las causas fue de 5/577 participantes de los grupos de pioglitazona versus 2/579 participantes de los grupos placebo (odds ratio de Peto 2,38; IC del 95%: 0,54 a 10,50; p = 0,25; cuatro estudios, 1156 participantes; evidencia de certeza muy baja). La incidencia de DMT2 fue de 80/700 participantes de los grupos de pioglitazona versus 131/695 participantes de los grupos placebo (RR 0,40; IC del 95%: 0,17 a 0,95; p = 0,04; seis estudios, 1395 participantes; evidencia de certeza baja). Hubo 3/93 participantes con EAG de los grupos de pioglitazona versus 1/94 participantes de los grupos placebo (RR 3,00; IC del 95%: 0,32 a 28,22; p = 0,34; dos estudios, 187 participantes; evidencia de certeza muy baja). Sin embargo, el estudio más grande de esta comparación no distinguió entre eventos adversos graves y no graves. Este estudio informó que 121/303 (39,9%) participantes del grupo de pioglitazona versus 151/299 (50,5%) participantes del grupo placebo presentaron un evento adverso (p = 0,03). Un estudio observó la mortalidad cardiovascular en 2/181 participantes del grupo de pioglitazona versus 0/186 participantes del grupo placebo (RR 5,14; IC del 95%: 0,25 a 106,28; p = 0,29; evidencia de certeza muy baja). Un estudio observó un infarto de miocardio no mortal en 2/303 participantes del grupo de pioglitazona versus 1/299 participantes del grupo placebo (RR 1,97; IC del 95%: 0,18 a 21,65; p = 0,58; evidencia de certeza muy baja).

Veintiún estudios compararon la pioglitazona con ninguna intervención. Ningún estudio informó sobre mortalidad cardiovascular, IM/ACNM, CdVRS ni efectos socioeconómicos. La mortalidad por todas las causas fue de 11/441 participantes de los grupos de pioglitazona versus 12/425 participantes de los grupos de ninguna intervención (RR 0,85; IC del 95%: 0,38 a 1,91; p = 0,70; tres estudios, 866 participantes; evidencia de certeza muy baja). La incidencia de DMT2 fue de 60/1034 participantes de los grupos de pioglitazona versus 197/1019 participantes de los grupos de ninguna intervención (RR 0,31; IC del 95%: 0,23 a 0,40; p < 0,001; 16 estudios, 2053 participantes; evidencia de certeza moderada). Los estudios informaron de EAG en 16/610 participantes de los grupos de pioglitazona versus 21/601 participantes den los grupos de ninguna intervención (RR 0,71; IC del 95%: 0,38 a 1,32; p = 0,28; siete estudios, 1211 participantes; evidencia de certeza baja).

Se identificaron dos estudios en curso que comparan la pioglitazona con el placebo y con otros fármacos hipoglucemiantes. Estos estudios, con 2694 participantes, pueden aportar evidencia en futuras actualizaciones de esta revisión.

Conclusiones de los autores

La pioglitazona redujo o retrasó el desarrollo de la DMT2 en personas con mayor riesgo de DMT2 en comparación con el placebo (evidencia de certeza baja) y en comparación con ninguna intervención (evidencia de certeza moderada). No está claro si el efecto de la pioglitazona se mantiene una vez que se interrumpe. La pioglitazona, en comparación con la metformina, no mostró ni ventajas ni inconvenientes en cuanto a la aparición de DMT2 en personas con mayor riesgo (evidencia de certeza baja).

Los datos y la información sobre la mortalidad por todas las causas, los EAG y las complicaciones micro y macrovasculares en general fueron escasos. Ninguno de los estudios incluidos informó sobre la CdVRS o los efectos socioeconómicos.

PICO

Population

Intervention

Comparison

Outcome

El uso y la enseñanza del modelo PICO están muy extendidos en el ámbito de la atención sanitaria basada en la evidencia para formular preguntas y estrategias de búsqueda y para caracterizar estudios o metanálisis clínicos. PICO son las siglas en inglés de cuatro posibles componentes de una pregunta de investigación: paciente, población o problema; intervención; comparación; desenlace (outcome).

Para saber más sobre el uso del modelo PICO, puede consultar el Manual Cochrane.

Resumen en términos sencillos

disponible en

¿La pioglitazona previene o retrasa la diabetes tipo 2 y sus complicaciones en personas con riesgo de desarrollar diabetes mellitus de tipo 2?

¿Qué es la diabetes tipo 2?

La diabetes tipo 2, también conocida como diabetes de aparición en la edad adulta, es el tipo más común de diabetes. Evita que el cuerpo utilice la insulina de forma adecuada. La insulina es una hormona que ayuda al cuerpo a regular los niveles de azúcar en la sangre. Las personas con diabetes tipo 2 pueden tener efectos a largo plazo (complicaciones de la diabetes), como enfermedades de los ojos o los riñones, o desarrollar úlceras en los pies. Se dice que las personas con niveles de azúcar en sangre moderadamente elevados (a menudo denominado "prediabetes") tienen un mayor riesgo de desarrollar diabetes. La pioglitazona es un medicamento para reducir el nivel de azúcar en la sangre, que se utiliza para tratar a las personas con diabetes tipo 2.

¿Qué se quería averiguar?

Se quería saber si la pioglitazona también se puede utilizar para prevenir o retrasar la diabetes tipo 2 en personas con un mayor riesgo de desarrollar la enfermedad. Se examinaron los efectos de la pioglitazona sobre desenlaces importantes para los pacientes, como las complicaciones de la diabetes, la muerte por cualquier causa, la calidad de vida relacionada con la salud y los efectos no deseados del tratamiento.

¿Qué se hizo?

Se buscaron estudios que investigaran la pioglitazona utilizada para prevenir o retrasar la aparición de la diabetes tipo 2. Los participantes debían tener niveles elevados de azúcar en la sangre, pero más bajos que los niveles de diagnóstico para la diabetes, y no debían presentar otras enfermedades. Los estudios debían aplicar la intervención (pioglitazona) por al menos 24 semanas.

Qué se encontró

Se encontraron 27 ensayos controlados aleatorizados (estudios clínicos en los que los participantes son asignados al azar a uno de dos o más grupos de tratamiento) con un total de 4186 participantes. Los estudios compararon la pioglitazona con otros medicamentos antidiabéticos, dieta y ejercicio, placebo (un tratamiento "simulado") o ninguna intervención. Veintitrés de 27 estudios se realizaron en China. Los estudios duraron entre 24 semanas y tres años.

Esta evidencia está actualizada hasta noviembre de 2019.

Resultados clave

Cinco estudios compararon la pioglitazona con otros medicamentos antidiabéticos (metformina, acarbosa o repaglinida) y un estudio comparó la pioglitazona con la dieta y el ejercicio. No hubo efectos beneficiosos ni perjudiciales claros sobre el riesgo de desarrollar diabetes al comparar las medicinas.

Seis estudios compararon la pioglitazona con el placebo. Hubo una reducción o retraso en el desarrollo de la diabetes tipo 2: 188 de 1000 personas tratadas con placebo desarrollaron diabetes tipo 2 en comparación con 75 de cada 1000 personas tratadas con pioglitazona (posible dispersión: 32 por 1000 a 179 por 1000).

Veintitrés estudios compararon la pioglitazona con ninguna intervención. Hubo una reducción o retraso en el desarrollo de la diabetes tipo 2: 193 de 1000 personas con ninguna intervención desarrollaron diabetes tipo 2 en comparación con 60 de 1000 tratadas con pioglitazona (posible dispersión: 44 por 1000 a 77 por 1000).

Sólo unos pocos estudios informaron sobre las muertes por cualquier causa, los efectos graves no deseados, los ataques cardíacos no mortales o los ictus. No fue posible detectar efectos beneficiosos ni perjudiciales claros de la pioglitazona en estos desenlaces. Ninguno de los estudios incluidos informó sobre la calidad de vida relacionada con la salud o los efectos socioeconómicos (como los costos de la intervención, la ausencia al trabajo, el consumo de medicamentos).

Se encontraron dos estudios en curso que potencialmente se podrían incluir en esta revisión. Estos estudios pueden contribuir con datos de alrededor de 2694 participantes en futuras actualizaciones de esta revisión.

Los estudios de investigación futuros se deberían centrar en si el efecto de la pioglitazona se mantiene después de que las personas dejan de tomarla. Además, los estudios de investigación se deberían centrar en los desenlaces importantes para el paciente, como los efectos no deseados y las complicaciones de la diabetes.

Calidad de la evidencia

Todos los estudios tuvieron problemas en sus métodos o en la forma en que informaron los resultados. Además, muchos desenlaces no se informaron en ningún estudio o sólo se informaron en unos pocos. Por lo tanto, no se sabe con certeza si la pioglitazona previene o retrasa la diabetes tipo 2 en las personas con riesgo de desarrollar la enfermedad.

Authors' conclusions

Implications for practice

There is low‐certainty evidence that pioglitazone compared with placebo reduces or delays the development of type 2 diabetes mellitus (T2DM) in people with increased risk of T2DM (risk ratio (RR) 0.40, 95% confidence interval (CI) 0.17 to 0.95); 95% prediction interval ranging between 0.03 and 4.68). There is moderate‐certainty evidence that pioglitazone compared with no intervention reduces or delays the development of T2DM in people with increased risk of T2DM (RR 0.31, 95% CI 0.23 to 0.40; 95% prediction interval ranging between 0.23 and 0.41). There is no advantage or disadvantage of pioglitazone treatment compared with metformin on the incidence of T2DM in people with increased risk of T2DM (RR 0.98, 95% CI 0.40 to 2.38; low‐certainty evidence).

Data and reporting of serious adverse events, micro‐ and macrovascular complications and mortality were generally sparse. None of the included studies reported on health‐related quality of life or socioeconomic effects of pioglitazone intervention.

Most studies applied behaviour‐changing interventions in the intervention and comparator groups. Most studies identified people at increased risk of T2DM by impaired fasting glucose (IFG) and/or impaired glucose tolerance (IGT). None of the included studies identified increased risk of T2DM by elevated glycosylated haemoglobin (HbA1c), and the effect in this population is unknown.

Implications for research

We are uncertain whether the beneficial effects of pioglitazone on the incidence of T2DM when compared to placebo or no intervention are sustained after discontinuation of the drug. Additionally, the effects of pioglitazone treatment on diabetes‐specific complications are unclear. There are no data on the effects of pioglitazone on health‐related quality of life and socioeconomic effects in people with increased risk of developing T2DM. Future studies should focus on these factors.

Summary of findings

Open in table viewer

Summary of findings 1. Summary of findings: pioglitazone monotherapy versus another pharmacological glucose‐lowering monotherapy

Outcomes	Acarbose, metformin, repaglinide	Pioglitazone	Relative effect (95% CI)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Pioglitazone monotherapy versus another pharmacological glucose‐lowering monotherapy (acarbose, metformin, repaglinide)
Population: people at increased risk of developing T2DM Settings: outpatients Intervention: pioglitazone Comparison: acarbose, metformin, repaglinide
All‐cause mortality
Metformin	Not reported
Acarbose	Not reported
Repaglinide	Not reported
Incidence of T2DM (N)
Metformin Diagnostic criteria: 2 studies applied the WHO 1999 criteria for 'prediabetes' (FPG > 6.1 mmol/L and < 7.0 mmol/L and/or a 2‐h glucose < 11.1 mmol/L after a 75 g OGTT; Chen 2007a; Zeng 2013) 1 study applied the ADA 2010 criteria (FPG ≥ 5.6 mmol/L and < 6.9 mmol/L and 2‐h glucose ≥ 7.8 mmol/L or < 11.0 mmol/L after a 75 g OGTT; Zhang 2007) Follow‐up: 12‐24 months	55 per 1000	54 per 1000 (22 to 131)	RR 0.98 (0.40 to 2.38)	331 (3)	⊕⊕⊝⊝ Low^a	9/168 participants developed T2DM in the pioglitazone groups vs 9/163 participants in the metformin groups
Acarbose Diagnostic criteria: 1 study applied the WHO 1999 criteria for prediabetes (FPG > 6.1 mmol/L and < 7.0 mmol/L and/or a 2‐h glucose < 11.1 mmol/L after a 75 g OGTT; Chen 2007a) Follow‐up: 12 months	See comment			96 (1)	⊕⊝⊝⊝ Verylow^b	1/50 participants in the pioglitazone group vs 2/46 participants in the acarbose group developed T2DM (Chen 2007a)
Repaglinide Diagnostic criteria: 1 study applied the ADA 2010 criteria (FPG ≥ 5.6 mmol/L and < 6.9 mmol/L and 2‐h glucose ≥ 7.8 mmol/L or < 11.0 mmol/L after a 75 g OGTT; Zhang 2007) Follow‐up: 12 months	See comment			96 (1)	⊕⊕⊝⊝ Low^c	2/48 participants in the pioglitazone group vs 1/48 participants in the repaglinide group developed T2DM (Zhang 2007)
Serious adverse events (N)
Metformin Follow‐up: 12‐24 months	See comment			201 (2)	⊕⊕⊝⊝ Low^a	No participant experienced a serious adverse event (Chen 2007a; Zhang 2007)
Acarbose Follow‐up: 12 months	See comment			96 (1)	⊕⊝⊝⊝ Verylow^b	No participant experienced a serious adverse event (Chen 2007a)
Repaglinide Follow‐up: 12 months	See comment			103 (1)	⊕⊕⊝⊝ Low^c	No participant experienced a serious adverse event (Zhang 2007)
Cardiovascular mortality
Metformin	Not reported
Acarbose	Not reported
Repaglinide	Not reported
Non‐fatal myocardial infarction/stroke
Metformin	Not reported
Acarbose	Not reported
Repaglinide	Not reported
Health‐related quality of life
Metformin	Not reported
Acarbose	Not reported
Repaglinide	Not reported
Socioeconomic effects
Metformin	Not reported
Acarbose	Not reported
Repaglinide	Not reported
ADA: American Diabetes Association; CI: confidence interval; FPG: fasting plasma glucose; OGTT: oral glucose tolerance test; RR: risk ratio; HbA1c: glycosylated haemoglobin A1c; T2DM: type 2 diabetes mellitus; WHO: World Health Organization
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded by two levels because of serious imprecision (low median sample size, small number of studies), see Appendix 1. ^bDowngraded by one level because of attrition bias and by two levels because of serious imprecision (one study only, low median sample size), see Appendix 2. ^cDowngraded by two levels because of serious imprecision (one study only, low median sample size), see Appendix 3.

Open in table viewer

Summary of findings 2. Summary of findings: pioglitazone monotherapy versus intensive behaviour‐changing intervention

Outcomes	Personalised diet + exercise consultation	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Pioglitazone monotherapy versus intensive behaviour‐changing intervention
Population: people at increased risk for developing T2DM Settings: outpatients Intervention: pioglitazone Comparison: intensive diet plus exercise intervention
All‐cause mortality	Not reported
Incidence of T2DM (N) Follow‐up: 12 months	See comment	96^a (1)	⊕⊕⊝⊝ Low^b	2/48 participants developed T2DM in the pioglitazone group vs 5/48 participants in the personalised diet and exercise consultation group (Zhang 2007)
Serious adverse events (N)	See comment	103^a (1)	⊕⊕⊝⊝ Low^b	No participant experienced a serious adverse event (Zhang 2007)
Cardiovascular mortality	Not reported
Non‐fatal myocardial infarction/stroke	Not reported
Health‐related quality of life	Not reported
Socioeconomic effects	Not reported
CI: confidence interval; T2DM: type 2 diabetes mellitus
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
^aOnly people finishing the study were included in analysis of incidence of type 2 diabetes mellitus. All randomised participants were included in the remaining analyses. ^bDowngraded by two levels because of serious imprecision (one study only, low sample size), see Appendix 4.

Open in table viewer

Summary of findings 3. Summary of findings: pioglitazone monotherapy versus placebo

Outcomes	Placebo	Pioglitazone	Relative effect (95% CI)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Pioglitazone monotherapy versus placebo (identical behaviour‐changing interventions in both groups)
Population: people at increased risk for developing T2DM Settings: outpatients Intervention: pioglitazone Comparison: placebo
All‐cause mortality (N) Follow‐up: 44 weeks to 2 years	3 per 1000	8 per 1000 (2 to 36)	Peto OR2.38 (0.54 to 10.50)	1156 (4)	⊕⊝⊝⊝ Very low^a	Only 2 studies reported events (ACT NOW; IDPP‐2)
Incidence of T2DM (N) Diagnostic criteria: 3 studies applied the WHO 1999 criteria for 'prediabetes' (FPG > 6.1 mmol/L and < 7.0 mmol/L and/or a 2‐h glucose < 11.1 mmol/L after a 75 g OGTT; Attallah 2007; IDPP‐2; Xu 2011) 1 study applied the ADA 2003 criteria (FPG ≥ 5.3 mmol/L and < 6.9 mmol/L or 2‐h glucose ≥ 7.8 mmol/L and < 11.0 mmol/L after a 75 g OGTT; Bone 2013) 1 study applied the ADA 2010 criteria (FPG ≥ 5.6 mmol/L and < 6.9 mmol/L and 2‐h glucose ≥ 7.8 mmol/L or < 11.0 mmol/L after a 75 g OGTT; Shi 2014) 1 study applied its own definition of 'prediabetes' (FPG = 95‐125 mg/dL and 2‐h glucose ≥ 140 mg/dL or < 199 mg/dL after a 75 g OGTT or FPG = 90‐125 mg/dL and 2‐h glucose of 170‐199 mg/dL plus at least one additional high risk characteristic of diabetes; ACT NOW) Follow‐up: 44 weeks to 2 years	188 per 1000	75 per 1000 (32 to 179)	RR 0.40 (0.17 to 0.95)	1395 (6)	⊕⊕⊝⊝ Low^b	The 95% prediction interval ranged from 0.03 to 4.68 Three studies reported the incidence of T2DM after extended follow‐up periods ranging between 3 weeks and 11.4 months. Data for T2DM at the end of the extension periods showed that the intervention effect was dismissed RR 0.89, 95% CI 0.49 to 1.60 (ACT NOW; Attallah 2007; Bone 2013)
Serious adverse events (N) Follow‐up: 44 weeks to 2 years	See comment		RR3.00 (0.32 to 28.22)	187 (2)	⊕⊝⊝⊝ Very low^c	Only Bone 2013 observed events: 3/93 participants experienced serious adverse events in the pioglitazone group vs 1/94 participants in the placebo group. The largest study for this comparison reported on adverse events without distinguishing between serious and non‐serious events (ACT NOW). 121/303 (39.9%) participants in the pioglitazone group compared with 151/299 (50.5%) participants in the placebo group experienced an adverse event (P = 0.03; ACT NOW)
Cardiovascular mortality (N) Follow‐up: 44 weeks to 2 years	See comment		RR 5.14 (0.25 to 106.28)	1156 (4)	⊕⊝⊝⊝ Very low^d	Only IDPP‐2 observed events: 2/181 participants in the pioglitazone group vs 0/186 participants in the placebo group 3 studies reported no cardiovascular deaths
Non‐fatal myocardial infarction/stroke (N) Follow‐up: 44 weeks to 2 years	See comment		RR 1.97 (0.18 to 21.65)	789 (3)	⊕⊝⊝⊝ very low^a	Only ACT NOW observed non‐fatal myocardial infarction events: 2/303 participants in the pioglitazone group vs 1/299 participants in the placebo group 2 studies did not observe non‐fatal myocardial infarctions Non‐fatal strokes were not observed
Health‐related quality of life	Not reported
Socioeconomic effects	Not reported
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). ADA: American Diabetes Association; CI: confidence interval;FPG: fasting plasma glucose; OGTT: oral glucose tolerance test; OR: odds ratio; RR: risk ratio; T2DM: type 2 diabetes mellitus; WHO: World Health Organization
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded by one level because of attrition bias and indirectness (insufficient time frame) and by two levels because of serious imprecision (95% confidence interval consistent with benefit and harm, small number of studies), see Appendix 5. ^bDowngraded by one level because of attrition bias and by one level because of imprecision (low median sample size), see Appendix 5. ^cDowngraded by one level because of attrition bias and by two levels because of severe imprecision (95% confidence interval consistent with benefit and harm, low median sample size, small number of studies), see Appendix 5.

Open in table viewer

Summary of findings 4. Summary of findings: pioglitazone monotherapy versus no intervention

Outcomes	No intervention	Pioglitazone	Relative effect (95% CI)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Pioglitazone as monotherapy versus no intervention (identical behaviour‐changing interventions in both groups)
Population: people at increased risk of developing T2DM Settings: outpatients (20 studies), inpatients (1 study) Intervention: pioglitazone Comparison: no intervention
All‐cause mortality (N) Follow‐up: 24 weeks to mean of 2.2 years	28 per 1000	24 per 1000 (11 to 54)	RR 0.85 (0.38 to 1.91)	866 (3)	⊕⊝⊝⊝ Verylow^a
Incidence of T2DM (N) Diagnostic criteria: 9 studies applied the WHO 1999 criteria for 'prediabetes' (FPG > 6.1 mmol/L and < 7.0 mmol/L and/or a 2‐h glucose < 11.1 mmol/L after a 75 g OGTT; Che 2014; Chen 2007a; Gao 2011; Ke 2006; Liang 2004; Xiu 2015; Yi 2015; Yu 2011; Zeng 2013) 1 study applied the ADA 2003 criteria (FPG ≥ 5.3 mmol/L and < 6.9 mmol/L or 2‐h glucose ≥ 7.8 mmol/L and < 11.0 mmol/L after a 75 g OGTT; Fang 2013) 4 studies applied the ADA 2010 criteria (FPG ≥ 5.6 mmol/L and < 6.9 mmol/L and 2‐h glucose ≥ 7.8 mmol/L or < 11.0 mmol/L after a 75 g OGTT; Deng 2013; Wu 2013; Zhang 2007; Zhang 2015) 2 studies did not specify their diagnostic criteria (Chen 2007b; Li 2017) Follow‐up: 24 weeks to a mean of 2.2 years	193 per 1000	60 per 1000 (44 to 77)	RR 0.31 (0.23 to 0.40)	2053 (16)	⊕⊕⊕⊝ Moderate^b	The 95% prediction interval ranged between 0.23 and 0.41
Serious adverse events (N) Follow‐up: 6 months to a mean of 2.2 years	35 per 1000	25 per 1000 (13 to 46)	RR 0.71 (0.38 to 1.32)	1211 (7)	⊕⊕⊝⊝ Low^c	The 95% prediction interval ranged between 0.32 and 1.60
Cardiovascular mortality	Not reported
Non‐fatal myocardial infarction/stroke	Not reported
Health‐related quality of life	Not reported
Socioeconomic effects	Not reported
ADA: American Diabetes Association; CI: confidence interval;FPG: fasting plasma glucose; OGTT: oral glucose tolerance test; RR: risk ratio; T2DM: type 2 diabetes mellitus; WHO: World Health Organization
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded by one level because of indirectness (insufficient time frame) and by two levels because of serious imprecision (small sample size, small number of studies, 95% confidence interval consistent with benefit and harm), see Appendix 6. ^bDowngraded by one level because of imprecision (low median sample size) and unclear selection bias. ^cDowngraded by two levels because of imprecision (95% confidence interval consistent with benefit and harm, low median sample size).

Background

Description of the condition

'Prediabetes', 'borderline diabetes', the 'prediabetic stage', 'high risk of diabetes' or 'intermediate hyperglycaemia' (WHO/IDF 2006), are often characterised by various measurements of elevated blood glucose concentrations (such as isolated impaired fasting glucose (IFG), isolated impaired glucose tolerance (IGT), isolated elevated glycosylated haemoglobin A1c (HbA1c) or combinations thereof). These elevated blood glucose levels indicating hyperglycaemia are considered too high to be normal but below the diagnostic threshold for type 2 diabetes mellitus (T2DM). Therefore, because of the continuous spectrum from the normal to the diabetic stage, a sound evidence base is needed to define thresholds for conditions of 'sub‐diabetes'. It is obvious that the different terms used to describe various stages of hyperglycaemia might induce different emotional reactions. For example, the term 'prediabetes' may imply (at least for the lay person) that diabetes is unavoidable whereas (high) risk of diabetes has the positive connotation that the disease may be avoided altogether. We will use all of the above‐mentioned terms throughout this systematic review, however we will focus on 'prediabetes' because many people associate this label with dire consequences ‐ despite the disputable construct of intermediate health states termed prediseases (Viera 2011). On the other hand, any diagnosis of 'prediabetes' might be an opportunity to review, for example, eating habits and physical activity levels, thus enabling affected individuals to actively change their way of life.

The most commonly used criteria to define people with a high risk of developing T2DM were established by the American Diabetes Association (ADA) and the World Health Organization (WHO). The first glycaemic measurement used to define the prediabetic stage by the US National Diabetes Data Group was IGT (NDDG 1979). IGT is based on the measurement of plasma glucose two hours after ingestion of 75 g glucose. The prediabetic range is defined as a plasma glucose level between 7.8 mmol/L and 11.1 mmol/L (140 mg/dL to 200 mg/dL) two hours after the glucose load. Studies have indicated that IGT is caused by insulin resistance and defective insulin secretion (Abdul‐Ghani 2006). In 1997 the ADA and later on the WHO introduced the IFG concept to define 'prediabetes' (ADA 1997; WHO 1999). The initial definition of IFG was a fasting plasma glucose level of 6.1 mmol/L to 6.9 mmol/L (110 mg/dL to 125 mg/dL). Later on, the ADA reduced the lower threshold for defining IFG to 5.6 mmol/L (100 mg/dL) (ADA 2003). However, this lower cut‐off point for IFG to define 'prediabetes' was not endorsed by the WHO (WHO/IDF 2006). IFG seems to be associated with ß‐cell dysfunction (impaired insulin secretion) and an increase of the hepatic glucose output (DeFronzo 1989). More recently, HbA1c has been introduced for identifying people with a high risk of developing T2DM. In 2009, the International Expert Committee (IEC) suggested the HbA1c to identify people with a high risk of T2DM. People with HbA1c measurements between 6.0% and 6.4% fulfilled this criterion (IEC 2009). Shortly after, the ADA redefined this HbA1c level as 5.7% to 6.4% to identify people with a high risk of developing T2DM (ADA 2010). Unlike IFG and IGT, HbA1c reflects longer‐term glycaemic control, that is, how the blood glucose levels have been during the previous two to three months (Inzucchi 2012).

In 2010, the International Diabetes Federation (IDF) estimated the prevalence of IGT to be 343 million, and this number is predicted to increase to 471 million by 2035 (IDF 2013). Studies have shown poor correlations between HbA1c and IFG/IGT (Gosmanov 2014; Selvin 2011). Moreover, the various glycaemic tests do not seem to identify the same people (Gosmanov 2014; Selvin 2011). The risk of progression from 'prediabetes' to T2DM depends on the diagnostic criteria used to identify 'prediabetes' (Richter 2018). Some people diagnosed with 'prediabetes' will never develop T2DM, and some will return to normoglycaemia (Richter 2018). IFG and HbA1c are both thought to predict a different risk spectrum for developing T2DM (Cheng 2006; Morris 2013). Most importantly, dysglycaemia is commonly an asymptomatic condition, and naturally often remains 'undiagnosed' (CDC 2015).

Currently, ADA recommends reduced calorie intake and increased physical activity for people with increased risk of T2DM (ADA 2017). It is still not clarified if any particular intervention, especially glucose‐lowering drugs, should be recommended for people with 'prediabetes' (Yudkin 2014). Studies have indicated that the progression from 'prediabetes' to T2DM is reduced, or maybe just delayed with 'lifestyle' interventions (usually increased physical activity, dietary changes or both) (Diabetes Prevention Program 2002; Diabetes Prevention Program FU 2009; Finnish Diabetes Prevention Study Group 2001). A recent meta‐analysis of 22 studies with behaviour‐changing interventions in people at high risk of T2DM concluded that the effect of behaviour‐changing interventions on longer‐term diabetes prevention had not been clarified (Dunkley 2014).

The prescription of pharmacological glucose‐lowering interventions for the prevention of T2DM is not generally accepted among international diabetes associations and clinicians. Several groups of pharmacological glucose‐lowering interventions have been investigated in people with 'prediabetes'. Some findings indicate that the progression from 'prediabetes' to T2DM is reduced or maybe just delayed (Diabetes Prevention Program 2002; Diabetes Prevention Program FU 2009). However, the ADA recommends metformin for people with 'prediabetes' and a body mass index (BMI) more than 35 kg/m², aged less than 60 years, and for women with prior gestational diabetes mellitus (ADA 2015). Several Cochrane Reviews have not shown prevention of T2DM after use of insulin secretagogues, sodium‐glucose cotransporter (SGLT) 2 inhibitors, dipeptidyl‐peptidase (DDP) 4 inhibitors, and glucagon‐like peptide (GLP) 1 (Hemmingsen 2016a; Hemmingsen 2016b; Hemmingsen 2017).

Description of the intervention

Glucose‐lowering drugs are most often used to treat T2DM in its initial stages if behaviour‐changing modifications have failed. Pioglitazone affects many tissues and parts of the body. For people with T2DM, pioglitazone can be prescribed as monotherapy or in combination with existing glucose‐lowering interventions (Davies 2019). Currently, pioglitazone is not recommended for people with intermediate hyperglycaemia. The cost of pioglitazone is low in most countries (Davies 2019).

Pioglitazone is a member of the thiazolidinedione group, which also encompasses troglitazone (withdrawn due to hepatic toxicity) and rosiglitazone (withdrawn from several markets due to suspected cardiovascular risk) (Babai 2018; Cohen 2010).

Adverse effects of the intervention

The most common adverse effects of pioglitazone are heart failure, bone fracture, oedema and weight gain (Liao 2017; Richter 2006).

How the intervention might work

Pioglitazone is known to improve insulin sensitivity, glycaemic control, hypertension and dyslipidaemia in people with T2DM (Schernthaner 2013). Pioglitazone decreases fasting and postprandial plasma glucose levels by improving the sensitivity of hepatic and peripheral (muscle) tissue to insulin (Schernthaner 2013). By improving insulin sensitivity, thiazolidinediones may exert beneficial effects on cardiovascular risk factors (Schernthaner 2013).

Pioglitazone exerts its action through a selective and potent inhibition of peroxisome proliferator‐activated receptor (PPAR)‐gamma (Ninomiya 2016). PPARs are transcription factors that reside in the nucleus and become activated by ligands such as thiazolidinediones. Thiazolidinediones enter the cell, bind to the nuclear receptors, and alter the expression of genes. There are several PPARs including PPAR‐alpha, PPAR‐beta/delta, and PPAR‐gamma. PPAR‐gamma receptors are present in adipose, hepatic and skeletal muscle tissue and control insulin‐responsive genes, which have a wide‐ranging influence (Ninomiya 2016).

Why it is important to do this review

There has been an increased focus on the prevention or delay of T2DM with non‐pharmacological interventions and glucose‐lowering medications. Currently, several studies are being conducted to clarify whether the progression from an at‐risk status to T2DM can be stopped or postponed with glucose‐lowering agents (ClinicalTrials.gov). However, a more important issue for people with dysglycaemia is whether the intervention reduces the risk of death or complications ‐ especially cardiovascular disease ‐ related to T2DM.

One systematic review evaluated the effects of pioglitazone in people with T2DM, insulin resistance and intermediate hyperglycaemia. However, the investigators did not search for unpublished data nor apply rigorous Cochrane methodology (Liao 2017). This review concluded that pioglitazone was associated with a reduced risk of macrovascular complications (all types of participants combined). However, the risks of heart failure, bone fracture, oedema and weight gain were increased (Liao 2017).

Objectives

To assess the effects of pioglitazone for prevention or delay of T2DM and its associated complications in people at risk of developing T2DM.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs).

Types of participants

We included nondiabetic individuals at risk of developing T2DM, that is, diagnosed with intermediate hyperglycaemia or 'prediabetes'.

Diagnostic criteria for 'prediabetes'

To be consistent with changes in the classification of and diagnostic criteria for 'prediabetes' (IFG, IGT and elevated HbA1c) over the years, the diagnosis should have been established using the standard criteria valid at the time the study started (for example ADA 1997; ADA 2010; NDDG 1979; WHO 1999). Ideally, the diagnostic criteria should be described. If necessary, we used the study authors' definition of 'prediabetes'. Differences of glycaemic measurements used to define 'prediabetes' may introduce substantial heterogeneity. We therefore subjected diagnostic criteria to a subgroup analysis.

Types of interventions

We planned to investigate the following comparisons of intervention versus control/comparator intervention.

Intervention

Pioglitazone as monotherapy
Pioglitazone as a part of a dual combination therapy

Comparisons

Any pharmacological glucose‐lowering monotherapy intervention (e.g. acarbose, metformin, sulphonylurea) compared with pioglitazone as monotherapy
Behaviour‐changing interventions (e.g. diet, exercise, diet and exercise) compared with pioglitazone as monotherapy
Placebo compared with pioglitazone as monotherapy
No intervention compared with pioglitazone as monotherapy
Any pharmacological glucose‐lowering agent (e.g. acarbose, metformin, sulphonylurea) compared with pioglitazone as a part of a dual combination therapy. This glucose‐lowering agent had to be identical in both the intervention and comparator groups (e.g. pioglitazone plus metformin versus metformin or pioglitazone plus acarbose versus acarbose plus metformin).

Concomitant interventions (e.g. educational programmes or additional pharmacotherapy) had to be identical in both the intervention and comparator groups to establish fair comparisons.

For studies including multiple arms, we included any arms that met the review inclusion criteria.

Minimum duration of intervention

We included studies with a duration of the intervention of 24 weeks or more.

Minimum duration of follow‐up

The minimum duration of follow‐up was 24 weeks (end of the intervention period).

We defined any follow‐up period going beyond the original time frame for the primary outcome measure as specified in the power calculation of the study's protocol as an extended follow‐up period, also called 'open‐label extension study' (Buch 2011; Megan 2012).

Summary of specific exclusion criteria

We excluded studies of people exclusively diagnosed with 'metabolic syndrome' because this is a special population that is not representative of people with only intermediate hyperglycaemia. Also, the composite of risk indicators such as elevated blood lipids, insulin resistance, obesity and hypertension, which is termed 'metabolic syndrome' is of doubtful clinical usefulness and uncertain distinct disease entity.
We excluded studies evaluating participants with intermediate hyperglycaemia in combination with another condition (e.g. cystic fibrosis) except for hypertension.
We excluded studies evaluating participants with intermediate hyperglycaemia because of other medical interventions (e.g. glucocorticoids).

We included studies in obese people and participants with previous gestational diabetes, if study investigators described that the participants had intermediate hyperglycaemia. We planned to include studies that did not report one or more of our primary or secondary outcome measures in the publication. If a study had been included this way, we planned to contact the corresponding study author for supplementary data.

Types of outcome measures

Primary outcomes

All‐cause mortality
Incidence of T2DM
Serious adverse events

Secondary outcomes

Cardiovascular mortality
Non‐fatal myocardial infarction
Non‐fatal stroke
Congestive heart failure
Amputation of lower extremity
Blindness or severe vision loss
End‐stage renal disease
Non‐serious adverse events
Hypoglycaemia
Health‐related quality of life
Time to progression to T2DM
Measures of blood glucose control
Socioeconomic effects

Method of outcome measurement

All‐cause mortality: defined as death from any cause
Incidence of T2DM and time to progression to T2DM: defined according to diagnostic criteria valid at the time the diagnosis was established, using the standard criteria valid at the time the study started (e.g. ADA 2008; WHO 1998). If necessary, we used the study authors' definition of T2DM.
Serious adverse events: defined according to the International Conference on Harmonisation Guidelines as any event that led to death, that was life‐threatening, required in‐patient hospitalisation or prolongation of existing hospitalisation, resulted in persistent or significant disability, and any important medical event that may have had jeopardised the participant or required intervention to prevent it, or as reported in studies (ICH 1997).
Cardiovascular mortality, non‐fatal myocardial infarction, non‐fatal stroke, congestive heart failure, amputation of lower extremity, blindness or severe vision loss, hypoglycaemia (mild, moderate, severe/serious): defined as reported in studies
End‐stage renal disease: defined as dialysis, renal transplantation or death due to renal disease
Non‐serious adverse events: defined as number of participants with any untoward medical occurrence not necessarily having a causal relationship with the intervention
Health‐related quality of life: defined as mental and physical health‐related quality of life, separate and combined, evaluated by a validated instrument such as Short‐Form 36 (SF‐36)
Measures of blood glucose control: fasting blood glucose, blood glucose two hours after ingestion of 75 g glucose and HbA1c measurements
Socioeconomic effects: for example, costs of the intervention, absence from work, medication consumption

Timing of outcome measurement

Measured at the end of the intervention and the end of follow‐up: all‐cause mortality, cardiovascular mortality, non‐fatal myocardial infarction, non‐fatal stroke, congestive heart failure, amputation of lower extremity, blindness or severe vision loss, hypoglycaemia (mild, moderate, severe/serious), health‐related quality of life, measures of blood glucose control, socioeconomic effects
Measured at the end of the intervention and the longest reported end of follow‐up: incidence of T2DM
Measured at any time of the intervention and during follow‐up: serious adverse events, non‐serious adverse events, hypoglycaemia (mild, moderate, severe/serious)

Search methods for identification of studies

Electronic searches

We searched the following sources from the inception of each database to the date of search and placed no restrictions on the language of publication:

Cochrane Central Register of Controlled Trials (CENTRAL; 2019, Issue 11) via the Cochrane Register of Studies Online (CRSO), searched 12 November 2019;
MEDLINE (Ovid MEDLINE ALL 1946 to Daily Update) searched 12 November 2019;
ClinicalTrials.gov (www.clinicaltrials.gov) searched 12 November 2019;
World Health Organization International Clinical Trials Registry Platform (ICTRP; www.who.int/trialsearch/) searched 12 November 2019;
China Network Knowledge Infrastructure (CNKI; cnki.net) searched 4 March 2020;
Chinese Scientific Journals Database (VIP; cqvip.com) searched 4 March 2020;
Wan Fang data (wanfangdata.com.cn/index.html) searched 4 March 2020;
SinoMed (sinomed.ac.cn) searched 4 March 2020.

For detailed search strategies, see Appendix 7.

We did not include Embase in our search, as RCTs indexed in Embase are now prospectively added to CENTRAL via a highly sensitive screening process (Cochrane 2018).

Searching other resources

We tried to identify other potentially eligible studies or ancillary publications by searching the reference lists of included studies, systematic reviews, meta‐analyses, and health technology assessment reports. In addition, we contacted the authors of included studies to obtain additional information on the retrieved studies and establish whether we may have missed further studies.

We also searched databases from regulatory agencies (European Medicines Agency (EMA) and US Food and Drugs Administration (FDA); (Hart 2012; Schroll 2015)).

We did not use abstracts or conference proceedings for data extraction unless full data were available from study authors because this information source does not fulfil the CONSORT requirements, which consists of, "an evidence‐based, minimum set of recommendations for reporting randomized trials" (CONSORT 2018). We defined grey literature as records detected in ClinicalTrials.gov, WHO ICTRP or databases from regulatory agencies.

Data collection and analysis

Selection of studies

Two review authors (EI, BH) independently screened the abstract, title, or both, of every record retrieved by the literature searches, to determine which studies should be further assessed. We obtained the full text of all potentially relevant records. We resolved disagreements through consensus or by recourse to a third review author (KM). If we could not resolve a disagreement, we planned to categorise the study as a 'Study awaiting classification' and contacted the study authors for clarification. We made an adapted PRISMA flow diagram to show the process of study selection (Liberati 2009). All articles excluded after full‐text assessment are listed with reason for exclusion in Characteristics of excluded studies.

Data extraction and management

For English language studies that fulfilled our inclusion criteria, two review authors (EI, KM) independently extracted key participant and intervention characteristics. For Chinese language studies that fulfilled our inclusion criteria, two review authors (EI, YC) independently extracted key participant and intervention characteristics. We described interventions according to the 'template for intervention description and replication' (TIDieR) checklist (Hoffmann 2014; Hoffmann 2017).

We reported data on efficacy outcomes and adverse events using standardised data extraction sheets from Cochrane Metabolic and Endocrine Disorders (CMED). We resolved disagreements by discussion or, if required, by consultation with a third review author (BH) (for details see Characteristics of included studies; Table 1; Appendix 8; Appendix 9; Appendix 10; Appendix 11; Appendix 12; Appendix 13; Appendix 14; Appendix 15; Appendix 16; Appendix 17; Appendix 18; Appendix 19; Appendix 20; Appendix 21; Appendix 1; Appendix 2; Appendix 3; Appendix 4; Appendix 5; Appendix 6.

Open in table viewer

Table 1. Overview of study populations

Study ID (study design)	Intervention(s) and comparator(s)	Description of power and sample size calculation	Screened/eligible (N)	Randomised (N)	Analysed (primary outcome) (N)	Finishing study (N)	Randomised finishing study (%)	Follow‐up (extended follow‐up)^a
ACT NOW (parallel RCT)	I: pioglitazone 45 mg/day titrated up from 30 mg/day after 1 month if no adverse effects plus BCI	"Based upon this information (see assumptions below), it can be calculated that approximately 600 subjects with IGT will be required to achieve 90% statistical power that pioglitazone decreases the conversion rate of IGT to type 2 diabetes by 50%. This power calculation assumes that randomized individuals drop out prior to the confirmed diagnosis of diabetes with an exponential hazard rate of 0.10 (or less) per year. The following assumptions were used to calculate the sample size: (i) The primary endpoint is the development of diabetes, (ii) Participants are randomized over a 21 month period and followed for a total of 3.75 years, starting from the time that the first IGT subject is recruited, (iii) Type I error rate (alpha) is 0.05, (iv) The desired power is 90%, (v) The development of diabetes in the placebo‐treated group is 11% per year,(vi) The hazard rate for the development of diabetes in the pioglitazone‐treated group is reduced by 50%. (vii) The drop out rate is 10% per year"	1827/602	303	303	213	70.3	24‐48 months (extended follow‐up with median of 11.4 months)
	C: placebo once daily, placebo dose titrated up after 1 month if no adverse effects plus BCI		1827/602	299	299	228	76.3
	Total:			602	602	441	73.3
Attallah 2007 (2 x 2 factorial RCT)	I: pioglitazone 30 mg/day titrated up from 15 mg/day after 4 weeks plus GH placebo from 4‐40 weeks	"Sample size was determined using data from Johannsson et al. [19], in which nine months of GH resulted in a 17.9% ± 3.5% reduction in visceral fat along with an improvement in insulin sensitivity in abdominally obese men. Using this information, we determined that 12 participants were needed per group to have an 80% chance of detecting a change in visceral fat of at least 17.9%. However, since women were also included in the study and are known to respond less effectively to GH than men, we increased the sample size to 15."	185/81	22	15	15	68.1	43‐44 weeks
	C: pioglitazone placebo plus GH placebo from 4‐40 weeks		185/81	19	16	16	84.2
	Total:			41	31	31	75.6
Bone 2013 (parallel RCT)	I: pioglitazone 45 mg/day titrated up from 30 mg/day plus vitamin D and calcium	"The a priori sample size calculation projected that 75 women per treatment group would be required for the two‐sided 95% CI of the estimated between‐group treatment difference to be within the target precision limits of ± 2.3%, assuming a SD of 6.5% and dropout rate of 18%. In a post hoc power calculation using the observed SD of 3% and 60 women per group, it was determined that the study had greater than 90% power to detect a between group difference in percentage change in BMD from baseline to month 12 if the true difference were 1.7%	386/156	78	78	57	73.1	18 months
	C: matching placebo once daily plus vitamin D and calcium		386/156	78	78	61	76.9
	Total:			156	156	117	75
Che 2014 (parallel RCT)	I: L‐amlodipine 5 mg/day plus pioglitazone 30 mg/day	—	—	38	38	38^b	100^c	6 months
	C: L‐amlodipine 5 mg/day	—	—	35	35	35^b	100^c
	Total:			73	73	73^b	100^c
Chen 2007a (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	50	50	50	100	12 months
	C1: acarbose 150 mg/day plus BCI			50	46	46	92
	C2: metformin 750 mg/day plus BCI			50	46	46	92
	C3: BCI			50	50	50	100
	C4: no intervention			50	50	50	100
	Total:			250	242	242	96.8
Chen 2007b (parallel RCT)	I: pioglitazone 15 mg/day plus captopril 50 mg/day and Betaloc 50 mg/day	—	—	88	83	83	94.3	24 months
	C: captopril 50 mg/day and Betaloc 50 mg/day	—	—	80	73	73	91.3
	Total:			168	156	156	94.0
Deng 2013 (parallel RCT)	I: pioglitazone 30 mg/day plus BCI	—	—	303	303^b	303^b	100^b	Mean of 2.2 years
	C: BCI	—	—	302	302^b	302^b	100^b
	Total:			605	605^c	605^c	100^c
Fang 2013 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	60	60	55	91.7	12 months
	C: BCI	—	—	60	60	56	93.3
	Total:			120	120	111	92.5
Gao 2011 (parallel RCT)	I: pioglitazone 30 mg/day, captopril and other antihypertensive drug(s) plus BCI	—	—	50	50	48	96	18 months
	C: captopril and other antihypertensive drug(s) plus BCI	—	—	50	50	46	92
	Total:			100	100	94	94
Guo 2009 (parallel RCT)	I: pioglitazone 30 mg/day plus BCI	—	—	30	30	30^b	100^b	6 months
	C: BCI	—	—	30	30	30^b	100^b
	Total:			60	60	60^c	100^c
Guo 2010 (parallel RCT)	I: pioglitazone 30 mg/day plus BCI	—	—	32	32	32	100	12 months
	C: BCI	—	—	32	32	32	100
	total:			64	64	64	100
Han 2007 (parallel RCT)	I: pioglitazone 15 mg/day plus enalapril (dose not reported)	—	—	23	23	23	100	6 months
	C: enalapril (dose not reported)	—	—	20	20	20	100
	Total:			43	43	43	100
IDPP‐2 (parallel RCT)	I: pioglitazone 30 mg/day titrated up from 15 mg once daily plus BCI	"It was assumed that the cumulative incidence of diabetes in 3 years would be 40% in the control group with lifestyle modification and placebo and 25% in the group receiving lifestyle modification and pioglitazone. The sample sizes required in each of the two groups were 165 with a type 1 error of 5%, with 80% power. Higher numbers were recruited (204 in group A and 203 in group B) to allow for drop out. The intention‐to‐treat approach was used."	6589/882/407 (double screening procedure)	204	181	181	88.7	36 months
	C: placebo in matching doses plus BCI		6589/882/407 (double screening procedure)	203	186	186	91.6
	Total:			407	367	367	90.2
Ke 2006 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	32	30	30	93.8	12 months
	C: BCI	—	—	30	28	28	93.3
	Total:			62	58	58	93.5
Li 2017 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	75	75	75^b	100c	6 months
	C: BCI	—	—	75	75	75^b	100^c
	Total:			150	150	150^b	100^c
Liang 2004 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	30	29	29	96.7	12 months
	C: BCI	—	—	30	30	30	100
	Total:			60	59	59	98.3
Shi 2014 (parallel RCT)	I: pioglitazone 30 mg/day plus BCI	—	—	40	40	40^b	40^b	10 months
	C: placebo in matching doses plus BCI	—	—	40	40	40^b	40^b
	Total:			80	80	80^c	100^c
Tian 2015 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	36	36	36^b	100^b	24 weeks
	C: BCI	—	—	36	36	36^b	100^b
	Total:			72	72	72^c	100^c
Wu 2013 (parallel RCT)	I: pioglitazone 30 mg/day plus amlodipine 5 mg/day and valsartan 80 mg/day plus BCI	—	—	34	34	34^b	100^c	6 months
	C: amlodipine 5 mg/day and valsartan 80 mg/day plus BCI	—	—	33	33	33^b	100^c
	Total:			67	67	67^c	100^c
Xiu 2015 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	43	43	43	100	48 weeks
	C: BCI	—	—	43	43	43	100
	Total:			86	86	86	100
Xu 2011 (parallel RCT)	I: pioglitazone 30 mg/day plus BCI	—	—	60	60	60	100	48 weeks
	C: placebo in matching doses plus BCI	—	—	59	59	59	100
	Total:			119	119	119	100
Yi 2015 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	35	35	35	100	12 months
	C: BCI	—	—	35	35	35	100
	Total:			70	70	70	100
Yu 2011 (parallel RCT)	I: pioglitazone 4 mg/day plus BCI	—	—	60	60	60^b	100^c	6 months
	C: BCI	—	—	60	60	60^b	100^c
	Total:			120	120	120^b	100^c
Zeng 2013 (parallel RCT)	I1: pioglitazone 38 mg/day plus BCI	—	—	70	70	70	100	24 months
	I2: metformin 38 mg/day plus BCI			68	68	68	100
	C: BCI			66	66	66	100
	Total:			204	204	204	100
Zhang 2007 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	52	48	48	100	12 months
	C1: metformin 250 mg/day plus BCI			53	49	49	100
	C2: repaglinide 3 x 0.25 mg/day plus BCI			51	48	48	100
	C3: intensive BCI			51	48	48	100
	C4: BCI			56	54	54	100
	Total:			263	263	247	93.9
Zhang 2015 (parallel RCT)	I: pioglitazone 15 mg/day	—	—	26	26	26	100	6 months
	C: BCI	—	—	26	25	25	96.2
	Total:			52	51	51	98.1
Zhao 2009 (parallel RCT)	I: pioglitazone 15 mg/day	—	—	47	47	47^b	100^c	6 months
	C: BCI	—	—	45	45	45^b	100^c
	Total:			92	92	92^b	100^c

Grand total (all included studies)	All interventions			1921		NA^d
	All comparators			2265		NA^d
	All interventions and comparators			4186		NA^d
Total (studies reporting on dropouts)	All interventions			1258		1098	87.2
	All comparators			1765		1462	82.8
	All interventions and comparators			3023		2560	84.7
— denotes not reported BCI: behaviour‐changing intervention; C: comparator; GH: growth hormone; I: intervention; IGT: impaired glucose tolerance; NA: not applicable; RCT: randomised controlled trial; SD: standard deviation.

^aFollow‐up under randomised conditions until end of study (= duration of intervention + follow‐up post‐intervention or identical to duration of intervention); extended follow‐up refers to follow‐up of participants once the original study was terminated as specified in the power calculation.
^bStudies did not report on dropouts or number of participants finishing study (assumed value).
^cValues calculated include assumptions on finishing rates of studies not reporting on dropout or finishing rates.
^dNot all included studies reported on dropouts.

We provided information including study identifier for potentially relevant ongoing studies in the Characteristics of ongoing studies table and in Appendix 15, 'Matrix of study endpoint (publications and study documents)'. We tried to find the protocol for each included study and reported primary, secondary, and other outcomes in comparison with data in publications.

We emailed all authors with available contact information of included studies to enquire whether they would be willing to answer questions regarding their studies. We presented the results of this survey in Appendix 21. We thereafter sought relevant missing information on the study from the primary study author(s), if required.

Dealing with duplicate and companion publications

In the event of duplicate publications, companion documents, or multiple reports of a primary study, we maximised the information yield by collating all available data and used the most complete data set aggregated across all known publications. All duplicate publications, companion documents, multiple reports of a primary study, and trial documents of included studies (such as trial registry information) are listed as secondary references under the study ID of the included study. All duplicate publications, companion documents, multiple reports of a study, and trial documents of excluded studies (such as trial registry information) are listed as secondary references under the study ID of the excluded study.

Data from clinical trials registers

If data from included studies were available as study results in clinical trials registers, such as ClinicalTrials.gov or similar sources, we made full use of this information and extracted the data. If there was also a full publication of the study, we collated and critically appraised all available data. If an included study had been marked as a completed study in a clinical trials register but no additional information (study results, publication, or both) was available, we planned to add this study to the 'Characteristics of studies awaiting classification' table.

Assessment of risk of bias in included studies

Two review authors (EI, KM) independently assessed the risk of bias for each included English language study. Two review authors (EI, YC) independently assessed the risk of bias for each included Chinese language study. We resolved disagreements by consensus or by consulting a third review author (BH). In the case of disagreement, we planned to consult the remainder of the review author team and make a judgement based on consensus. If adequate information was unavailable from the publications, study protocols, or other sources, we contacted the study authors for more detail to request missing data on 'Risk of bias' items.

We used the Cochrane 'Risk of bias' assessment tool (Higgins 2019b), assigning assessments of low, high, or unclear risk of bias (for details, see Appendix 8; Appendix 9). We evaluated individual bias items as described in the Cochrane Handbook for Systematic Reviews of Interventions, according to the criteria and associated categorisations contained therein (Higgins 2019b).

Summary assessment of risk of bias

We generated a 'Risk of bias' graph and a 'Risk of bias' summary figure (Figure 2; Figure 3).

We distinguished between self‐reported, investigator‐assessed and adjudicated outcome measures.

We considered the following self‐reported outcomes.

Non‐serious adverse events
Hypoglycaemia, if reported by participants
Health‐related quality of life
Blood glucose control, if measured by study participants

We considered the following outcomes to be investigator‐assessed.

All‐cause mortality
Incidence of T2DM
Time to progression to T2DM
Serious adverse events
Cardiovascular mortality
Non‐fatal myocardial infarction
Non‐fatal stroke
Congestive heart failure
Amputation of lower extremity
Blindness or severe vision loss
End‐stage renal disease
Hypoglycaemia, if measured by study personnel
Blood glucose control, if measured by study personnel
Socioeconomic effects

Risk of bias for a study across outcomes

Some 'Risk of bias' domains, such as selection bias (sequence generation and allocation sequence concealment), affect risk of bias across all outcome measures in a study. In case of high risk of selection bias, we planned to mark all endpoints investigated in the associated study as being at high risk. Otherwise, we did not perform a summary assessment of the risk of bias across all outcomes for a study.

Risk of bias for an outcome within a study and across domains

We assessed the risk of bias for an outcome measure by including all entries relevant to that outcome (i.e. both study‐level entries and outcome‐specific entries). We considered low risk of bias to denote a low risk of bias for all key domains, unclear risk to denote an unclear risk of bias for one or more key domains and high risk to denote a high risk of bias for one or more key domains.

Risk of bias for an outcome across studies and across domains

These are the main summary assessments that we incorporated into our judgements about the certainty of the evidence in the 'Summary of findings' tables. We defined outcomes as at low risk of bias when most information came from studies at low risk of bias, unclear risk when most information came from studies at low or unclear risk of bias and high risk when a sufficient proportion of information came from studies at high risk of bias.

Measures of treatment effect

When at least two included studies were available for a comparison of a given outcome, we expressed dichotomous data as a risk ratio (RR) or a Peto odds ratio with 95% confidence intervals (CIs). For continuous outcomes measured on the same scale (e.g. weight loss in kg) we estimated the intervention effect using the mean difference (MD) with 95% CIs. For continuous outcomes that measure the same underlying concept (e.g. health‐related quality of life) but use different measurement scales, we planned to calculate the standardised mean difference (SMD).

Unit of analysis issues

We considered the level at which randomisation occurred, such as cross‐over studies, cluster‐randomised trials, and multiple observations for the same outcome. If more than one comparison from the same study was eligible for inclusion in the same meta‐analysis, we planned either to combine groups to create a single pair‐wise comparison, or to appropriately reduce the sample size so that the same participants did not contribute data to the meta‐analysis more than once (splitting the 'shared' group into two or more groups). Although the latter approach offers some solution for adjusting the precision of the comparison, it does not account for correlation arising from inclusion of the same set of participants in multiple comparisons (Higgins 2019a).

We planned to re‐analyse cluster‐RCTs that did not appropriately adjust for potential clustering of participants within clusters in their analyses. Variance of the intervention effects would have been inflated by a design effect. Calculation of a design effect involves estimation of an intra‐cluster correlation coefficient (ICC). We planned to obtain estimates of ICCs by contacting study authors, or by imputing ICC values using either estimates from other included studies that report ICCs or external estimates from empirical research (e.g. Bell 2013). We planned to examine the impact of clustering by performing sensitivity analyses. We did not include any cluster‐RCTs in this review.

Dealing with missing data

If possible, we obtained missing data from the authors of included studies and carefully evaluated important numerical data such as screened, randomly assigned participants, as well as intention‐to‐treat and as‐treated and per‐protocol populations. We investigated attrition rates (e.g. dropouts, losses to follow‐up, withdrawals), and we critically appraised issues concerning missing data and use of imputation methods (e.g. last observation carried forward).

For studies in which the standard deviation (SD) of the outcome was not available at follow‐up or we could not recreate it, we standardised by the mean of the pooled follow‐up SD from studies that reported this information.

When included studies did report means and SDs for outcomes, and we did not receive requested information from study authors, we planned to impute these values by estimating the mean and the variance from the median, the range and the size of the sample (Hozo 2005).

We investigated the impact of imputation on meta‐analyses by performing sensitivity analyses.

Assessment of heterogeneity

In the event of substantial clinical or methodological heterogeneity, we did not report study results as the pooled effect estimate in a meta‐analysis.

We identified heterogeneity (inconsistency) by visually inspecting the forest plots and by using a standard Chi² test with a significance level of α = 0.1 (Deeks 2019). In view of the low power of this test, we also considered the I² statistic, which quantifies inconsistency across studies, to assess the impact of heterogeneity on the meta‐analysis (Higgins 2002; Higgins 2003).

When we found heterogeneity, we attempted to determine possible reasons for this by examining individual study and subgroup characteristics.

Assessment of reporting biases

For outcomes with 10 or more included studies that investigated a particular outcome, we used funnel plots to assess small‐study effects.

Data synthesis

We only displayed a meta‐analysis if we judged participants, interventions, comparisons, and outcomes to be sufficiently similar to ensure an answer that is clinically meaningful. We planned primarily to summarise low risk of bias data using a random‐effects model (Wood 2008), this was not possible because of methodological bias judgements (see Characteristics of included studies). We interpreted random‐effects meta‐analyses with due consideration for the whole distribution of effects and presented a prediction interval (Borenstein 2017a; Borenstein 2017b; Higgins 2009). A prediction interval requires at least three studies to be included in the calculations and specifies a predicted range for the true treatment effect in an individual study (Riley 2013). For rare events such as event rates below 1%, we used the Peto odds ratio method, provided there was no substantial imbalance between intervention and comparator group sizes, and intervention effects were not exceptionally large. In addition, we performed statistical analyses according to the statistical guidelines presented in the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2019).

Subgroup analysis and investigation of heterogeneity

We expected the following characteristics to introduce clinical heterogeneity, and we carried out the following subgroup analyses including investigation of interactions (Altman 2003).

Studies with a long duration (four years and longer) versus studies with a shorter duration (less than four years)
Diagnostic 'prediabetes' criteria (IFG, IGT, elevated HbA1c)
Pioglitazone monotherapy versus pioglitazone dual combination therapy
Age, depending on data
Sex
Ethnicity, depending on data
Comorbid conditions, such as hypertension or obesity
Women with and without previous gestational diabetes mellitus

Sensitivity analysis

We performed sensitivity analyses to explore the influence of the following factors (when applicable) on effect sizes by restricting analysis to the following.

Published studies
Effect of risk of bias, as specified in the Assessment of risk of bias in included studies section
Very long or large studies to establish the extent to which they dominate the results
Use of the following filters: diagnostic criteria, imputation, language of publication, source of funding (industry versus other) or country

We also tested the robustness of results by repeating analyses using different measures of effect size (i.e. RR, odds ratio, etc.) and different statistical models (fixed‐effect and random‐effects models).

Summary of findings and assessment of the certainty of the evidence

We presented the overall certainty for each outcome specified below, according to the GRADE approach, which took into account issues related to internal validity (risk of bias, inconsistency, imprecision, publication bias) and also to external validity, such as directness of results. Two review authors (BH, BR) independently rated the certainty of the evidence for each outcome. We resolved differences in assessment by discussion or by consultation with a third review author (KM).

We included Appendix 1; Appendix 2; Appendix 3; Appendix 4; Appendix 5 and Appendix 6 entitled 'Checklist to aid consistency and reproducibility of GRADE assessments', to help with standardisation of the 'Summary of findings' tables (Meader 2014). Alternatively, we planned to use the GRADEpro GDT software and would have presented evidence profile tables as an appendix (GRADEproGDT). We presented results for outcomes as described in the Types of outcome measures section. When meta‐analysis was not possible, we presented the results in a narrative format in the 'Summary of findings' table. We justified all decisions to downgrade the certainty of the evidence using footnotes, and we made comments to aid the reader's understanding of the Cochrane Review when necessary.

'Summary of findings' table

We presented a summary of the evidence in summary of findings Table 1; summary of findings Table 2; summary of findings Table 3; summary of findings Table 4. This provides key information about the best estimate of the magnitude of effect, in relative terms and as absolute differences for each relevant comparison of alternative management strategies, numbers of participants and studies addressing each important outcome, and a rating of overall confidence in effect estimates for each outcome. We created the 'Summary of findings' tables using the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2019), along with Review Manager 5 software (Review Manager 2020).

We presented 'Summary of findings' tables for pioglitazone as monotherapy versus the comparator interventions: other glucose‐lowering drugs; behaviour‐changing interventions; placebo; and no intervention. We planned to present in the 'Summary of findings' tables the interventions pioglitazone as part of a dual combination therapy but we did not include any studies with this comparison.

We reported the following outcomes, listed according to priority.

All‐cause mortality
Incidence of T2DM
Serious adverse events
Cardiovascular mortality
Non‐fatal myocardial infarction/stroke
Health‐related quality of life
Socioeconomic effects

Results

Description of studies

For a detailed description of studies, see Table 1, 'Characteristics of included studies', 'Characteristics of excluded studies, and 'Characteristics of ongoing studies'.

Results of the search

The search of MECIR‐mandatory databases resulted in 779 records and of Chinese databases in 350 records, which after deduplication and use of the Cochrane Screen4Me service (known assessments and RCT classifier (Marshall 2018; McDonald 2017; Noel‐Storr 2018; Thomas 2017)), were reduced to 691 records.

We identified a total of 123 references as potentially eligible after screening title or abstract, or both. Of these, we excluded 57 after checking the full text. We also excluded one study on the basis of clinicaltrials.gov status as 'terminated' with a total enrolment of three participants and after we had received no response from the principal investigator (NCT01006018). Of the remaining eligible 65 records, there were two ongoing studies (Beijing prediabetes reversion programme (BPRP); NCT02969798). We did not categorise any studies as 'awaiting classification'. We identified one systematic review through the search (Norris 2007). We did not find any additional references for this review by evaluation of references. We identified 27 studies (60 records) meeting our inclusion criteria. The screening process is presented in the flowchart Figure 1.

Figure 1

Study flow diagram

HTA: health technology assessment; RCT: randomised controlled trial: Screen4Me: Cochrane´s screening service.

Included studies

A detailed description of the characteristics of included studies is presented in Characteristics of included studies; Appendix 10; Appendix 11; Appendix 12; Appendix 13; Appendix 14; Appendix 15; Appendix 16; Appendix 17; Appendix 18; Appendix 19; Appendix 20. The following is a succinct overview.

Source of data

All reported data were published in medical journals. One study reported data in a trials register (ACT NOW), no additional data were available from the clinical trials register. We contacted all study authors and one pharmaceutical company (Takeda Pharmaceutical Company) for additional data. One study provided extra data on events of hypoglycaemia through correspondence (Attallah 2007).

Comparisons

Six studies compared pioglitazone as monotherapy with placebo (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2; Shi 2014; Xu 2011). One study compared pioglitazone with placebo through a two by two factorial design with growth hormone (Attallah 2007). This study included four intervention groups; we could include only two of these in our review (pioglitazone plus placebo growth hormone versus placebo pioglitazone plus placebo growth hormone; Attallah 2007).

Twenty‐one studies compared pioglitazone with no intervention (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Tian 2015; Wu 2013; Xiu 2015; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009). Five of these studies applied identical antihypertensive drugs to intervention and comparator groups (Che 2014; Chen 2007b; Gao 2011; Han 2007; Wu 2013). Three studies compared pioglitazone with other glucose‐lowering drugs (Chen 2007a; Zeng 2013; Zhang 2007). One study compared pioglitazone with personalised diet and exercise consultation (Zhang 2007).

Overview of study populations

Four studies provided information on sample size calculation (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2). The same four studies reported a total number of participants screened (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2). A total of 4344 participants were randomised in the included studies. Studies randomised a total of 1921 participants to pioglitazone as monotherapy or combined with antihypertensive drugs or calcium plus D‐vitamin with identical regimens in intervention and comparator groups. Studies randomised a total of 2423 participants to a comparator group. Eighteen studies reported on participant dropout (ACT NOW; Attallah 2007; Bone 2013; Chen 2007a; Chen 2007b; Fang 2013; Gao 2011; Guo 2010; Han 2007; IDPP‐2; Ke 2006; Liang 2004; Xiu 2015; Xu 2011; Yi 2015; Zeng 2013; Zhang 2007; Zhang 2015). Finishing rates in the 18 studies reporting on dropout were 87.3% (1098/1258) of the participants randomised to an intervention group and 82.9% (1464/1767) of the participants randomised to a comparator group (Table 1). The individual sample size across intervention and comparator groups ranged between 43 participants (Han 2007), and 605 participants (Deng 2013), see Table 1.

Study design

All but one studies were parallel RCTs. The one study was a two by two factorial RCT with growth hormone (Attallah 2007). Three studies included multiple arms of behaviour‐changing intervention and other glucose‐lowering drugs (Chen 2007a; Zeng 2013; Zhang 2007). Six studies applied placebo (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2; Shi 2014; Xu 2011). Twenty‐one studies compared pioglitazone with no intervention (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Tian 2015; Wu 2013; Xiu 2015; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009). Two out of 27 studies were conducted in multiple centres (ACT NOW; Bone 2013), the number of centres ranged between 8 and 31. Four studies reported blinding of participants, personnel and outcome assessors (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2). The remaining 23 studies reported no blinding or did not report on blinding (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Shi 2014; Tian 2015; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009). Studies were conducted between 2002 and 2014. Duration of intervention ranged between 6 and 36 months. Duration of follow‐up ranged between 6 and 48 months. One study reported a run‐in period of four weeks (Attallah 2007). Three studies reported titration periods (ACT NOW; Bone 2013; IDPP‐2). Titration periods ranged between four weeks and six months, after which pioglitazone and placebo doses were increased. None of the included studies reported early termination. Three studies reported wash‐out periods or extended follow‐ups (ACT NOW; Attallah 2007; Bone 2013). Three studies reported receiving commercial funding (ACT NOW; Attallah 2007; Bone 2013). Three studies reported receiving only non‐commercial funding (Fang 2013; IDPP‐2; Wu 2013). The remaining 21 studies did not report any details on funding (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Shi 2014; Tian 2015; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009).

Settings

Three out of 27 studies were conducted in the USA (ACT NOW; Attallah 2007; Bone 2013), 23 studies were conducted in China (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Shi 2014; Tian 2015; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009), and one study was conducted in India (IDPP‐2). All but one study reported an outpatient setting (Wu 2013).

Participants

Four out of 27 studies reported distribution of ethnicity in study populations (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2). The remaining 23 studies did not report on ethnicity of participants but were all assumed to have included participants of Chinese ethnicity (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Shi 2014; Tian 2015; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009). One study included only female participants (Bone 2013), the remaining studies reported distribution of sexes. The mean age of included participants ranged between 45 and 73 years (see Appendix 13). Four out of 27 studies included only elderly participants (Guo 2009; Guo 2010; Wu 2013; Xu 2011).

All but two studies reported baseline fasting plasma glucose (FPG) and two‐hour glucose post‐oral glucose tolerance test (OGTT; Chen 2007b; Gao 2011). Fifteen out of 27 studies reported baseline values for glycosylated haemoglobin (HbA1c; ACT NOW; Attallah 2007; Bone 2013; Che 2014; Deng 2013; Fang 2013; Guo 2009; Guo 2010; IDPP‐2; Ke 2006; Li 2017; Liang 2004; Shi 2014; Wu 2013; Zhang 2007). Mean values of FPG at baseline varied between 5.5 mmol/L and 6.9 mmol/L. Two‐hour glucose means at baseline varied between 7.9 mmol/L and 10.3 mmol/L. HbA1c means at baseline varied between 5.5% and 6.5%.

Five studies had hypertension as an inclusion criterion (Che 2014; Chen 2007b; Gao 2011; Han 2007; Wu 2013). None of the remaining 22 included studies reported data on comorbidities or comedications on study populations.

Twenty out of 27 studies reported exclusion criteria (ACT NOW; Attallah 2007; Bone 2013; Chen 2007a; Fang 2013; Guo 2009; Guo 2010; Han 2007; IDPP‐2; Li 2017; Shi 2014; Tian 2015; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhao 2009). Major exclusion criteria were diabetes; receiving glucose‐lowering drugs or drugs known to alter glucose metabolism; known disease of the heart, kidneys, liver or endocrine organs; acromegaly; chronic infections; pregnancy or breastfeeding; and excessive alcohol consumption.

Diagnosis

Seventeen out of 27 studies used the WHO 1999 criteria for intermediate hyperglycaemia or prediabetes (FPG > 6.1 mmol/L and < 7.0 mmol/L or two‐hour glucose ≥ 7.8 mmol/L and < 11.1 mmol/L: Attallah 2007; Che 2014; Chen 2007a; Gao 2011; Guo 2009; Guo 2010; Han 2007; IDPP‐2; Ke 2006; Liang 2004; Tian 2015; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhao 2009). Two studies applied the ADA 2003 criteria for intermediate hyperglycaemia or prediabetes (FPG ≥ 5.3 mmol/L and < 6.9 mmol/L and two‐hour glucose ≥ 7.8 mmol/L and < 11.0 mmol/L: Bone 2013; Fang 2013). Five out of 27 studies used ADA 2010 definitions (FPG ≥ 5.6 mmol/L and < 6.9 mmol/L and two‐hour glucose ≥ 7.8 mmol/L and < 11.0 mmol/L; Deng 2013; Shi 2014; Wu 2013; Zhang 2007; Zhang 2015). One study used authors' own definitions of intermediate hyperglycaemia or prediabetes (ACT NOW; see Appendix 10). The last two studies did not report their definitions of IGT, IFG or prediabetes (Chen 2007b; Li 2017).

Eleven out of 27 studies included participants on the basis of IGT only (Che 2014; Chen 2007a; Chen 2007b; Gao 2011; Guo 2009; Guo 2010; IDPP‐2; Ke 2006; Li 2017; Liang 2004; Zeng 2013). Eleven studies included participants with IGT plus IFG (Attallah 2007; Han 2007; Tian 2015; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zhang 2007; Zhang 2015; Zhao 2009). Four studies included participants with IGT or IFG, or both (Bone 2013; Deng 2013; Fang 2013; Shi 2014). One study included participants with 'prediabetes' and did not specify their definition of 'prediabetes' (Tian 2015). None of the studies included participants based on elevated glycosylated haemoglobin (HbA1c).

Interventions

None of the included studies reported treatment at baseline. All studies applied pioglitazone in intervention groups. Two out of 27 studies applied a final dose of 45 mg once daily titrated up from 30 mg once daily after four weeks or one month (ACT NOW; Bone 2013). One study applied a final dose of 30 mg once daily with titration up from 15 mg after six months (IDPP‐2). Another study applied 30 mg once daily titrated up from 15 mg once daily after four weeks, with growth hormone injections after the fourth week in a two‐by‐two factorial design (Attallah 2007). We included the intervention arms with growth hormone placebo injections (Attallah 2007). One study applied a stable dose of 38 mg once daily (Zeng 2013). Nine studies applied a stable dose of 30 mg once daily (Che 2014; Deng 2013; Gao 2011; Guo 2009; Guo 2010; Shi 2014; Tian 2015; Wu 2013; Xu 2011). Twelve studies applied a stable dose of 15 mg once daily (Chen 2007a; Chen 2007b; Fang 2013; Han 2007; Ke 2006; Li 2017; Liang 2004; Xiu 2015; Yi 2015; Zhang 2007; Zhang 2015; Zhao 2009). One study applied a stable dose of 4 mg once daily (Yu 2011). All pharmacological interventions, except placebo growth hormone injections, were administered as pills.

One out of 27 studies applied intensive behaviour‐changing intervention as a sole intervention in one intervention arm (Zhang 2007).

Outcomes

Twenty‐two out of 27 included studies explicitly stated a primary outcome in their publication (ACT NOW; Attallah 2007; Bone 2013; Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; IDPP‐2; Ke 2006; Li 2017; Liang 2004; Shi 2014; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015). All included studies stated secondary outcomes. Included studies collected a median of seven (range 4 to 21) outcomes. Seven out of 27 studies reported on all‐cause mortality (ACT NOW; Attallah 2007; Bone 2013; Chen 2007b; Deng 2013; Gao 2011; IDPP‐2). A total of 22 out of 27 included studies reported incidence of type 2 diabetes mellitus (T2DM) (ACT NOW; Attallah 2007; Bone 2013; Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; IDPP‐2; Ke 2006; Li 2017; Liang 2004; Shi 2014; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015), this was the outcome most commonly defined as the primary outcome in the included studies. Sixteen out of 27 studies reported serious adverse events or non‐serious adverse events, or both (ACT NOW; Attallah 2007; Bone 2013; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Han 2007; IDPP‐2; Ke 2006; Li 2017; Liang 2004; Zhang 2007; Zhang 2015; Zhao 2009), see Appendix 18; Appendix 19; Appendix 20. Two studies reported on cardiovascular mortality, non‐fatal stroke and myocardial infarction (ACT NOW; IDPP‐2). Three out of 27 studies reported having external judiciary panels for assessing serious adverse events and deaths (ACT NOW; Bone 2013; IDPP‐2). We identified one out of the 27 studies reporting study results in a clinical trials register (ACT NOW). There were differences in data regarding adverse and serious adverse events in publications versus clinicaltrials.gov. We contacted the study authors and the sponsoring pharmaceutical company without clarification (see Appendix 21).

Excluded studies

We excluded 30 studies (58 records) after full‐text evaluation (Figure 1; Characteristics of excluded studies). We excluded 16 studies (44 records) due to population not being 'prediabetes' without comorbidity except obesity or hypertension (e.g. non‐alcoholic steatohepatitis (NASH) or prior stroke) (Belfort 2006; ChiCTR‐TRC‐08000099; ChiCTR‐TRC‐08000111; Cusi 2016; IRIS 2016; J‐SPIRIT 2015; Li 2009; Nam 2013; NCT00015626; NCT00722631; UMIN000001035; Wu 2008; Yang 2012; Yokoyama 2007; Zhang 2014; Zhao 2011). We excluded four studies (four records) because of wrong comparison (EudraCT2005‐004421‐26; Karim 2016; NCT00306826; Zhou 2006); four studies (four records) because of an intervention period of fewer than 24 weeks (EudraCT2006‐002084‐49; Liu 2015; NCT00470262; NCT00633282); four records as they were not RCTs (Kawamori 1996; Kobayashi 2005; Lan 2012; Norris 2007); one study (one record) because of wrong intervention (Durbin 2004); and one study (one record) because of being listed as terminated with three enrolled participants and no answer from the principal investigator following contact (NCT01006018).

Risk of bias in included studies

For details on the risk of bias of the included studies see Characteristics of included studies.

For an overview of review authors' judgements about each risk of bias item for individual studies and across all studies see Figure 2 and Figure 3.

Figure 2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies (blank cells indicate that the particular outcome was not measured in some studies)

Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included study (blank cells indicate that the particular outcome was not measured in some studies)

Allocation

We judged three studies to be at low risk of selection bias due to randomisation sequence and allocation concealment (ACT NOW; Attallah 2007; Bone 2013). One study did not describe a truly randomised sequence, but groups appeared similar for all important prognostic variables at baseline; we judged the study as low risk of selection bias (IDPP‐2; see Appendix 8). We judged 23 studies as unclear risk of selection bias due to not adequately describing their randomisation sequence or allocation concealment and reporting limited baseline characteristics (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Shi 2014; Tian 2015; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009; see Characteristics of included studies and Appendix 8). We judged no included studies as having a high risk of selection bias.

Blinding

All‐cause mortality, incidence of T2DM, serious adverse events and measures of blood glucose control are unlikely to be influenced by performance or detection bias. We judged all studies as having low risk of performance or detection bias across these outcomes. Four out of 27 studies explicitly stated that they had undertaken blinding of participants and personnel (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2). We judged all four of these studies as having low risk of performance and detection bias. The remaining 23 studies did not report any blinding (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Shi 2014; Tian 2015; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009). We judged 12 of these studies as having high risk of performance and detection bias for non‐serious adverse events, due to adverse events being likely to be influenced by the lack of blinding (Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Han 2007; Ke 2006; Li 2017; Liang 2004; Zhang 2007; Zhang 2015; Zhao 2009).

Incomplete outcome data

Eighteen out of 27 studies reported on total participants randomised and completion rates (ACT NOW; Attallah 2007; Bone 2013; Chen 2007a; Chen 2007b; Fang 2013; Gao 2011; Guo 2010; Han 2007; IDPP‐2; Ke 2006; Liang 2004; Xiu 2015; Xu 2011; Yi 2015; Zeng 2013; Zhang 2007; Zhang 2015), the attrition rate among these studies ranged from 0% to 31.8% for the intervention groups and from 0% to 23.7% for comparator groups. Two out of 27 studies did intention‐to‐treat analyses (Attallah 2007; Zhang 2007).

We judged one study as low risk of attrition bias across all outcomes due to balanced minor dropout between groups (Zhang 2007). We judged 21 out of 27 studies as unclear risk of bias due to missing or insufficient information regarding dropouts, reasons for dropout, evaluation or handling of missing data (Che 2014; Deng 2013; Fang 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; IDPP‐2; Ke 2006; Liang 2004; Li 2017; Shi 2014; Tian 2015; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhao 2009; Zhang 2015). We judged five studies out of 27 as having high risk of attrition bias regarding at least one outcome in the scope of this review due to high dropouts or dropouts not balanced across intervention and comparator groups (ACT NOW; Attallah 2007; Bone 2013; Chen 2007a; Chen 2007b).

Selective reporting

One out of 27 studies provided a full‐length protocol (ACT NOW).

We judged two studies as having low risk of reporting bias (Attallah 2007; IDPP‐2). We judged three studies as unclear risk of reporting bias because of insufficient information (Bone 2013; Liang 2004; Zhang 2015). We judged one study as having high risk of reporting bias due to serious adverse events not sufficiently reported and analysed in publications despite being recorded according to protocol (ACT NOW). We judged 21 studies as having high risk of reporting bias due to no or insufficient reporting on key outcomes for this review (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Shi 2014; Tian 2015; Wu 2013; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhao 2009). The outcomes mostly unreported were mortality, incidence of T2DM, serious and non‐serious adverse events and hypoglycaemia (see Appendix 15; Appendix 16; Characteristics of included studies).

Other potential sources of bias

We judged all studies as being at unclear risk of bias regarding other potential sources of bias.

Three out of 27 studies were funded completely or partly by pharmaceutical companies (ACT NOW; Attallah 2007; Bone 2013). Twenty‐one studies did not report funding source (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Gao 2011; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Shi 2014; Tian 2015; Xiu 2015; Xu 2011; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009). One study only gave 10 participants from each group the planned dose because of reported side effects in both groups (IDPP‐2). The remaining participants were treated with a lower dosage of pioglitazone. Another study altered inclusion criteria from the protocol and extended the recruitment period (ACT NOW). One study was conducted in 31 sites including 156 participants; six sites did not include anyone, details on recruitment and dropout across study sites were not available (Bone 2013).

Effects of interventions

See: Summary of findings 1 Summary of findings: pioglitazone monotherapy versus another pharmacological glucose‐lowering monotherapy; Summary of findings 2 Summary of findings: pioglitazone monotherapy versus intensive behaviour‐changing intervention; Summary of findings 3 Summary of findings: pioglitazone monotherapy versus placebo; Summary of findings 4 Summary of findings: pioglitazone monotherapy versus no intervention

Baseline characteristics

For details of baseline characteristics, see Appendix 12; Appendix 13.

Pioglitazone monotherapy versus any pharmacological glucose‐lowering monotherapy intervention (acarbose, metformin, sulphonylurea)

Three studies compared pioglitazone with another glucose‐lowering monotherapy intervention with matching behaviour‐changing interventions in each group (Chen 2007a; Zeng 2013; Zhang 2007). All three studies included a metformin group. Additionally, Chen 2007a included an acarbose group, and Zhang 2007 included a repaglinide group. For an overview of major results see summary of findings Table 1.

Primary outcomes

All‐cause mortality

None of the three included studies reported all‐cause mortality.

Incidence of T2DM

Pioglitazone versus metformin

All three studies reported on incidence of T2DM compared with metformin. A total of 9 out of 168 participants developed T2DM in the pioglitazone group compared with 9 out of 163 participants in the metformin group (RR 0.98, 95% CI 0.40 to 2.38; P = 0.96; 3 studies, 331 participants; low‐certainty evidence; Analysis 1.1). The 95% prediction interval did not provide a meaningful estimate.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT' with IGT only did not show interaction between subgroups (P = 0.96; Analysis 1.2).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, sex, ethnicity, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis
- Including studies only diagnosing by WHO 1999 did not substantially change the effect estimate (RR 0.96, 95% CI 0.35 to 2.64; P = 0.94; 2 studies, 234 participants; Chen 2007a; Zeng 2013). One study applied the ADA 2010 criteria and showed that 2 out of 48 participants in the pioglitazone group compared with 2 out of 49 participants in the metformin group developed T2DM (Zhang 2007).
- Studies had similar size, publications status, bias judgements, language, reported funding and no imputation was done. Therefore, we did not perform further sensitivity analyses.

Pioglitazone versus acarbose

One study reported on the incidence of T2DM comparing pioglitazone with acarbose: 1 out of 50 participants in the pioglitazone group developed T2DM compared with 2 out of 46 participants in the acarbose group (Chen 2007a; very low‐certainty evidence).

Pioglitazone versus repaglinide

One study reported on the incidence of T2DM comparing pioglitazone with repaglinide: 2 out of 48 participants in the pioglitazone group developed T2DM compared with 1 out of 48 participants in the repaglinide group (Zhang 2007; low‐certainty evidence).

Serious adverse events

Pioglitazone versus metformin

Two studies that compared pioglitazone with metformin reported that no participants experienced serious adverse events (0 out of 102 participants in the pioglitazone group and 0 out of 99 participants in the metformin group; low‐certainty evidence; Chen 2007a; Zhang 2007).

Pioglitazone versus acarbose

One study that compared pioglitazone with acarbose reported that no participants experienced serious adverse events (0 out of 50 participants in the pioglitazone group and 0/46 participants in the acarbose group; very low‐certainty evidence; Chen 2007a).

Pioglitazone versus repaglinide

One study that compared pioglitazone with repaglinide reported that no participants experienced serious adverse events (0 out of 52 participants in the pioglitazone group and 0 out of 51 participants in the repaglinide group; low‐certainty evidence; Zhang 2007).

Secondary outcomes

None of the included studies that compared pioglitazone with another glucose‐lowering drug reported on cardiovascular mortality, non‐fatal myocardial infarction, non‐fatal stroke, congestive heart failure, amputation of lower extremity, blindness or severe vision loss, end‐stage renal disease, hypoglycaemia, health‐related quality of life, time to progression to T2DM or socioeconomic effects.

Non‐serious adverse events

Pioglitazone versus metformin

Two studies reported on non‐serious adverse events (Chen 2007a; Zhang 2007).

A total of 4 out of 102 participants experienced non‐serious adverse events in the pioglitazone group compared with 9 out of 99 participants in the metformin group (RR 0.48, 95% CI 0.04 to 5.51; P = 0.55; 2 studies, 201 participants; Analysis 1.3).

Subgroup analysis: comparing 'prediabetes' criteria of 'IFG plus IGT' with IGT only did not indicate subgroup interaction (P = 0.07; Analysis 1.4). Subgroup analyses for duration of intervention, mono/dual‐therapy, age, sex, ethnicity, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis: could not be performed due to lack of data.

Pioglitazone versus acarbose

One study reported 1 out of 50 participants in the pioglitazone group experienced a non‐serious adverse event compared with 2 out of 46 participants receiving acarbose (Chen 2007a).

Pioglitazone versus repaglinide

One study reported 3 out of 52 participants in the pioglitazone group experienced a non‐serious adverse event compared with 0 out of 51 participants receiving repaglinide (Zhang 2007).

Measures of blood glucose control

All three included studies that compared pioglitazone with another glucose‐lowering drug reported on measures of blood glucose control. None of the studies reported the statistical variation of the changes from baseline values. We imputed these using the correlation coefficient from IDPP‐2.

Fasting plasma glucose

Pioglitazone versus metformin

Pioglitazone compared with metformin showed a MD of 0.03 mmol/L (95% CI −0.17 to 0.23; P = 0.77; 3 studies, 339 participants; Analysis 1.5). The 95% prediction interval ranged between −1.3 mmol/L and 1.3 mmol/L.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT' with IGT only did not show interaction between subgroups (P = 0.40; Analysis 1.6).
- Subgrouping by comorbidity into studies with hypertension as an inclusion criterion and studies without comorbidity in their inclusion criteria did not show subgroup interaction (P = 0.49; Analysis 1.7).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, sex, ethnicity and previous gestational diabetes were not possible.
Sensitivity analysis
- Two studies diagnosed the participants according to the WHO 1999 criteria (Chen 2007a; Zeng 2013). Analysing these two studies did not substantially change the effect estimate (MD 0.06 mmol/L, 95% CI −0.15 to 0.26; P = 0.59; 234 participants).
- The included studies were similar regarding publication status, 'Risk of bias' judgements, size and duration, imputation, language, funding and country. No further sensitivity analyses were possible.

Pioglitazone versus acarbose

One study that compared pioglitazone with acarbose reported changes in FPG from baseline (Chen 2007a). The change from baseline was −1.26 mmol/L (SD 1.0) in the pioglitazone group compared with −1.19 mmol/L (SD 0.8) in the acarbose group (assumed standard errors (SEs) were recalculated to SDs).

Pioglitazone versus repaglinide

One study that compared pioglitazone with metformin reported changes in FPG from baseline (Zhang 2007). The change from baseline was −0.66 mmol/L (SD 1.7) in the pioglitazone group compared with −0.42 mmol/L (SD 1.8) in the repaglinide group (assumed SEs were recalculated to SDs).

Two‐hour glucose

Pioglitazone versus metformin

Pioglitazone compared with metformin for two‐hour glucose after an OGTT showed a MD of −0.23 mmol/L (95% CI −0.43 to −0.02; P = 0.03; 3 studies, 339 participants; Analysis 1.8) in favour of pioglitazone. The 95% prediction interval ranged between −1.59 mmol/L and 1.13 mmol/L.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT' with IGT only did not show subgroup interaction (P = 0.19; Analysis 1.9).
- Subgrouping by comorbidity into populations with hypertension as an inclusion criterion and studies without comorbidity in their inclusion criteria did not show subgroup interaction (P = 0.92; Analysis 1.10).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, sex, ethnicity and previous gestational diabetes were not possible.
Sensitivity analysis
- Two studies diagnosed the participants according to the WHO 1999 criteria (Chen 2007a; Zeng 2013). Analysing these two studies did not substantially change the effect estimate (MD −0.27 mmol/L, 95% CI −0.47 to −0.08; P = 0.007; 234 participants).
- The included studies were similar regarding publication status, 'Risk of bias' judgements, size and duration, imputation, language, funding and country. No sensitivity analyses were possible.

Pioglitazone versus acarbose

One study that compared pioglitazone with metformin reported changes in two‐hour glucose from baseline (Chen 2007a). The change from baseline was −1.15 mmol/L (SD 1.2) in the pioglitazone group compared with −1.13 mmol/L (SD 1.4) in the acarbose group (assumed SEs were recalculated to SDs).

Pioglitazone versus repaglinide

One study that compared pioglitazone with repaglinide reported changes in two‐hour glucose from baseline (Zhang 2007). The change from baseline was −1.48% (SD 1.6) in the pioglitazone group compared with 1.39% (SD 1.5) in the repaglinide group (assumed SEs were recalculated to SDs).

HbA1c

Pioglitazone versus metformin

One study that compared pioglitazone with metformin reported change in HbA1c from baseline (Zhang 2007). The change from baseline was −0.15% (SD 1.8) in the pioglitazone group compared with −0.19% (SD 2.4) in the metformin group (assumed SEs were recalculated to SDs).

Pioglitazone versus acarbose

HbA1c at the end of follow‐up or change from baseline was not reported for this comparison.

Pioglitazone versus repaglinide

One study that compared pioglitazone with repaglinide reported change in HbA1c from baseline (Zhang 2007). The change from baseline was −0.15% (SD 1.8) in the pioglitazone group compared with −0.32% (SD 2.0) in the repaglinide group (assumed SEs were recalculated to SDs).

Pioglitazone monotherapy versus behaviour‐changing interventions (diet, exercise, diet and exercise)

One study compared pioglitazone combined with diet and exercise with an intensive behaviour‐changing intervention consisting of personalised diet and exercise consultation (Zhang 2007). For an overview of major results see summary of findings Table 2.

Primary outcomes

All‐cause mortality

The included study did not report all‐cause mortality.

Incidence of T2DM

A total of 2 out of 48 participants developed T2DM in the pioglitazone group compared with 5 out of 48 in the personalised diet and exercise consultation group (low‐certainty evidence).

Serious adverse events

Zhang 2007 reported that no participants experienced a serious adverse event (low‐certainty evidence).

Secondary outcomes

The included study did not report on cardiovascular mortality, non‐fatal myocardial infarction, non‐fatal stroke, congestive heart failure, amputation of lower extremity, blindness or severe vision loss, end‐stage renal disease, hypoglycaemia, health‐related quality of life, time to progression to T2DM or socioeconomic effects.

Non‐serious adverse events

Zhang 2007 reported that 3 out of 52 participants in the pioglitazone group experienced non‐serious adverse events compared with 0 out of 51 participants in the personalised diet and exercise consultation group.

Measures of blood glucose control

Zhang 2007 did not report statistical variation on changes from baseline values. We imputed these using correlation coefficients from the IDPP‐2.

Fasting plasma glucose

The change from baseline was −0.66 mmol/L (SD 1.7) in 52 participants in the pioglitazone group compared with −0.15 mmol/L (SD 1.2) in 51 participants in the personalised diet and exercise consultation group (assumed SEs were recalculated to SDs).

Two‐hour glucose

The change from baseline was −1.48 mmol/L (SD 1.6) in 52 participants in the pioglitazone group compared with 0.17 mmol/L (SD 1.2) in 51 participants in the personalised diet and exercise consultation group (assumed SEs were recalculated to SDs).

HbA1c

The change from baseline was −0.15% (SD 1.8) in 52 participants in the pioglitazone group compared with −0.12% (SD 1.1) in 51 participants in the personalised diet and exercise consultation group (assumed SEs were recalculated to SDs)

Pioglitazone monotherapy versus placebo

Six studies compared pioglitazone as monotherapy with placebo (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2; Shi 2014; Xu 2011). Five out of six studies implemented behaviour‐changing interventions, which were identical in intervention and placebo groups (ACT NOW; Bone 2013; IDPP‐2; Shi 2014; Xu 2011). Two studies administered a dose of 45 mg once daily (ACT NOW; Bone 2013). Four studies administered a dose of 30 mg once daily (Attallah 2007; IDPP‐2; Shi 2014; Xu 2011). Two studies had wash‐out follow‐ups of three to four weeks (Attallah 2007), and six months (Bone 2013). One study did an extended follow‐up on a subpopulation with a median follow‐up of 11.4 months after medication stopped (ACT NOW). Three studies included participants of mixed ethnicity in the USA (ACT NOW; Attallah 2007; Bone 2013). One study included participants of Asian Indian ethnicity (IDPP‐2). Two studies did not report on ethnicity but we assumed the population to be of Chinese ethnicity (Shi 2014; Xu 2011). We judged all studies to have a high risk of bias in at least one 'Risk of bias' domain (see Figure 3). For an overview of major results see summary of findings Table 3.

Primary outcomes

All‐cause mortality

Four out of six studies reported on all‐cause mortality (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2). Two studies reported events (ACT NOW; IDPP‐2).

Five out of 577 participants in the pioglitazone groups died compared with 2 out of 579 participants in the placebo groups (Peto OR 2.38, 95% CI 0.54 to 10.50; 4 studies, 1156 participants; very low‐certainty evidence; Analysis 2.1).

Subgroup analysis
- The two included studies with events could be subgrouped by country of origin, ethnicity of population, funding and diagnostic criteria. All these factors showed no subgroup interaction (P = 0.85; Analysis 2.2).
- Subgroup analysis for duration of intervention, mono/dual‐therapy, age, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis
- Two studies diagnosed participants according to the WHO 1999 criteria and showed that 2 out of 196 participants died in the pioglitazone groups compared with 1 out of 202 in the placebo groups (Attallah 2007; IDPP‐2). One study applied the ADA 2003 criteria and showed that no participants died (Bone 2013).
- The included studies were similar regarding publication status, bias judgements, duration and size, imputation, language, funding and country. No further sensitivity analyses were possible.

Incidence of T2DM

Intervention periods

All studies reported on incidence of T2DM. One study reported a cumulative incidence of participants developing T2DM as percentages, with 29.8% in the pioglitazone group developing T2DM compared with 31.6% in the placebo group (IDPP‐2). We calculated the incidence based on all randomised participants.

A total of 80 out of 700 participants developed T2DM in the pioglitazone groups compared with 131 out of 695 participants in the placebo groups (RR 0.40, 95% CI 0.17 to 0.95; P = 0.04; 6 studies, 1395 participants; low‐certainty evidence; Analysis 2.3), favouring pioglitazone. However, the 95% prediction interval ranged between 0.03 and 4.68.

Subgroup analysis: comparing 'prediabetes' criteria of 'IFG plus IGT' with 'IFG or IGT' with IGT only showed interaction in subgroups favouring pioglitazone.
- IGT only compared with 'IFG plus IGT' showed a reduced incidence of T2DM for the 'IFG plus IGT' criteria (P < 0.001; Analysis 2.4).
- Subgrouping by ethnicity by mainly white (mixed American ethnicities), Asian Indian and assumed Chinese showed subgroup interaction and favoured pioglitazone with reduced T2DM incidence in studies including mixed American ethnicities (P < 0.001; Analysis 2.5).
- Subgrouping by age (cut‐off 65 years of age) did not show subgroup interaction (P = 1.0; Analysis 2.6).
- Subgrouping by sex (men and women versus women only) did not show subgroup interaction (P = 0.25; Analysis 2.7).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis
- Including only smaller studies (N < 100 participants) changed the effect estimate (RR 0.25, 95% CI 0.03 to 2.14; P = 0.21; 2 studies, 111 participants; Attallah 2007; Shi 2014), no longer favouring pioglitazone.
- Sensitivity analysis including only studies published in English also changed the effect estimate (RR 0.42, 95% CI 0.14 to 1.26; P = 0.12; 4 studies, 1196 participants; ACT NOW; Attallah 2007; Bone 2013; IDPP‐2), no longer favouring pioglitazone.
- Three studies diagnosed the participants according to the WHO 1999 criteria (Attallah 2007; IDPP‐2; Xu 2011). Analysing these studies changed the effect estimate (RR 0.88, 95% CI 0.54 to 1.43; P = 0.60; 3 studies, 557 participants; (Attallah 2007; IDPP‐2; Xu 2011), no longer favouring pioglitazone. One study diagnosed the participants according to the ADA 2003 criteria; this showed that 1 out of 78 participants in the pioglitazone group developed T2DM compared with 8 out of 78 participants in the placebo group (Bone 2013). One study diagnosed the participants according to the ADA 2010 criteria; this showed that 1 out of 40 participants in the pioglitazone group developed T2DM compared with 4 out of 40 participants in the placebo group (Shi 2014).
- The included studies were similar regarding publications status, 'Risk of bias' judgements, duration of intervention and imputation.

Wash‐out and extended follow‐up

Three studies reported data for T2DM after an extended observational period ranging between three weeks and 11.4 months (ACT NOW; Attallah 2007; Bone 2013). ACT NOW invited completers of their main study to continue biannual follow‐ups after the end of intervention. About half of the participants randomised completed one‐year post‐treatment follow‐up. ACT NOW reported a hazard ratio (HR) of the cumulative incidence of T2DM across intervention and post‐treatment periods of pioglitazone compared with placebo during the intervention period of HR 0.44 (95% CI 0.29 to 0.67; P < 0.005; 290 participants), favouring pioglitazone.

A total of 19 out of 244 participants with prior pioglitazone treatment compared with 20 out of 224 participants with prior placebo treatment developed T2DM after the wash‐out follow‐up (RR 0.89, 95% CI 0.49 to 1.60; P = 0.70; 3 studies, 468 participants; Analysis 2.8).

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT' with 'IFG or IGT' did not show subgroup interaction (P = 0.50; Analysis 2.9).
- Subgroup analysis for age was identical to the one with 'prediabetes' criteria.
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, ethnicity, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis
- There was one smaller‐size study (N < 100 participants; Attallah 2007). During wash‐out and within three to four weeks, 2 out of 15 participants developed T2DM in the pioglitazone group compared with 2 out of 16 participants in the placebo group (Attallah 2007).
- Sensitivity analysis strictly including protocol prespecified extension did not substantially change the effect estimate (RR 0.78, 95% CI 0.16 to 3.81; P = 0.76; 2 studies, 178 participants; Attallah 2007; Bone 2013).
- One study diagnosed the participants according to the WHO 1999 criteria (Attallah 2007). One study diagnosed the participants according to the ADA 2003 criteria, this showed that 0 out of 77 participants in the pioglitazone group developed T2DM compared with 1 out of 70 participants in the placebo group (Bone 2013).
- The included studies were similar regarding publication status, 'Risk of bias' judgements, duration of intervention, imputation, language, funding and country.

Serious adverse events

Two studies reported on serious adverse events (Attallah 2007; Bone 2013). A total of 3 out of 93 participants experienced serious adverse events in the pioglitazone group compared with 1 out of 94 participants in the placebo group (RR 3.00, 95% CI 0.32 to 28.22; P = 0.34; 2 studies, 187 participants; very low‐certainty evidence; Analysis 2.10). Only Bone 2013 observed events.

IDPP‐2 reported 10 "major adverse events" in 181 participants in the pioglitazone group compared with 13 "major adverse events" in 186 participants in the placebo group.

ACT NOW reported only deaths as serious adverse events and listed all other adverse events without distinguishing between serious and non‐serious adverse events. We contacted study authors and the sponsoring pharmaceutical company but did not receive additional data.

Subgroup and sensitivity analyses were not possible because only one study reported any events.

Secondary outcomes

Cardiovascular mortality

Four included studies reported on cardiovascular mortality (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2). Only IDPP‐2 observed two events of cardiovascular mortality in 181 participants in the pioglitazone group compared with no event in 186 participants in the placebo group (RR 5.14, 95% CI 0.25 to 106.28; P = 0.29; very low‐certainty evidence; Analysis 2.11).

One study reported two unexplained deaths, one in each group of 303 participants in the pioglitazone group and 299 participants in the placebo group (ACT NOW).

Non‐fatal myocardial infarction

Three studies reported on non‐fatal myocardial infarction (ACT NOW; Attallah 2007; Bone 2013). Only ACT NOW observed two events in 303 participants in the pioglitazone group compared with one event in 299 participants in the placebo group (RR 1.97, 95% CI 0.18 to 21.65; very low‐certainty evidence; Analysis 2.12).

One study reported two events of heart disease in 181 participants in the pioglitazone group and 186 participants in the placebo group without specifying the kind of event (IDPP‐2). We contacted study authors for details but did not receive a reply.

Non‐fatal stroke

None of the included studies reported any events of non‐fatal stroke. One study reported two events of transient ischaemic attacks in 303 participants in the pioglitazone group compared with no transient ischaemic attacks in 299 participants in the placebo group (ACT NOW).

Congestive heart failure

Two studies reported on events of congestive heart failure (ACT NOW; Bone 2013). Two studies provided complete lists of adverse events (Attallah 2007; IDPP‐2). Three studies were available for meta‐analysis (ACT NOW; Attallah 2007; Bone 2013).

There was a total of two events among 396 participants in the pioglitazone groups compared with one event among 393 participants in the placebo groups (Peto OR 1.93, 95% CI 0.20 to 18.60; P = 0.57; 3 studies, 789 participants; Analysis 2.13).

One study reported 2 out of 181 participants with unspecified heart disease events in the pioglitazone group compared with 2 out of 186 in the placebo group (IDPP‐2).

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT' with 'IFG or IGT' did not show interaction in subgroups (P = 0.61; Analysis 2.14).
- Subgroup analysis for sex was identical to 'prediabetes' criteria (Analysis 2.15).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, ethnicity, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis
- Including only smaller studies was inconclusive due to no events in the one small study (N < 100 participants; Attallah 2007).
- The included studies were similar regarding publication status, 'Risk of bias' judgements, duration of intervention, imputation, language, funding and country.

Amputation of lower extremity

None of the included studies reported amputation of lower extremity.

Blindness or severe vision loss

None of the included studies reported blindness or severe vision loss.

End‐stage renal disease

None of the included studies reported end‐stage renal disease.

Non‐serious adverse events

Three included studies reported on non‐serious adverse events (ACT NOW; Attallah 2007; Bone 2013).

Two studies were available for meta‐analysis (Attallah 2007; Bone 2013). A total of 51 participants experienced a non‐serious adverse event in pioglitazone groups among 93 participants compared with 49 participants among 94 participants in placebo groups (RR 1.03, 95% CI 0.81 to 1.31; P = 0.82; 2 studies, 187 participants; Analysis 2.16).

ACT NOW reported only deaths as serious adverse events and listed all other adverse events without distinguishing between serious and non‐serious adverse events. A total of 121 out of 303 (39.9%) participants in the pioglitazone group compared with 151 out of 299 (50.5%) participants in the placebo group experienced an adverse event (P = 0.03).

IDPP‐2 only reported the total number of non‐serious adverse events and not the number of participants with a non‐serious adverse event; they reported a total of 28 non‐serious adverse events among 181 participants in the pioglitazone group compared with 22 among 186 participants in the placebo group.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT' with 'IFG or IGT' did not show significant interaction between subgroups (P = 0.53; Analysis 2.17).
- Subgroup analysis for sex was identical to 'prediabetes' criteria analysis.
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, ethnicity, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis
- There was one smaller‐sized study (N < 100 participants; Attallah 2007). This study diagnosed the participants according to the WHO 1999 criteria with 2 out of 15 participants experiencing non‐serious adverse events in the pioglitazone group compared with 1 out of 16 participants in the placebo group (Attallah 2007). One study diagnosed the participants according to the ADA 2003 criteria and showed 49 out of 78 participants in the pioglitazone group experiencing non‐serious adverse events compared with 48 out of 78 participants in the placebo group (Bone 2013).
- The included studies were similar regarding publication status, 'Risk of bias' judgements, duration, imputation, language, funding and country.

Hypoglycaemia

Two studies reported on hypoglycaemic events (Attallah 2007; IDPP‐2).

Attallah 2007 reported by correspondence that no participant experienced hypoglycaemia.

IDPP‐2 reported a total of 15 hypoglycaemic events among 181 participants in the pioglitazone group compared with 9 hypoglycaemic events among 186 participants in the placebo group. They did not specify the number of participants who experienced hypoglycaemia.

Health‐related quality of life

None of the included studies reported health‐related quality of life.

Time to progression to T2DM

Two studies reported on time to progression to T2DM (ACT NOW; IDPP‐2). The studies showed considerable heterogeneity with an I² of 94% (P < 0.001). Therefore, we did not perform a meta‐analysis (Analysis 2.18).

ACT NOW reported time to progression to T2DM with a HR of 0.28 (95% CI 0.16 to 0.49; P = 0.001; 602 participants). Adjustment for baseline characteristics did not alter the HR. The annual incidence of T2DM was 2.1% in the pioglitazone group compared with 7.6% in the placebo group.

IDPP‐2 reported time to progression to T2DM with an unadjusted HR of 1.08 (95% CI 0.75 to 1.56; P = 0.665; 407 participants). The HR adjusted for age, sex, study group, BMI, two‐hour glucose, insulin resistance, insulinogenic index, hypertension and family history of T2DM showed HR 0.98 (95% CI 0.67 to 1.44; P = 0.94). The cumulative incidence of T2DM was 29.8% in the pioglitazone group compared with 31.6% in the placebo group after 36 months of intervention.

Subgroup analysis
- The two included studies could be subgrouped by country of origin, ethnicity of population, funding and diagnostic criteria. All these factors showed subgroup interaction (P < 0.001; Analysis 2.19).
- Subgroup analysis for duration of intervention, mono/dual‐therapy, age, comorbidity and previous gestational diabetes was not possible.
Sensitivity analysis
- One study diagnosed participants by WHO 1999 criteria and reported an unadjusted HR of 1.08 (95% CI 0.75 to 1.56; P = 0.665; 407 participants; IDPP‐2). The incidence of T2DM was 29.8% in the pioglitazone group and 31.6% in the placebo group after 36 months of intervention.
- The included studies were similar regarding publication status, bias judgements, duration and size, imputation, language, funding and country.

Measures of blood glucose control

All included studies reported on measures of blood glucose control for intervention periods. One study reported measures of blood glucose control for a wash‐out period (Bone 2013).

Fasting plasma glucose

Intervention periods

Five studies reported FPG (ACT NOW; Bone 2013; IDPP‐2; Shi 2014; Xu 2011). Shi 2014 and Xu 2011 did not report statistical variation for changes from baseline, we imputed these values using IDPP‐2's correlation coefficients. The studies showed considerable heterogeneity with an I² of 96% (P < 0.001). Excluding studies with imputed values still showed an I^{2 of} 86% (Shi 2014; Xu 2011). The MD between the pioglitazone and placebo groups ranged between ‐1.35 mmol/L and 0.05 mmol/L (1125 participants; 5 studies; Analysis 2.20).

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT' with 'IFG or IGT' or IGT only showed interaction between subgroups (P = 0.002; Analysis 2.21), favouring pioglitazone, contrasting 'IFG or IGT' with IGT only.
- Subgrouping by ethnicity into mainly white (mixed American ethnicities), Asian Indians and assumed Chinese ethnicity showed an interaction between groups with a lesser effect on Asian Indians (P = 0.0006; Analysis 2.22) favouring pioglitazone, contrasting mainly white people with Asian Indian people.
- Subgrouping by sex into studies with men and women versus women only did not show interaction between subgroups (P = 0.91; Analysis 2.23).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis
- There was one smaller‐sized study (N < 100 participants; Shi 2014).
- The same study was the only study diagnosing participants by ADA 2010 criteria. The reported change in FPG from baseline was −0.44 mmol/L (SD 1.3) in 40 participants in the pioglitazone group compared with −0.02 mmol/L (SD 1.3) in 40 participants in the placebo group (Shi 2014). Two studies diagnosed participants with the WHO 1999 criteria (IDPP‐2; Xu 2011). Including only these studies showed a MD of −0.65 mmol/L (95% CI −2.02 to 0.72; P = 0.35; 2 studies, 486 participants). One study applied the ADA 2003 criteria and reported a change from baseline of −0.16 mmol/L (SD 0.7) in 57 participants in the pioglitazone group compared with 0.33 mmol/L (SD 0.7) in 61 participants in the placebo group (Bone 2013).
- One study received non‐commercial funding and reported a change in FPG from baseline of 0.24 mmol/L (SD 1.1) in 181 participants in the pioglitazone group compared with 0.19 mmol/L (SD 1) mmol/L in 186 participants in the placebo group (IDPP‐2).
- Including only studies published in English showed a MD of −0.27 mmol/L (95% CI −0.55 to 0.01; P = 0.06; 3 studies, 926 participants; ACT NOW; Bone 2013; IDPP‐2), in favour of placebo.
- Sensitivity analysis regarding imputation was identical to the analysis regarding language.
- Including only studies from the USA showed a MD of −0.39 mmol/L (95% CI −0.48 to −0.29; P < 0.001; 2 studies, 559 participants; ACT NOW; Bone 2013).
- The included studies were similar regarding publication status, 'Risk of bias' judgements and duration; no sensitivity analyses regarding these factors were possible.

Wash‐out and extended follow‐up

One study reported change in FPG from baseline for wash‐out periods (Bone 2013). The MD was 0.06 mmol/L (SD 0.1) in 57 participants in the pioglitazone group compared with 0.02 mmol/L (SD 0.1) in 54 participants in the placebo group (Bone 2013).

Two‐hour glucose

Intervention periods

All included studies reported on two‐hour glucose after OGTT. Five studies were available for meta‐analysis (ACT NOW; Bone 2013; IDPP‐2; Shi 2014; Xu 2011). Shi 2014 and Xu 2011 did not report statistical variation for changes from baseline; we imputed these values using IDPP‐2's correlation coefficients. The studies showed considerable heterogeneity with an I² of 81%. The MD was −0.89 mmol/L (95% CI −1.43 to −0.34; P = 0.001; 5 studies, 1119 participants; Analysis 2.24), in favour of pioglitazone. The 95% prediction interval ranged between −2.87 mmol/L and 1.09 mmol/L.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT' with 'IFG or IGT', or with IGT only showed interaction between subgroups demonstrating a stronger effect in favour of pioglitazone for the 'IFG plus IGT' categorisation than for IGT only (P = 0.0003; Analysis 2.25).
- Subgrouping by ethnicity of mainly white ethnicity (mixed American ethnicity), Asian Indian and assumed Chinese ethnicity did not show subgroup interaction (P = 0.05; Analysis 2.26).
- Subgrouping by sex into groups of men and women versus women only did not show subgroup interaction (P = 0.35; Analysis 2.27).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis
- There was one smaller‐sized study (N < 100 participants; Shi 2014).
- The same study was the only study that diagnosed participants by ADA 2010 criteria. The reported change in two‐hour glucose from baseline was −0.81 mmol/L (SD 1.5) in 40 participants in the pioglitazone group compared with −0.05 mmol/L (SD 1.3) in 40 participants in the placebo group (Shi 2014). Two studies diagnosed participants with the WHO 1999 criteria (IDPP‐2; Xu 2011). Including only these studies did not substantially change the effect estimate: MD −0.88 mmol/L (95% CI −2.42 to 0.65; P = 0.26; 42 studies, 86 participants; IDPP‐2; Xu 2011). One study applied the ADA 2003 criteria and reported a change from baseline of −1.47 mmol/L (SD 1.8) in 54 participants in the pioglitazone group compared with −0.94 mmol/L (SD 1.7) in 58 participants in the placebo group (Bone 2013).
- One study received non‐commercial funding and reported a change in two‐hour glucose from baseline of 0.59 mmol/L (SD 3) in 181 participants in the pioglitazone group compared with 0.68 mmol/L (SD 2.6) in 186 participants in the placebo group (IDPP‐2).
- Including only studies published in English did not substantially change the effect estimate (MD −0.65 mmol/L, 95% CI −1.40 to 0.09; P = 0.09; 3 studies, 920 participants; ACT NOW; Bone 2013; IDPP‐2).
- Sensitivity analysis regarding imputation was identical to the analysis regarding language.
- Including only studies from the USA did not substantially change the effect estimate (MD −0.94 mmol/L, 95% CI −1.65 to −0.23; P = 0.009; 2 studies, 559 participants; ACT NOW; Bone 2013).
- The included studies were similar regarding publication status, 'Risk of bias' judgements and duration of intervention.

Wash‐out and extended follow‐up

No studies reported on two‐hour glucose for wash‐out or extended follow‐up periods.

HbA1c

Intervention periods

All but one of the included studies reported on HbA1c (Xu 2011). Shi 2014 did not report statistical variation for changes from baseline; we imputed these values using IDPP‐2's correlation coefficients. Effects of pioglitazone compared with placebo showed a MD of −0.13% (95% CI −0.20 to −0.07; P < 0.001; 5 studies, 1001 participants; Analysis 2.28), in favour of pioglitazone. The 95% prediction interval ranged between −0.23% and −0.03%.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT' with 'IFG or IGT', or IGT only did not show subgroup interaction (P = 0.09; Analysis 2.29).
- Subgrouping by ethnicity of mainly white ethnicity (mixed American races), Asian Indian and assumed Chinese ethnicities did not show subgroup interaction (P = 0.70; Analysis 2.30).
- Subgrouping by sex into studies with men and women versus women only did not show subgroup interaction (P = 0.18; Analysis 2.31).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis
- Including only smaller studies (N < 100 participants) did not substantially change the effect estimate (MD −0.07%, 95% CI −0.23 to 0.09; P 0.37; 2 studies, 111 participants; Attallah 2007; Shi 2014).
- Two studies diagnosed participants by WHO 1999 criteria (Attallah 2007; IDPP‐2). Analysing these studies did not substantially change the effect estimate (MD −0.10%, 95% CI −0.24 to 0.04; P = 0.16; 2 studies, 398 participants; Attallah 2007; IDPP‐2). One study applied the ADA 2003 criteria and reported a HbA1c change from baseline of −0.15% (SD 0.2) in 58 participants in the pioglitazone group compared with −0.06% (SD 0.2) in 61 participants in the placebo group (Bone 2013).
- One study only received non‐commercial funding (IDPP‐2). The reported change in HbA1c from baseline was 0.45% (SD 0.8) in 181 participants in the pioglitazone group compared with 0.55% (0.8) in 186 participants in the placebo group (IDPP‐2).
- Including only studies published in English did not substantially change the effect estimate (MD −0.14%, 95% CI −0.21 to −0.07; P < 0.001; 4 studies, 921 participants; ACT NOW; Attallah 2007; Bone 2013; IDPP‐2).
- Sensitivity analysis regarding imputation was identical to the analysis regarding language.
- Including only studies from the USA showed a MD of −0.15% (95% CI −0.24 to −0.05; P = 0.003; 3 studies, 554 participants; ACT NOW; Attallah 2007; Bone 2013).
- The included studies were similar regarding publication status, 'Risk of bias' judgements and duration of intervention.

Wash‐out and extended follow‐up

One study reported change in HbA1c from baseline to the end of the wash‐out period (Bone 2013). The reported change showed a difference of −0.07% (SD 0.03) in 57 participants in the pioglitazone group compared with −0.03% (SD 0)% in 53 participants in the placebo group (Bone 2013).

Socioeconomic effects

None of the included studies reported on socioeconomic effects.

Pioglitazone monotherapy versus no intervention

Twenty‐one studies compared pioglitazone with no intervention; all but one (Han 2007), had matching behaviour‐changing interventions. Five studies applied identical antihypertensive medication to intervention and comparator groups (Che 2014; Chen 2007b; Gao 2011; Han 2007; Wu 2013; see Characteristics of included studies). One of the 21 studies reported a run‐in period (Xu 2011). One study applied a daily dose of 38 mg (Zeng 2013). Six studies applied a dose of 30 mg once daily (Che 2014; Deng 2013; Gao 2011; Guo 2009; Guo 2010; Tian 2015; Wu 2013). Thirteen studies applied a dose of 15 mg once daily (Chen 2007a; Chen 2007b; Fang 2013; Han 2007; Ke 2006; Li 2017; Liang 2004; Tian 2015; Xiu 2015; Yi 2015; Zhang 2007; Zhang 2015; Zhao 2009). One study applied a daily dose of 4 mg (Yu 2011). None of the included studies reported any wash‐out or extension periods. We judged all 21 studies as having a high risk of bias in at least one risk of bias domain (see Figure 3). For an overview of major results see summary of findings Table 4.

Primary outcomes

All‐cause mortality

Three of the 21 studies reported on all‐cause mortality (Chen 2007b; Deng 2013; Gao 2011).

A total of 11 deaths occurred among 441 participants randomised to the pioglitazone groups compared with 12 out of 425 in comparator groups (RR 0.85, 95% CI 0.38 to 1.91; P = 0.70; 3 studies, 866 participants; very low‐certainty evidence; Analysis 3.1). The 95% prediction interval did not provide a meaningful estimate.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG or IGT' with IGT only did not show subgroup interaction (P = 0.76; Analysis 3.2).
- Subgroup analysis comparing studies with comorbidity as part of the inclusion criteria (hypertension) compared with no comorbidity in the inclusion criteria showed no interaction between subgroups (P = 0.90; Analysis 3.3).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, sex, ethnicity and previous gestational diabetes were not possible.
Sensitivity analysis
- One study diagnosed participants by the WHO 1999 criteria and reported that 2 out of 50 participants died in the pioglitazone group compared with 3 out of 50 participants in the comparator group (Gao 2011). One study diagnosed participants by the ADA 2010 criteria and reported that 4 out of 303 participants died in the pioglitazone group compared with 5 out of 302 participants in the comparator group (Deng 2013).
- The included studies were similar regarding publication status, 'Risk of bias' judgements, length and size, language, imputation, source of funding and country.

Incidence of T2DM

Sixteen of the 21 studies reported on the incidence of T2DM (Che 2014; Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Ke 2006; Li 2017; Liang 2004; Wu 2013; Xiu 2015; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015).

A total of 60 participants out of 1034 participants in the pioglitazone groups developed T2DM compared with 197 out of 1019 participants in the comparator groups (RR 0.31, 95% CI 0.23 to 0.40; P < 0.001; 16 studies, 2053 participants; moderate‐certainty evidence; Analysis 3.4), in favour of pioglitazone. The 95% prediction interval ranged between 0.23 and 0.41.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT', 'IFG or IGT', or IGT only did not show subgroup interaction (P = 0.79; Analysis 3.5).
- Subgrouping by age showed no interaction between subgroups (P = 0.76; Analysis 3.6).
- Subgroup analysis comparing studies with comorbidity as part of inclusion criteria (hypertension) versus no comorbidity in inclusion criteria showed no interaction (P = 0.56; Analysis 3.7).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, sex, ethnicity and previous gestational diabetes were not possible.
Sensitivity analysis
- Including only smaller‐sized studies (N < 100 participants) did not substantially change the effect estimate (RR 0.25, 95% CI 0.15 to 0.44; P < 0.001; 7 studies, 464 participants; Che 2014; Ke 2006; Tian 2015; Wu 2013; Xiu 2015; Yi 2015; Zhang 2015).
- Nine studies diagnosed participants by WHO 1999 criteria (Che 2014; Chen 2007a; Gao 2011; Ke 2006; Liang 2004; Xiu 2015; Yi 2015; Yu 2011; Zeng 2013). Analysing these studies did not substantially change the effect estimate (RR 0.24, 95% CI 0.15 to 0.38; P < 0.001; 9 studies, 802 participants). Four studies diagnosed participants by ADA 2010 criteria (Deng 2013; Wu 2013; Zhang 2007; Zhang 2015). Including only these studies did not substantially change the effect estimate (RR 0.33, 95% CI 0.21 to 0.50; P < 0.001; 4 studies, 825 participants). One study diagnosed participants by ADA 2003 criteria and reported that 5 out of 60 participants developed T2DM in the pioglitazone group compared with 12 out of 60 participants in the no‐intervention group (Fang 2013).
- Including only studies that reported on funding did not substantially change the effect estimate (RR 0.41, 95% CI 0.18 to 0.94; P = 0.03; 2 studies, 187 participants; Fang 2013; Wu 2013).
- The included studies were similar regarding publication status, 'Risk of bias' judgements, length, imputation, language and country.

Serious adverse events

Seven out of 21 studies reported on serious adverse events (Chen 2007a; Chen 2007b; Deng 2013; Gao 2011; Zhang 2007; Zhang 2015; Zhao 2009). Three of these reported actual events (Chen 2007b; Deng 2013; Gao 2011).

A total of 16 out of 610 participants in the pioglitazone groups experienced serious adverse events compared with 21 out of 601 in the comparator groups (RR 0.71, 95% CI 0.38 to 1.32; P = 0.28; 7 studies, 1211 participants; low‐certainty evidence; Analysis 3.8). The 95% prediction interval ranged between 0.32 and 1.60.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT' with IGT only did not show subgroup interaction (P = 0.94; Analysis 3.9).
- Subgroup analysis comparing studies with comorbidity as part of inclusion criteria (hypertension) versus no comorbidity in inclusion criteria showed no subgroup interaction (P = 0.68; Analysis 3.10).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, sex, ethnicity and previous gestational diabetes were not possible.
Sensitivity analysis
- Three studies diagnosed participants by WHO 1999 criteria (Chen 2007a; Gao 2011; Zhao 2009). The three studies reported that a total of 2 out of 147 participants experienced serious adverse events in the pioglitazone groups compared with 3 out of 145 participants in the no‐intervention groups. Three studies diagnosed participants by ADA 2010 criteria (Deng 2013; Zhang 2007; Zhang 2015). The three studies reported that a total of 6 out of 380 participants experienced serious adverse events in the pioglitazone groups compared with 3 out of 383 participants in the no‐intervention groups (Deng 2013; Zhang 2007; Zhang 2015).
- The included studies were similar regarding publication status, 'Risk of bias' judgements, size and length, language, source of funding, country and no imputation was done.

Secondary outcomes

None of the included studies that compared pioglitazone with no intervention reported on cardiovascular mortality, non‐fatal myocardial infarction, non‐fatal stroke, amputation of lower extremity, blindness or severe vision loss, end‐stage renal disease, hypoglycaemia, health‐related quality of life, time to progression to T2DM or socioeconomic effects.

Cardiovascular events

One study reported unspecified coronary heart disease events in 3 out of 83 participants in the pioglitazone group compared with 7 events in 73 participants in the no‐intervention group (Chen 2007b). One study reported 27 unspecified cardiovascular events in 303 participants in the pioglitazone group compared with 23 events among 302 participants in the no‐intervention group (Deng 2013). One study reported two unspecified events of coronary heart disease among 50 participants in the pioglitazone group compared with six events among 50 participants in the placebo group (Gao 2011). We contacted the study authors but did not receive a reply (see Appendix 21).

One study reported one unspecified cerebrovascular disease event among 50 participants in the pioglitazone group compared with three events among 50 participants in the no‐intervention group (Gao 2011).

Congestive heart failure

Deng 2013 reported on congestive heart failure. Two out of 303 participants in the pioglitazone group experienced congestive heart failure compared with 2 out of 302 participants in the no‐intervention group.

Non‐serious adverse events

Eleven studies reported non‐serious adverse events (Chen 2007a; Chen 2007b; Deng 2013; Fang 2013; Gao 2011; Han 2007; Ke 2006; Li 2017; Liang 2004; Zhang 2015; Zhao 2009).

A total of 189 out of 781 participants experienced non‐serious adverse events in the pioglitazone groups compared with 151 out of 765 participants in the no‐intervention groups (RR 1.26, 95% CI 0.81 to 1.97; P = 0.31; 11 studies, 1546 participants; Analysis 3.11). The 95% prediction interval ranged between 0.49 and 3.22.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT', 'IFG or IGT', and IGT only did not show subgroup interaction (P = 0.47; Analysis 3.12).
- Subgroup analysis comparing studies with hypertension as part of the inclusion criteria versus no hypertension as an inclusion criterion showed interaction between subgroups (P = 0.01; Analysis 3.13). The hypertension subgroup showed fewer non‐serious adverse events in the pioglitazone groups.
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, sex, ethnicity and previous gestational diabetes were not possible.
Sensitivity analysis
- Including only smaller studies (N < 100 participants) changed the effect estimate, substantially favouring no‐intervention (RR 4.05, 95% CI 1.41 to 11.66; P = 0.04; 5 studies, 303 participants; Han 2007; Ke 2006; Liang 2004; Zhang 2015; Zhao 2009).
- Six studies diagnosed participants by WHO 1999 criteria (Chen 2007a; Gao 2011; Han 2007; Ke 2006; Liang 2004; Zhao 2009). Including only these studies did not substantially change the effect estimate (RR 1.90, 95% CI 0.57 to 6.36; P = 0.29; 6 studies, 452 participants). Two studies diagnosed by ADA 2010 criteria (Deng 2013; Zhang 2015). Including these studies only did not substantially influence the effect estimate (RR 1.23, 95% CI 1.02 to 1.47; P = 0.03; 2 studies, 656 participants). One study applied ADA 2003 criteria and reported that 16 out of 60 participants in the pioglitazone group experienced non‐serious adverse events compared with 12 out of 60 participants in the no‐intervention group (Fang 2013).
- Fang 2013 was also the only study that reported its funding source.
- The included studies were similar regarding publication status, length, 'Risk of bias' judgements, language, country and no imputation was done.

Hypoglycaemia

Four studies reported on hypoglycaemic events (Fang 2013; Han 2007; Liang 2004; Zhao 2009). Two studies reported actual events (Fang 2013; Liang 2004).

A total of 8 out of 159 participants experienced hypoglycaemia in the pioglitazone groups compared with 3 out of 155 participants in the no‐intervention groups (RR 2.71, 95% CI 0.76 to 9.75; P = 0.13; 4 studies, 314 participants; Analysis 3.14).

Subgroup analysis
- Comparing 'prediabetes' criteria of IGT only with a pooled population of IFG only, IGT only and 'IFG and IGT' did not show subgroup interaction (P = 0.91; Analysis 3.15).
- Subgroup analyses for duration of intervention, mono/dual‐therapy, age, sex, ethnicity, comorbidity and previous gestational diabetes were not possible.
Sensitivity analysis
- Three studies were of smaller size (N < 100 participants; Han 2007; Liang 2004; Zhao 2009). They reported a total of 5 out of 99 participants with hypoglycaemia in the pioglitazone groups compared with 2 out of 95 participants in the no‐intervention groups.
- One study reported on funding and noted that 3 out of 60 participants experienced hypoglycaemia in the pioglitazone group compared with 1 out of 60 in the no‐intervention groups (Fang 2013).
- Three studies diagnosed participants by WHO 1999 criteria (Han 2007; Liang 2004; Zhao 2009). The analysis was identical to the analysis for smaller‐sized studies (see above). One study diagnosed participants by ADA 2003 (Fang 2013). The sensitivity analysis was identical to analysis for funding status (see above).
- The included studies were similar regarding publication status, length, 'Risk of bias' judgements, language, country and no imputation was done.

Health‐related quality of life

None of the included studies reported on health‐related quality of life.

Time to progression to T2DM

One study reported on time to progression to T2DM (Deng 2013). The time to progression of T2DM showed a HR of 0.28 (95% CI 0.16 to 0.49; P < 0.001; 605 participants). The study did not specify whether the HR was unadjusted or adjusted for any factors. The annual incidence of T2DM across the intervention period with a mean duration of 2.2 years was 2.7% in the pioglitazone group compared with 8.1% in the comparator group.

Measures of blood glucose control

Nineteen studies reported on measures of blood glucose (Che 2014; Chen 2007a; Chen 2007b; Fang 2013; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Tian 2015; Wu 2013; Xiu 2015; Yi 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009). None of the included studies reported statistical variations for the changes from baseline, which we imputed using the correlation coefficient from IDPP‐2.

Fasting plasma glucose

All 19 studies reported FPG. All but one (Chen 2007b), were available for meta‐analysis. The studies showed considerable heterogeneity (I² = 93%). We transformed published SEs in three studies to SDs because the statistical variation appeared extremely small (Chen 2007a; Zhang 2007; Zhang 2015). The overall MD comparing pioglitazone with no intervention was −0.70 mmol/L (95% CI −0.96 to −0.44; P < 0.001; 18 studies, 1533 participants; Analysis 3.16), in favour of pioglitazone. The 95% prediction interval ranged between −1.84 mmol/L and 0.44 mmol/L.

One study reported a decrease in FPG in the pioglitazone group compared with the comparator group (P < 0.05) without specifying values (Chen 2007b).

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT', 'IFG or IGT', IGT only, categories 'unspecified definition' and 'mixed definitions' showed subgroup interaction (P = 0.003; Analysis 3.18) with greater reduction of FPG in the pioglitazone groups using the 'IFG or IGT' criterion compared with the categories 'mixed' and 'unspecified'.
- Subgrouping by age showed no subgroup interaction (P = 0.98; Analysis 3.17).
- Subgroup analysis comparing studies with comorbidity as part of inclusion criteria (hypertension) versus no comorbidity in inclusion criteria showed subgroup interaction (P = 0.002; Analysis 3.19). Studies with hypertensive cohorts showed greater reductions of FPG in the pioglitazone groups.
- Subgroup analyses for duration of intervention, mono/dual‐therapy sex, ethnicity and previous gestational diabetes were not possible.
Sensitivity analysis
- Including only smaller‐sized studies (N < 100 participants) did not substantially change the effect estimate (MD −0.79 mmol/L, 95% CI −1.01 to −0.57; P < 0.001; 12 studies, 800 participants; Che 2014; Guo 2009; Guo 2010; Han 2007; Ke 2006; Liang 2004; Tian 2015; Wu 2013; Xiu 2015; Yi 2015; Zhang 2015; Zhao 2009).
- Thirteen studies diagnosed participants by WHO 1999 criteria (Che 2014; Chen 2007a; Guo 2009; Guo 2010; Han 2007; Ke 2006; Liang 2004; Tian 2015; Xiu 2015; Yi 2015; Yu 2011; Zeng 2013; Zhao 2009). Including these studies only did not substantially change the effect estimate (MD −0.60 mmol/L, 95% CI −0.92 to −0.28; P < 0.001; 13 studies, 1037 participants). Three studies diagnosed participants by ADA 2010 criteria (Wu 2013; Zhang 2007; Zhang 2015). Including these studies only did not substantially change the effect estimate (MD −1.22 mmol/L, 95% CI −1.45 to −1.00; P < 0.001; 3 studies, 226 participants). One study applied the ADA 2003 criteria (Fang 2013). The reported change in FPG from baseline was −0.5 mmol/L (SD 0.6) in 60 participants in the pioglitazone group compared with −0.1 mmol/L (SD 0.4) in 60 participants in the no‐intervention group (Fang 2013).
- Restricting meta‐analysis to studies reporting on funding did not substantially change the effect estimate (MD −0.77 mmol/L, 95% CI −1.66 to 0.11; P = 0.09; 2 studies, 187 participants; Fang 2013; Wu 2013).
- The included studies were similar regarding publication status, 'Risk of bias' judgements, length, imputation, language and country.

Two‐hour glucose

Seventeen studies reported on two‐hour glucose (Che 2014; Chen 2007a; Fang 2013; Guo 2009; Guo 2010; Han 2007; Ke 2006; Li 2017; Liang 2004; Tian 2015; Wu 2013; Xiu 2015; Yu 2011; Zeng 2013; Zhang 2007; Zhang 2015; Zhao 2009). We transformed published SEs in three studies to SDs because the statistical variation appeared extremely small (Chen 2007a; Zhang 2007; Zhang 2015). There was considerable heterogeneity (I² = 86%). Comparing pioglitazone with no intervention showed a MD of −1.58 mmol/L (95% CI −1.88 to −1.28; P < 0.001; 17 studies, 1463 participants; Analysis 3.20). The 95% prediction interval ranged between −2.81 mmol/L and −0.35 mmol/L.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT', 'IFG or IGT', IGT only, categories 'unspecified definition' and 'mixed definitions' showed subgroup interaction (P = 0.0007; Analysis 3.21) but only for the category 'unspecified definition' versus IGT only.
- Subgrouping by age showed statistical subgroup interaction (P = 0.02; Analysis 3.23). However, this was not reliable because the CIs slightly overlapped.
- Subgroup analysis comparing studies with comorbidity as part of the inclusion criteria (hypertension) versus no comorbidity in the inclusion criteria showed statistical subgroup interaction (P = 0.04; Analysis 3.24). However, this was not reliable because the CIs slightly overlapped.
- Subgroup analyses for duration of intervention, mono/dual‐therapy sex, ethnicity and previous gestational diabetes were not possible.
Sensitivity analysis
- Including only smaller‐sized studies (N < 100 participants) did not substantially change the effect estimate (MD −1.61 mmol/L, 95% CI −2.11 to −1.10; P < 0.001; 11 studies, 730 participants; Che 2014; Guo 2009; Guo 2010; Han 2007; Ke 2006; Liang 2004; Tian 2015; Wu 2013; Xiu 2015; Zhang 2015; Zhao 2009).
- Twelve studies diagnosed participants by WHO 1999 criteria (Che 2014; Chen 2007a; Guo 2009; Guo 2010; Han 2007; Ke 2006; Liang 2004; Tian 2015; Xiu 2015; Yu 2011; Zeng 2013; Zhao 2009). Restricting meta‐analysis to these studies did not substantially change the effect estimate (MD −1.55 mmol/L, 95% CI −1.92 to −1.18; P < 0.001; 12 studies, 967 participants). Three studies diagnosed participants by ADA 2010 criteria (Wu 2013; Zhang 2007; Zhang 2015). Including these studies only did not substantially change the effect estimate (MD −1.93 mmol/L, 95% CI −2.37 to −1.48; P < 0.001; 3 studies, 226 participants). One study applied the ADA 2003 criteria (Fang 2013). The reported change in two‐hour glucose from baseline was −1.0 mmol/L (SD 1.4) in 60 participants in the pioglitazone group compared with −0.1 mmol/L (SD 1.2) in 60 participants in the no‐intervention groups.
- Including only studies that reported funding did not change the effect estimate (MD −1.54 mmol/L, 95% CI −2.92 to −0.17; P = 0.03; 2 studies, 187 participants; Fang 2013; Wu 2013). The included studies were similar regarding publication status, 'Risk of bias' judgements, length, imputation, language and country.

HbA1c

Nine studies reported on HbA1c (Che 2014; Fang 2013 Guo 2009; Guo 2010; Ke 2006; Li 2017; Liang 2004; Wu 2013; Zhang 2007). We transformed published SEs in two studies to SDs because statistical variation appeared extremely small (Fang 2013; Zhang 2007). There was considerable heterogeneity (I² = 92%). Comparing pioglitazone with no intervention showed a MD of −0.77% (95% CI −1.32 to −0.23; P < 0.005; 9 studies, 762 participants; Analysis 3.25). The 95% prediction interval ranged between −2.71% and to 1.17%.

Subgroup analysis
- Comparing 'prediabetes' criteria of 'IFG plus IGT'. IGT only or category 'unspecified' did not show subgroup interaction (P = 0.11; Analysis 3.26).
- Subgrouping by age showed no subgroup interaction (P = 0.80; Analysis 3.27).
- Subgroup analysis comparing studies with comorbidity as part of the inclusion criteria (hypertension) versus no comorbidity in the inclusion criteria showed no subgroup interaction (P = 0.67; Analysis 3.28).
- Subgroup analyses for duration of intervention, mono/dual‐therapy sex, ethnicity and previous gestational diabetes were not possible.
Sensitivity analysis
- Including only smaller‐sized studies (N < 100 participants) did not substantially change the effect estimate (MD −0.50%, 95% CI −0.68 to −0.32; P < 0.001; 6 studies, 385 participants; Che 2014; Guo 2009; Guo 2010; Ke 2006; Liang 2004; Wu 2013).
- Five studies diagnosed participants by WHO 1999 criteria (Che 2014; Guo 2009; Guo 2010; Ke 2006; Liang 2004). Including only these studies did not substantially change the effect estimate (MD −0.43%, 95% CI −0.63 to −0.24; P < 0.001; 5 studies, 318 participants). Two studies applied the ADA 2010 criteria (Wu 2013; Zhang 2007). Only including theses studies did not substantially change the effect estimate (MD −0.69%, 95% CI −1.18 to −0.20; P = 0.006; 2 studies, 174 participants). One study applied the ADA 2003 criteria (Fang 2013). The reported change in HbA1c from baseline was −1.5% (SD 1.0) in 60 participants in the pioglitazone group compared with −0.1% (SD 1.9) in 60 participants in the no‐intervention group (Fang 2013).
- Including only studies that reported funding did not substantially change the effect estimate (MD −1.15%, 95% CI −1.66 to −0.63; P < 0.001; 2 studies, 187 participants; Fang 2013; Wu 2013).
- Included studies were similar regarding publication status, 'Risk of bias' judgements, length, imputation, language and country.

Socioeconomic effects

None of the included studies reported on socioeconomic effects.

Pioglitazone as a part of a dual combination therapy versus any pharmacological glucose‐lowering agent (e.g. acarbose, metformin, sulphonylurea)

We identified no studies investigating pioglitazone as part of a dual combination therapy compared with another pharmacological glucose‐lowering agent.

Ongoing studies

We identified two ongoing studies, one comparing pioglitazone with placebo and one comparing pioglitazone with other glucose‐lowering drugs, which could provide data of interest for this review (Beijing prediabetes reversion programme (BPRP); NCT02969798). The studies plan to include 2645 randomised participants. Both studies will evaluate one or more of the primary or secondary outcomes of interest in this review. The durations of the studies are estimated to be 24 months (NCT02969798) and three years (Beijing prediabetes reversion programme (BPRP)).

One study has published baseline results and is in the process of publishing a final manuscript (Beijing prediabetes reversion programme (BPRP)). The other study is estimated to finish enrolment in 2023 (NCT02969798).

Studies awaiting assessment

We identified no RCTs that are awaiting assessment.

Discussion

Summary of main results

This Cochrane Review is the first systematic review investigating the effects of pioglitazone in people at increased risk of developing T2DM compared with other glucose‐lowering pharmacological interventions, behaviour‐changing interventions, placebo or no intervention. We included 27 studies with a total of 4186 participants. We judged all included studies as having high risk of bias in at least one 'Risk of bias' domain.

Pioglitazone compared with placebo reduced or delayed the incidence of developing T2DM (low‐certainty evidence), but the intervention effect disappeared after an extended post‐follow‐up period. There was limited information on beneficial or harmful effects of pioglitazone compared with placebo regarding all‐cause mortality, serious adverse events and macro‐ and microvascular complications (very low‐certainty evidence). Notably, the largest study for this comparison reported adverse events without distinguishing between serious and non‐serious events (ACT NOW). The study showed a decreased risk of adverse events with pioglitazone compared with placebo. Pioglitazone compared with no intervention showed a reduction or delay in the incidence of developing T2DM (moderate‐certainty evidence). However, none of the studies with this comparison reported T2DM after an observational, drug‐free study period. The comparison with no intervention did not show any beneficial or harmful effects regarding all‐cause mortality (very low‐certainty evidence) or serious adverse events (low‐certainty evidence), the reporting of these outcomes was sparse. None of the included studies reported on macro‐ or microvascular complications for the comparison with no intervention. Pioglitazone compared with metformin did not reduce or delay the incidence of T2DM (low‐certainty evidence), data on other outcomes for the comparison were minimal. We found only one study comparing pioglitazone with a behaviour‐changing intervention, and no studies compared pioglitazone as part of dual therapy with other glucose‐lowering drugs. None of the included studies reported on health‐related quality of life or socioeconomic effects of pioglitazone intervention.

Overall completeness and applicability of evidence

We performed an extensive search for studies with no restriction on language. We tried to contact all authors of included studies for additional data and questions regarding 'Risk of bias' assessments. We did not receive any information from the majority of study authors. One study author provided additional data (Attallah 2007). We also contacted one pharmaceutical company for additional data on two included studies (ACT NOW; Bone 2013), without further data being provided. We screened other reviews and meta‐analyses for additional references and cross‐checked our extracted data (if the reviews included any studies included in this review), but we found no additional references or data in this way (Haw 2017; Merlotti 2014; Norris 2007).

The included studies' definition of 'prediabetes' varied. Most studies used criteria issued by medical associations for definitions of IFG and IGT (ADA 2003; ADA 2010; WHO 1999). Studies using these criteria included participants based on: people with IFG plus IGT; people with IFG or IGT; or people with IGT only. One study used the authors' own definition of 'prediabetes' (ACT NOW). Two studies did not report on their definitions of people at increased risk for development of T2DM (Chen 2007b; Li 2017). None of the included studies included participants on the basis of elevated HbA1c.

The vast majority of participants in this review were from Chinese studies without specification on ethnicity. We assumed participants of these studies to be of Chinese ethnicity. Three studies included people of mixed ethnicities, mainly white ethnicity from the USA (ACT NOW; Attallah 2007; Bone 2013). One study included people of Asian Indian ethnicity (IDPP‐2). Therefore, not all ethnicities were represented in the included studies.

The applied dosages of pioglitazone varied from 4 mg once daily to 45 mg once daily. All but three studies included some degree of diet and exercise intervention (Attallah 2007; Bone 2013; Han 2007). Information about these behaviour‐changing interventions was generally minimal. One study included only women (Bone 2013), the remaining studies included men and women.

A substantial amount of studies did not report on key outcomes for this review. Only 7 out of 27 studies reported on all‐cause mortality (ACT NOW; Attallah 2007; Bone 2013; Chen 2007b; Deng 2013; Gao 2011; IDPP‐2). Five studies did not report on incidence of T2DM (Guo 2009; Guo 2010; Han 2007; Tian 2015; Zhao 2009). Eleven out of 27 studies reported on serious adverse events (ACT NOW; Attallah 2007; Bone 2013; Chen 2007a; Chen 2007b; Deng 2013; Gao 2011; IDPP‐2; Zhang 2007; Zhang 2015; Zhao 2009). Nine out of 27 studies did not account for participants or reported on losses to follow‐up or missing data (Che 2014; Deng 2013; Guo 2009; Li 2017; Shi 2014; Tian 2015; Wu 2013; Yu 2011; Zhao 2009). The overall rate of reporting on all‐cause mortality, complications of T2DM and key study characteristics was low among the 23 studies conducted in China.

Quality of the evidence

We judged none of the 27 included studies as having low risk of bias across all 'Risk of bias' domains (Figure 3). Four out of 27 studies reported sufficient information on randomisation sequence generation and allocation concealment and we judged them as having low risk of selection bias (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2). Only four studies reported any blinding of participants, personnel and outcome assessors (ACT NOW; Attallah 2007; Bone 2013; IDPP‐2). We judged all studies as having low risk of performance or detection bias with regard to all‐cause mortality, incidence of T2DM, serious adverse events and measure of blood glucose control. These outcomes are objective and unlikely to be influenced by a lack of blinding. We judged all of the included studies as unclear or high risk of attrition bias for at least one outcome. Main reasons for judgement were substantial or unbalanced dropout rates, no report on missing data or handling of missing data. We were only able to retrieve one protocol for our included studies (ACT NOW). We judged all but five studies as having a high risk of outcome reporting bias (Attallah 2007; Bone 2013; IDPP‐2; Liang 2004; Zhang 2015), mainly due to not reporting on key outcomes for this review, which we would have expected for these kinds of studies (see Appendix 16). Six out of 27 studies reported some details on funding (ACT NOW; Attallah 2007; Bone 2013; Fang 2013; IDPP‐2; Wu 2013). Three studies received financial funding, free drugs or both from pharmaceutical companies (ACT NOW; Attallah 2007; Bone 2013). A prior Cochrane Review has shown that studies with commercial funding are more likely to report favourable outcomes for the intervention (Lundh 2017). Overall reporting of key study characteristics for 'Risk of bias' assessment and outcome reporting was sparse and the main contributor for the uncertainty of our effect estimates.

We judged the incidence of T2DM in comparison with no intervention as moderate‐certainty evidence. All other outcomes across all comparisons we judged as low‐ to very low‐certainty evidence, mainly due to few studies reporting outcomes and various 'Risk of bias' assessments (see Appendix 1; Appendix 2; Appendix 3; Appendix 4; Appendix 5; Appendix 6).

Potential biases in the review process

Out of 27 included studies only two study authors supplied contact information and replied to our inquiry about their study (ACT NOW; Attallah 2007). Therefore, a substantial part of our 'Risk of bias' assessment is judged unclear due to missing information. Most of the studies included in this review were underpowered, with a small number of participants and relatively short timeframe (< 3 years). The meta‐analyses were limited in their assessment of specific clinical characteristics due to not being able to retrieve individual participant data for any study. Few of the included studies reported the assessment and diagnosis of T2DM (e.g. by OGTT or FPG, and/or multiple tests on multiple days should be undertaken for certainty of diagnosis) besides cut‐off values. The diagnosis could have been subject to heterogeneity across studies and could affect the meta‐analysis. Only three of 27 included studies reported on their assessment of serious adverse events and stated that an external adjudication panel validated events. This could affect the effect estimates, depending on the remaining studies' assessment of serious adverse events. We excluded studies with participants with 'prediabetes' and concomitant diseases (e.g. non‐alcoholic steatohepatitis and prior stroke) except for hypertension. We examined the potential effects of hypertension by subgroup analysis. We limited the minimum intervention period for studies to 24 weeks and we excluded four studies based on this criteria (EudraCT2006‐002084‐49; Liu 2015; NCT00470262; NCT00633282).

Two review authors independently carried out data extraction for studies. However, authors were not blinded to which study they were extracting data from.

Agreements and disagreements with other studies or reviews

As T2DM is a major health problem worldwide (IDF 2013), interest in identifying people at increased risk of developing T2DM, and early intervention has increased. Several RCTs have assessed the effects of different pharmacological glucose‐lowering interventions for the prevention of T2DM (ACT NOW; Diabetes Prevention Program 2009; Pi‐Sunyer 2015). A pharmacological approach to the prevention or delay of T2DM is appealing to both the clinician and the pharmaceutical industry. However, although a reduction in or delay of the incidence of T2DM is important, the major public health impact of prevention studies will be determined by the extent to which prevention or delay of T2DM translates into a reduction in diabetes‐specific macrovascular and microvascular complications. Furthermore, the individual psychological effects of having a condition and receiving drugs, when potentially unnecessary, need to be discussed.

A recent meta‐analysis investigated the effects of glucose‐lowering interventions after treatment withdrawal in people at high risk of developing T2DM (Haw 2017). The review showed that the reduction in the incidence of T2DM was not sustained after drug intervention withdrawal (mixed group of glucose‐lowering drugs, no studies included with prior pioglitazone intervention). In the present review, we saw the same tendency, with three studies that compared pioglitazone to placebo and extended follow‐up after intervention end (ACT NOW; Attallah 2007; Bone 2013).

Other recent meta‐analyses have examined the effects of different interventions for the prevention of T2DM in people at increased risk (Merlotti 2014; Pang 2018). They have included glitazones as a group instead of focusing on pioglitazone. The results suggested for glitazones as a group were broadly similar to the results of our review. Although, the results are not completely comparable.

One older review investigated the effects of pioglitazone and rosiglitazone in people with T2DM, 'prediabetes' or the metabolic syndrome (Norris 2007). The review did not draw any conclusion for people at increased risk of development of T2DM due to insufficient data on the effects of pioglitazone in people with 'prediabetes'.

Figure 1

Study flow diagram

HTA: health technology assessment; RCT: randomised controlled trial: Screen4Me: Cochrane´s screening service.

Ir a la figura de la revisiónAbrir en una pestaña nueva

Figure 2

Ir a la figura de la revisiónAbrir en una pestaña nueva

Figure 3

Risk of bias summary: review authors' judgements about each risk of bias item for each included study (blank cells indicate that the particular outcome was not measured in some studies)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.1

Comparison 1: Pioglitazone versus another glucose‐lowering intervention, Outcome 1: Incidence of T2DM (comparator metformin)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.2

Comparison 1: Pioglitazone versus another glucose‐lowering intervention, Outcome 2: Incidence of T2DM by 'prediabetes' criteria (comparator metformin)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.3

Comparison 1: Pioglitazone versus another glucose‐lowering intervention, Outcome 3: Non‐serious adverse events (comparator metformin)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.4

Comparison 1: Pioglitazone versus another glucose‐lowering intervention, Outcome 4: Non‐serious adverse events by 'prediabetes' criteria (comparator metformin)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.5

Comparison 1: Pioglitazone versus another glucose‐lowering intervention, Outcome 5: Fasting plasma (comparator metformin)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.6

Comparison 1: Pioglitazone versus another glucose‐lowering intervention, Outcome 6: Fasting plasma glucose by 'prediabetes' criteria (comparator metformin)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.7

Comparison 1: Pioglitazone versus another glucose‐lowering intervention, Outcome 7: Fasting plasma glucose by comorbidity (comparator metformin)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.8

Comparison 1: Pioglitazone versus another glucose‐lowering intervention, Outcome 8: 2‐hour blood glucose (comparator metformin)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.9

Comparison 1: Pioglitazone versus another glucose‐lowering intervention, Outcome 9: 2‐hour blood glucose by 'prediabetes' criteria (comparator metformin)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.10

Comparison 1: Pioglitazone versus another glucose‐lowering intervention, Outcome 10: 2‐hour blood glucose by comorbidity (comparator metformin)

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.1

Comparison 2: Pioglitazone versus placebo, Outcome 1: All‐cause mortality

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.2

Comparison 2: Pioglitazone versus placebo, Outcome 2: All‐cause mortality by subgroup

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.3

Comparison 2: Pioglitazone versus placebo, Outcome 3: Incidence of T2DM

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.4

Comparison 2: Pioglitazone versus placebo, Outcome 4: Incidence of T2DM by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.5

Comparison 2: Pioglitazone versus placebo, Outcome 5: Incidence of T2DM by ethnicity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.6

Comparison 2: Pioglitazone versus placebo, Outcome 6: Incidence of T2DM by age

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.7

Comparison 2: Pioglitazone versus placebo, Outcome 7: Incidence of T2DM by sex

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.8

Comparison 2: Pioglitazone versus placebo, Outcome 8: Incidence of T2DM after intervention end, assessed at any time after intervention end

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.9

Comparison 2: Pioglitazone versus placebo, Outcome 9: Incidence of T2DM after intervention end, assessed at any time after intervention end by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.10

Comparison 2: Pioglitazone versus placebo, Outcome 10: Serious adverse events

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.11

Comparison 2: Pioglitazone versus placebo, Outcome 11: Cardiovascular mortality

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.12

Comparison 2: Pioglitazone versus placebo, Outcome 12: Non‐fatal myocardial infarction

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.13

Comparison 2: Pioglitazone versus placebo, Outcome 13: Congestive heart failure

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.14

Comparison 2: Pioglitazone versus placebo, Outcome 14: Congestive heart failure by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.15

Comparison 2: Pioglitazone versus placebo, Outcome 15: Congestive heart failure by sex

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.16

Comparison 2: Pioglitazone versus placebo, Outcome 16: Non‐serious adverse events

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.17

Comparison 2: Pioglitazone versus placebo, Outcome 17: Non‐serious adverse events by 'prediabetes' criterion

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.18

Comparison 2: Pioglitazone versus placebo, Outcome 18: Time to progression of T2DM

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.19

Comparison 2: Pioglitazone versus placebo, Outcome 19: Time to progression of T2DM by subgroup

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.20

Comparison 2: Pioglitazone versus placebo, Outcome 20: Fasting plasma glucose

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.21

Comparison 2: Pioglitazone versus placebo, Outcome 21: Fasting plasma glucose by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.22

Comparison 2: Pioglitazone versus placebo, Outcome 22: Fasting plasma glucose by ethnicity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.23

Comparison 2: Pioglitazone versus placebo, Outcome 23: Fasting plasma glucose by sex

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.24

Comparison 2: Pioglitazone versus placebo, Outcome 24: 2‐hour blood glucose

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.25

Comparison 2: Pioglitazone versus placebo, Outcome 25: 2‐hour blood glucose by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.26

Comparison 2: Pioglitazone versus placebo, Outcome 26: 2‐hour blood glucose by ethnicity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.27

Comparison 2: Pioglitazone versus placebo, Outcome 27: 2‐hour blood glucose by sex

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.28

Comparison 2: Pioglitazone versus placebo, Outcome 28: HbA1C

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.29

Comparison 2: Pioglitazone versus placebo, Outcome 29: HbA1C by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.30

Comparison 2: Pioglitazone versus placebo, Outcome 30: HbA1C by ethnicity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 2.31

Comparison 2: Pioglitazone versus placebo, Outcome 31: HbA1c by sex

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.1

Comparison 3: Pioglitazone versus no intervention, Outcome 1: All‐cause mortality

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.2

Comparison 3: Pioglitazone versus no intervention, Outcome 2: All‐cause mortality by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.3

Comparison 3: Pioglitazone versus no intervention, Outcome 3: All‐cause mortality by comorbidity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.4

Comparison 3: Pioglitazone versus no intervention, Outcome 4: Incidence of T2DM

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.5

Comparison 3: Pioglitazone versus no intervention, Outcome 5: Incidence of T2DM by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.6

Comparison 3: Pioglitazone versus no intervention, Outcome 6: Incidence of T2DM by age

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.7

Comparison 3: Pioglitazone versus no intervention, Outcome 7: Incidence of T2DM by comorbidity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.8

Comparison 3: Pioglitazone versus no intervention, Outcome 8: Serious adverse events

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.9

Comparison 3: Pioglitazone versus no intervention, Outcome 9: Serious adverse events by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.10

Comparison 3: Pioglitazone versus no intervention, Outcome 10: Serious adverse events by comorbidity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.11

Comparison 3: Pioglitazone versus no intervention, Outcome 11: Non‐serious adverse events

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.12

Comparison 3: Pioglitazone versus no intervention, Outcome 12: Non‐serious adverse events by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.13

Comparison 3: Pioglitazone versus no intervention, Outcome 13: Non‐serious adverse events by comorbidity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.14

Comparison 3: Pioglitazone versus no intervention, Outcome 14: Participants with events of hypoglycaemia

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.15

Comparison 3: Pioglitazone versus no intervention, Outcome 15: Participants with events of hypoglycaemia by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.16

Comparison 3: Pioglitazone versus no intervention, Outcome 16: Fasting plasma glucose

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.17

Comparison 3: Pioglitazone versus no intervention, Outcome 17: Fasting plasma glucose by age

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.18

Comparison 3: Pioglitazone versus no intervention, Outcome 18: Fasting plasma glucose by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.19

Comparison 3: Pioglitazone versus no intervention, Outcome 19: Fasting plasma glucose by comorbidity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.20

Comparison 3: Pioglitazone versus no intervention, Outcome 20: 2‐hour blood glucose

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.21

Comparison 3: Pioglitazone versus no intervention, Outcome 21: 2‐hour blood glucose by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.22

Comparison 3: Pioglitazone versus no intervention, Outcome 22: 2‐hour blood glucose by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.23

Comparison 3: Pioglitazone versus no intervention, Outcome 23: 2‐hour blood glucose by age

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.24

Comparison 3: Pioglitazone versus no intervention, Outcome 24: 2‐hour blood glucose by comorbidity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.25

Comparison 3: Pioglitazone versus no intervention, Outcome 25: HbA1C

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.26

Comparison 3: Pioglitazone versus no intervention, Outcome 26: HbA1C by 'prediabetes' criteria

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.27

Comparison 3: Pioglitazone versus no intervention, Outcome 27: HbA1C by age

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 3.28

Comparison 3: Pioglitazone versus no intervention, Outcome 28: HbA1c by comorbidity

Ir a la figura de la revisiónAbrir en una pestaña nueva

Summary of findings 1. Summary of findings: pioglitazone monotherapy versus another pharmacological glucose‐lowering monotherapy

Outcomes	Acarbose, metformin, repaglinide	Pioglitazone	Relative effect (95% CI)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Pioglitazone monotherapy versus another pharmacological glucose‐lowering monotherapy (acarbose, metformin, repaglinide)
Population: people at increased risk of developing T2DM Settings: outpatients Intervention: pioglitazone Comparison: acarbose, metformin, repaglinide
All‐cause mortality
Metformin	Not reported
Acarbose	Not reported
Repaglinide	Not reported
Incidence of T2DM (N)
Metformin Diagnostic criteria: 2 studies applied the WHO 1999 criteria for 'prediabetes' (FPG > 6.1 mmol/L and < 7.0 mmol/L and/or a 2‐h glucose < 11.1 mmol/L after a 75 g OGTT; Chen 2007a; Zeng 2013) 1 study applied the ADA 2010 criteria (FPG ≥ 5.6 mmol/L and < 6.9 mmol/L and 2‐h glucose ≥ 7.8 mmol/L or < 11.0 mmol/L after a 75 g OGTT; Zhang 2007) Follow‐up: 12‐24 months	55 per 1000	54 per 1000 (22 to 131)	RR 0.98 (0.40 to 2.38)	331 (3)	⊕⊕⊝⊝ Low^a	9/168 participants developed T2DM in the pioglitazone groups vs 9/163 participants in the metformin groups
Acarbose Diagnostic criteria: 1 study applied the WHO 1999 criteria for prediabetes (FPG > 6.1 mmol/L and < 7.0 mmol/L and/or a 2‐h glucose < 11.1 mmol/L after a 75 g OGTT; Chen 2007a) Follow‐up: 12 months	See comment			96 (1)	⊕⊝⊝⊝ Verylow^b	1/50 participants in the pioglitazone group vs 2/46 participants in the acarbose group developed T2DM (Chen 2007a)
Repaglinide Diagnostic criteria: 1 study applied the ADA 2010 criteria (FPG ≥ 5.6 mmol/L and < 6.9 mmol/L and 2‐h glucose ≥ 7.8 mmol/L or < 11.0 mmol/L after a 75 g OGTT; Zhang 2007) Follow‐up: 12 months	See comment			96 (1)	⊕⊕⊝⊝ Low^c	2/48 participants in the pioglitazone group vs 1/48 participants in the repaglinide group developed T2DM (Zhang 2007)
Serious adverse events (N)
Metformin Follow‐up: 12‐24 months	See comment			201 (2)	⊕⊕⊝⊝ Low^a	No participant experienced a serious adverse event (Chen 2007a; Zhang 2007)
Acarbose Follow‐up: 12 months	See comment			96 (1)	⊕⊝⊝⊝ Verylow^b	No participant experienced a serious adverse event (Chen 2007a)
Repaglinide Follow‐up: 12 months	See comment			103 (1)	⊕⊕⊝⊝ Low^c	No participant experienced a serious adverse event (Zhang 2007)
Cardiovascular mortality
Metformin	Not reported
Acarbose	Not reported
Repaglinide	Not reported
Non‐fatal myocardial infarction/stroke
Metformin	Not reported
Acarbose	Not reported
Repaglinide	Not reported
Health‐related quality of life
Metformin	Not reported
Acarbose	Not reported
Repaglinide	Not reported
Socioeconomic effects
Metformin	Not reported
Acarbose	Not reported
Repaglinide	Not reported
ADA: American Diabetes Association; CI: confidence interval; FPG: fasting plasma glucose; OGTT: oral glucose tolerance test; RR: risk ratio; HbA1c: glycosylated haemoglobin A1c; T2DM: type 2 diabetes mellitus; WHO: World Health Organization
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded by two levels because of serious imprecision (low median sample size, small number of studies), see Appendix 1. ^bDowngraded by one level because of attrition bias and by two levels because of serious imprecision (one study only, low median sample size), see Appendix 2. ^cDowngraded by two levels because of serious imprecision (one study only, low median sample size), see Appendix 3.

Summary of findings 1. Summary of findings: pioglitazone monotherapy versus another pharmacological glucose‐lowering monotherapy

Ir a la tabla de la revisión

Summary of findings 2. Summary of findings: pioglitazone monotherapy versus intensive behaviour‐changing intervention

Outcomes	Personalised diet + exercise consultation	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Pioglitazone monotherapy versus intensive behaviour‐changing intervention
Population: people at increased risk for developing T2DM Settings: outpatients Intervention: pioglitazone Comparison: intensive diet plus exercise intervention
All‐cause mortality	Not reported
Incidence of T2DM (N) Follow‐up: 12 months	See comment	96^a (1)	⊕⊕⊝⊝ Low^b	2/48 participants developed T2DM in the pioglitazone group vs 5/48 participants in the personalised diet and exercise consultation group (Zhang 2007)
Serious adverse events (N)	See comment	103^a (1)	⊕⊕⊝⊝ Low^b	No participant experienced a serious adverse event (Zhang 2007)
Cardiovascular mortality	Not reported
Non‐fatal myocardial infarction/stroke	Not reported
Health‐related quality of life	Not reported
Socioeconomic effects	Not reported
CI: confidence interval; T2DM: type 2 diabetes mellitus
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
^aOnly people finishing the study were included in analysis of incidence of type 2 diabetes mellitus. All randomised participants were included in the remaining analyses. ^bDowngraded by two levels because of serious imprecision (one study only, low sample size), see Appendix 4.

Summary of findings 2. Summary of findings: pioglitazone monotherapy versus intensive behaviour‐changing intervention

Ir a la tabla de la revisión

Summary of findings 3. Summary of findings: pioglitazone monotherapy versus placebo

Outcomes	Placebo	Pioglitazone	Relative effect (95% CI)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Pioglitazone monotherapy versus placebo (identical behaviour‐changing interventions in both groups)
Population: people at increased risk for developing T2DM Settings: outpatients Intervention: pioglitazone Comparison: placebo
All‐cause mortality (N) Follow‐up: 44 weeks to 2 years	3 per 1000	8 per 1000 (2 to 36)	Peto OR2.38 (0.54 to 10.50)	1156 (4)	⊕⊝⊝⊝ Very low^a	Only 2 studies reported events (ACT NOW; IDPP‐2)
Incidence of T2DM (N) Diagnostic criteria: 3 studies applied the WHO 1999 criteria for 'prediabetes' (FPG > 6.1 mmol/L and < 7.0 mmol/L and/or a 2‐h glucose < 11.1 mmol/L after a 75 g OGTT; Attallah 2007; IDPP‐2; Xu 2011) 1 study applied the ADA 2003 criteria (FPG ≥ 5.3 mmol/L and < 6.9 mmol/L or 2‐h glucose ≥ 7.8 mmol/L and < 11.0 mmol/L after a 75 g OGTT; Bone 2013) 1 study applied the ADA 2010 criteria (FPG ≥ 5.6 mmol/L and < 6.9 mmol/L and 2‐h glucose ≥ 7.8 mmol/L or < 11.0 mmol/L after a 75 g OGTT; Shi 2014) 1 study applied its own definition of 'prediabetes' (FPG = 95‐125 mg/dL and 2‐h glucose ≥ 140 mg/dL or < 199 mg/dL after a 75 g OGTT or FPG = 90‐125 mg/dL and 2‐h glucose of 170‐199 mg/dL plus at least one additional high risk characteristic of diabetes; ACT NOW) Follow‐up: 44 weeks to 2 years	188 per 1000	75 per 1000 (32 to 179)	RR 0.40 (0.17 to 0.95)	1395 (6)	⊕⊕⊝⊝ Low^b	The 95% prediction interval ranged from 0.03 to 4.68 Three studies reported the incidence of T2DM after extended follow‐up periods ranging between 3 weeks and 11.4 months. Data for T2DM at the end of the extension periods showed that the intervention effect was dismissed RR 0.89, 95% CI 0.49 to 1.60 (ACT NOW; Attallah 2007; Bone 2013)
Serious adverse events (N) Follow‐up: 44 weeks to 2 years	See comment		RR3.00 (0.32 to 28.22)	187 (2)	⊕⊝⊝⊝ Very low^c	Only Bone 2013 observed events: 3/93 participants experienced serious adverse events in the pioglitazone group vs 1/94 participants in the placebo group. The largest study for this comparison reported on adverse events without distinguishing between serious and non‐serious events (ACT NOW). 121/303 (39.9%) participants in the pioglitazone group compared with 151/299 (50.5%) participants in the placebo group experienced an adverse event (P = 0.03; ACT NOW)
Cardiovascular mortality (N) Follow‐up: 44 weeks to 2 years	See comment		RR 5.14 (0.25 to 106.28)	1156 (4)	⊕⊝⊝⊝ Very low^d	Only IDPP‐2 observed events: 2/181 participants in the pioglitazone group vs 0/186 participants in the placebo group 3 studies reported no cardiovascular deaths
Non‐fatal myocardial infarction/stroke (N) Follow‐up: 44 weeks to 2 years	See comment		RR 1.97 (0.18 to 21.65)	789 (3)	⊕⊝⊝⊝ very low^a	Only ACT NOW observed non‐fatal myocardial infarction events: 2/303 participants in the pioglitazone group vs 1/299 participants in the placebo group 2 studies did not observe non‐fatal myocardial infarctions Non‐fatal strokes were not observed
Health‐related quality of life	Not reported
Socioeconomic effects	Not reported
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). ADA: American Diabetes Association; CI: confidence interval;FPG: fasting plasma glucose; OGTT: oral glucose tolerance test; OR: odds ratio; RR: risk ratio; T2DM: type 2 diabetes mellitus; WHO: World Health Organization
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded by one level because of attrition bias and indirectness (insufficient time frame) and by two levels because of serious imprecision (95% confidence interval consistent with benefit and harm, small number of studies), see Appendix 5. ^bDowngraded by one level because of attrition bias and by one level because of imprecision (low median sample size), see Appendix 5. ^cDowngraded by one level because of attrition bias and by two levels because of severe imprecision (95% confidence interval consistent with benefit and harm, low median sample size, small number of studies), see Appendix 5.

Summary of findings 3. Summary of findings: pioglitazone monotherapy versus placebo

Ir a la tabla de la revisión

Summary of findings 4. Summary of findings: pioglitazone monotherapy versus no intervention

Outcomes	No intervention	Pioglitazone	Relative effect (95% CI)	Number of participants (studies)	Certainty of the evidence (GRADE)	Comments
Pioglitazone as monotherapy versus no intervention (identical behaviour‐changing interventions in both groups)
Population: people at increased risk of developing T2DM Settings: outpatients (20 studies), inpatients (1 study) Intervention: pioglitazone Comparison: no intervention
All‐cause mortality (N) Follow‐up: 24 weeks to mean of 2.2 years	28 per 1000	24 per 1000 (11 to 54)	RR 0.85 (0.38 to 1.91)	866 (3)	⊕⊝⊝⊝ Verylow^a
Incidence of T2DM (N) Diagnostic criteria: 9 studies applied the WHO 1999 criteria for 'prediabetes' (FPG > 6.1 mmol/L and < 7.0 mmol/L and/or a 2‐h glucose < 11.1 mmol/L after a 75 g OGTT; Che 2014; Chen 2007a; Gao 2011; Ke 2006; Liang 2004; Xiu 2015; Yi 2015; Yu 2011; Zeng 2013) 1 study applied the ADA 2003 criteria (FPG ≥ 5.3 mmol/L and < 6.9 mmol/L or 2‐h glucose ≥ 7.8 mmol/L and < 11.0 mmol/L after a 75 g OGTT; Fang 2013) 4 studies applied the ADA 2010 criteria (FPG ≥ 5.6 mmol/L and < 6.9 mmol/L and 2‐h glucose ≥ 7.8 mmol/L or < 11.0 mmol/L after a 75 g OGTT; Deng 2013; Wu 2013; Zhang 2007; Zhang 2015) 2 studies did not specify their diagnostic criteria (Chen 2007b; Li 2017) Follow‐up: 24 weeks to a mean of 2.2 years	193 per 1000	60 per 1000 (44 to 77)	RR 0.31 (0.23 to 0.40)	2053 (16)	⊕⊕⊕⊝ Moderate^b	The 95% prediction interval ranged between 0.23 and 0.41
Serious adverse events (N) Follow‐up: 6 months to a mean of 2.2 years	35 per 1000	25 per 1000 (13 to 46)	RR 0.71 (0.38 to 1.32)	1211 (7)	⊕⊕⊝⊝ Low^c	The 95% prediction interval ranged between 0.32 and 1.60
Cardiovascular mortality	Not reported
Non‐fatal myocardial infarction/stroke	Not reported
Health‐related quality of life	Not reported
Socioeconomic effects	Not reported
ADA: American Diabetes Association; CI: confidence interval;FPG: fasting plasma glucose; OGTT: oral glucose tolerance test; RR: risk ratio; T2DM: type 2 diabetes mellitus; WHO: World Health Organization
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded by one level because of indirectness (insufficient time frame) and by two levels because of serious imprecision (small sample size, small number of studies, 95% confidence interval consistent with benefit and harm), see Appendix 6. ^bDowngraded by one level because of imprecision (low median sample size) and unclear selection bias. ^cDowngraded by two levels because of imprecision (95% confidence interval consistent with benefit and harm, low median sample size).

Summary of findings 4. Summary of findings: pioglitazone monotherapy versus no intervention

Ir a la tabla de la revisión

Table 1. Overview of study populations

Study ID (study design)	Intervention(s) and comparator(s)	Description of power and sample size calculation	Screened/eligible (N)	Randomised (N)	Analysed (primary outcome) (N)	Finishing study (N)	Randomised finishing study (%)	Follow‐up (extended follow‐up)^a
ACT NOW (parallel RCT)	I: pioglitazone 45 mg/day titrated up from 30 mg/day after 1 month if no adverse effects plus BCI	"Based upon this information (see assumptions below), it can be calculated that approximately 600 subjects with IGT will be required to achieve 90% statistical power that pioglitazone decreases the conversion rate of IGT to type 2 diabetes by 50%. This power calculation assumes that randomized individuals drop out prior to the confirmed diagnosis of diabetes with an exponential hazard rate of 0.10 (or less) per year. The following assumptions were used to calculate the sample size: (i) The primary endpoint is the development of diabetes, (ii) Participants are randomized over a 21 month period and followed for a total of 3.75 years, starting from the time that the first IGT subject is recruited, (iii) Type I error rate (alpha) is 0.05, (iv) The desired power is 90%, (v) The development of diabetes in the placebo‐treated group is 11% per year,(vi) The hazard rate for the development of diabetes in the pioglitazone‐treated group is reduced by 50%. (vii) The drop out rate is 10% per year"	1827/602	303	303	213	70.3	24‐48 months (extended follow‐up with median of 11.4 months)
	C: placebo once daily, placebo dose titrated up after 1 month if no adverse effects plus BCI		1827/602	299	299	228	76.3
	Total:			602	602	441	73.3
Attallah 2007 (2 x 2 factorial RCT)	I: pioglitazone 30 mg/day titrated up from 15 mg/day after 4 weeks plus GH placebo from 4‐40 weeks	"Sample size was determined using data from Johannsson et al. [19], in which nine months of GH resulted in a 17.9% ± 3.5% reduction in visceral fat along with an improvement in insulin sensitivity in abdominally obese men. Using this information, we determined that 12 participants were needed per group to have an 80% chance of detecting a change in visceral fat of at least 17.9%. However, since women were also included in the study and are known to respond less effectively to GH than men, we increased the sample size to 15."	185/81	22	15	15	68.1	43‐44 weeks
	C: pioglitazone placebo plus GH placebo from 4‐40 weeks		185/81	19	16	16	84.2
	Total:			41	31	31	75.6
Bone 2013 (parallel RCT)	I: pioglitazone 45 mg/day titrated up from 30 mg/day plus vitamin D and calcium	"The a priori sample size calculation projected that 75 women per treatment group would be required for the two‐sided 95% CI of the estimated between‐group treatment difference to be within the target precision limits of ± 2.3%, assuming a SD of 6.5% and dropout rate of 18%. In a post hoc power calculation using the observed SD of 3% and 60 women per group, it was determined that the study had greater than 90% power to detect a between group difference in percentage change in BMD from baseline to month 12 if the true difference were 1.7%	386/156	78	78	57	73.1	18 months
	C: matching placebo once daily plus vitamin D and calcium		386/156	78	78	61	76.9
	Total:			156	156	117	75
Che 2014 (parallel RCT)	I: L‐amlodipine 5 mg/day plus pioglitazone 30 mg/day	—	—	38	38	38^b	100^c	6 months
	C: L‐amlodipine 5 mg/day	—	—	35	35	35^b	100^c
	Total:			73	73	73^b	100^c
Chen 2007a (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	50	50	50	100	12 months
	C1: acarbose 150 mg/day plus BCI			50	46	46	92
	C2: metformin 750 mg/day plus BCI			50	46	46	92
	C3: BCI			50	50	50	100
	C4: no intervention			50	50	50	100
	Total:			250	242	242	96.8
Chen 2007b (parallel RCT)	I: pioglitazone 15 mg/day plus captopril 50 mg/day and Betaloc 50 mg/day	—	—	88	83	83	94.3	24 months
	C: captopril 50 mg/day and Betaloc 50 mg/day	—	—	80	73	73	91.3
	Total:			168	156	156	94.0
Deng 2013 (parallel RCT)	I: pioglitazone 30 mg/day plus BCI	—	—	303	303^b	303^b	100^b	Mean of 2.2 years
	C: BCI	—	—	302	302^b	302^b	100^b
	Total:			605	605^c	605^c	100^c
Fang 2013 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	60	60	55	91.7	12 months
	C: BCI	—	—	60	60	56	93.3
	Total:			120	120	111	92.5
Gao 2011 (parallel RCT)	I: pioglitazone 30 mg/day, captopril and other antihypertensive drug(s) plus BCI	—	—	50	50	48	96	18 months
	C: captopril and other antihypertensive drug(s) plus BCI	—	—	50	50	46	92
	Total:			100	100	94	94
Guo 2009 (parallel RCT)	I: pioglitazone 30 mg/day plus BCI	—	—	30	30	30^b	100^b	6 months
	C: BCI	—	—	30	30	30^b	100^b
	Total:			60	60	60^c	100^c
Guo 2010 (parallel RCT)	I: pioglitazone 30 mg/day plus BCI	—	—	32	32	32	100	12 months
	C: BCI	—	—	32	32	32	100
	total:			64	64	64	100
Han 2007 (parallel RCT)	I: pioglitazone 15 mg/day plus enalapril (dose not reported)	—	—	23	23	23	100	6 months
	C: enalapril (dose not reported)	—	—	20	20	20	100
	Total:			43	43	43	100
IDPP‐2 (parallel RCT)	I: pioglitazone 30 mg/day titrated up from 15 mg once daily plus BCI	"It was assumed that the cumulative incidence of diabetes in 3 years would be 40% in the control group with lifestyle modification and placebo and 25% in the group receiving lifestyle modification and pioglitazone. The sample sizes required in each of the two groups were 165 with a type 1 error of 5%, with 80% power. Higher numbers were recruited (204 in group A and 203 in group B) to allow for drop out. The intention‐to‐treat approach was used."	6589/882/407 (double screening procedure)	204	181	181	88.7	36 months
	C: placebo in matching doses plus BCI		6589/882/407 (double screening procedure)	203	186	186	91.6
	Total:			407	367	367	90.2
Ke 2006 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	32	30	30	93.8	12 months
	C: BCI	—	—	30	28	28	93.3
	Total:			62	58	58	93.5
Li 2017 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	75	75	75^b	100c	6 months
	C: BCI	—	—	75	75	75^b	100^c
	Total:			150	150	150^b	100^c
Liang 2004 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	30	29	29	96.7	12 months
	C: BCI	—	—	30	30	30	100
	Total:			60	59	59	98.3
Shi 2014 (parallel RCT)	I: pioglitazone 30 mg/day plus BCI	—	—	40	40	40^b	40^b	10 months
	C: placebo in matching doses plus BCI	—	—	40	40	40^b	40^b
	Total:			80	80	80^c	100^c
Tian 2015 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	36	36	36^b	100^b	24 weeks
	C: BCI	—	—	36	36	36^b	100^b
	Total:			72	72	72^c	100^c
Wu 2013 (parallel RCT)	I: pioglitazone 30 mg/day plus amlodipine 5 mg/day and valsartan 80 mg/day plus BCI	—	—	34	34	34^b	100^c	6 months
	C: amlodipine 5 mg/day and valsartan 80 mg/day plus BCI	—	—	33	33	33^b	100^c
	Total:			67	67	67^c	100^c
Xiu 2015 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	43	43	43	100	48 weeks
	C: BCI	—	—	43	43	43	100
	Total:			86	86	86	100
Xu 2011 (parallel RCT)	I: pioglitazone 30 mg/day plus BCI	—	—	60	60	60	100	48 weeks
	C: placebo in matching doses plus BCI	—	—	59	59	59	100
	Total:			119	119	119	100
Yi 2015 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	35	35	35	100	12 months
	C: BCI	—	—	35	35	35	100
	Total:			70	70	70	100
Yu 2011 (parallel RCT)	I: pioglitazone 4 mg/day plus BCI	—	—	60	60	60^b	100^c	6 months
	C: BCI	—	—	60	60	60^b	100^c
	Total:			120	120	120^b	100^c
Zeng 2013 (parallel RCT)	I1: pioglitazone 38 mg/day plus BCI	—	—	70	70	70	100	24 months
	I2: metformin 38 mg/day plus BCI			68	68	68	100
	C: BCI			66	66	66	100
	Total:			204	204	204	100
Zhang 2007 (parallel RCT)	I: pioglitazone 15 mg/day plus BCI	—	—	52	48	48	100	12 months
	C1: metformin 250 mg/day plus BCI			53	49	49	100
	C2: repaglinide 3 x 0.25 mg/day plus BCI			51	48	48	100
	C3: intensive BCI			51	48	48	100
	C4: BCI			56	54	54	100
	Total:			263	263	247	93.9
Zhang 2015 (parallel RCT)	I: pioglitazone 15 mg/day	—	—	26	26	26	100	6 months
	C: BCI	—	—	26	25	25	96.2
	Total:			52	51	51	98.1
Zhao 2009 (parallel RCT)	I: pioglitazone 15 mg/day	—	—	47	47	47^b	100^c	6 months
	C: BCI	—	—	45	45	45^b	100^c
	Total:			92	92	92^b	100^c

Grand total (all included studies)	All interventions			1921		NA^d
	All comparators			2265		NA^d
	All interventions and comparators			4186		NA^d
Total (studies reporting on dropouts)	All interventions			1258		1098	87.2
	All comparators			1765		1462	82.8
	All interventions and comparators			3023		2560	84.7
— denotes not reported BCI: behaviour‐changing intervention; C: comparator; GH: growth hormone; I: intervention; IGT: impaired glucose tolerance; NA: not applicable; RCT: randomised controlled trial; SD: standard deviation.
^aFollow‐up under randomised conditions until end of study (= duration of intervention + follow‐up post‐intervention or identical to duration of intervention); extended follow‐up refers to follow‐up of participants once the original study was terminated as specified in the power calculation. ^bStudies did not report on dropouts or number of participants finishing study (assumed value). ^cValues calculated include assumptions on finishing rates of studies not reporting on dropout or finishing rates. ^dNot all included studies reported on dropouts.

Table 1. Overview of study populations

Ir a la tabla de la revisión

Comparison 1. Pioglitazone versus another glucose‐lowering intervention

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Incidence of T2DM (comparator metformin) Show forest plot	3	331	Risk Ratio (M‐H, Random, 95% CI)	0.98 [0.40, 2.38]

1.2 Incidence of T2DM by 'prediabetes' criteria (comparator metformin) Show forest plot	3	331	Risk Ratio (M‐H, Random, 95% CI)	0.98 [0.40, 2.38]

1.2.1 Only IGT	2	234	Risk Ratio (M‐H, Random, 95% CI)	0.96 [0.35, 2.64]
1.2.2 IFG and IGT	1	97	Risk Ratio (M‐H, Random, 95% CI)	1.02 [0.15, 6.96]
1.3 Non‐serious adverse events (comparator metformin) Show forest plot	2	201	Risk Ratio (M‐H, Random, 95% CI)	0.48 [0.04, 5.51]

1.4 Non‐serious adverse events by 'prediabetes' criteria (comparator metformin) Show forest plot	2	201	Risk Ratio (M‐H, Random, 95% CI)	0.48 [0.04, 5.51]

1.4.1 Only IGT	1	96	Risk Ratio (M‐H, Random, 95% CI)	0.13 [0.02, 1.03]
1.4.2 IFG and IGT	1	105	Risk Ratio (M‐H, Random, 95% CI)	1.53 [0.27, 8.78]
1.5 Fasting plasma (comparator metformin) Show forest plot	3	339	Mean Difference (IV, Random, 95% CI)	0.03 [‐0.17, 0.23]

1.6 Fasting plasma glucose by 'prediabetes' criteria (comparator metformin) Show forest plot	3	339	Mean Difference (IV, Random, 95% CI)	0.03 [‐0.17, 0.23]

1.6.1 IFG and IGT	1	105	Mean Difference (IV, Random, 95% CI)	‐0.24 [‐0.90, 0.42]
1.6.2 Only IGT	2	234	Mean Difference (IV, Random, 95% CI)	0.06 [‐0.15, 0.26]
1.7 Fasting plasma glucose by comorbidity (comparator metformin) Show forest plot	3	339	Mean Difference (IV, Random, 95% CI)	0.03 [‐0.17, 0.23]

1.7.1 No comorbidity in inclusion criteria	2	243	Mean Difference (IV, Random, 95% CI)	‐0.01 [‐0.24, 0.22]
1.7.2 Hypertension	1	96	Mean Difference (IV, Random, 95% CI)	0.15 [‐0.24, 0.54]
1.8 2‐hour blood glucose (comparator metformin) Show forest plot	3	339	Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.43, ‐0.02]

1.9 2‐hour blood glucose by 'prediabetes' criteria (comparator metformin) Show forest plot	3	339	Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.43, ‐0.02]

1.9.1 IFG and IGT	1	105	Mean Difference (IV, Random, 95% CI)	0.16 [‐0.46, 0.78]
1.9.2 Only IGT	2	234	Mean Difference (IV, Random, 95% CI)	‐0.27 [‐0.47, ‐0.08]
1.10 2‐hour blood glucose by comorbidity (comparator metformin) Show forest plot	3	339	Mean Difference (IV, Random, 95% CI)	‐0.23 [‐0.43, ‐0.02]

1.10.1 No comorbidity in inclusion criteria	2	243	Mean Difference (IV, Random, 95% CI)	‐0.16 [‐0.58, 0.25]
1.10.2 Hypertension	1	96	Mean Difference (IV, Random, 95% CI)	‐0.13 [‐0.63, 0.37]

Comparison 1. Pioglitazone versus another glucose‐lowering intervention

Ir a la tabla de la revisión

Comparison 2. Pioglitazone versus placebo

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 All‐cause mortality Show forest plot	4	1156	Peto Odds Ratio (Peto, Fixed, 95% CI)	2.38 [0.54, 10.50]

2.2 All‐cause mortality by subgroup Show forest plot	2	969	Peto Odds Ratio (Peto, Fixed, 95% CI)	2.38 [0.54, 10.50]

2.2.1 Mainly white ethnicity, from USA, commercial funding, IFG and IGT	1	602	Peto Odds Ratio (Peto, Fixed, 95% CI)	2.70 [0.38, 19.23]
2.2.2 Asian Indian ethnicity, from India, non‐commercial funding, only IGT	1	367	Peto Odds Ratio (Peto, Fixed, 95% CI)	2.01 [0.21, 19.44]
2.3 Incidence of T2DM Show forest plot	6	1395	Risk Ratio (M‐H, Random, 95% CI)	0.40 [0.17, 0.95]

2.4 Incidence of T2DM by 'prediabetes' criteria Show forest plot	6	1395	Risk Ratio (M‐H, Random, 95% CI)	0.40 [0.17, 0.95]

2.4.1 IFG and IGT	3	752	Risk Ratio (M‐H, Random, 95% CI)	0.31 [0.18, 0.52]
2.4.2 IFG or IGT	2	236	Risk Ratio (M‐H, Random, 95% CI)	0.17 [0.04, 0.77]
2.4.3 Only IGT	1	407	Risk Ratio (M‐H, Random, 95% CI)	0.95 [0.71, 1.27]
2.5 Incidence of T2DM by ethnicity Show forest plot	6	1395	Risk Ratio (M‐H, Random, 95% CI)	0.40 [0.17, 0.95]

2.5.1 Mixed American	3	789	Risk Ratio (M‐H, Random, 95% CI)	0.28 [0.16, 0.48]
2.5.2 Chinese (assumed)	2	199	Risk Ratio (M‐H, Random, 95% CI)	0.33 [0.09, 1.21]
2.5.3 Asian Indian	1	407	Risk Ratio (M‐H, Random, 95% CI)	0.95 [0.71, 1.27]
2.6 Incidence of T2DM by age Show forest plot	6	1395	Risk Ratio (M‐H, Random, 95% CI)	0.40 [0.17, 0.95]

2.6.1 Age ≤ 65 years	5	1276	Risk Ratio (M‐H, Random, 95% CI)	0.39 [0.14, 1.07]
2.6.2 Age > 65 years	1	119	Risk Ratio (M‐H, Random, 95% CI)	0.39 [0.08, 1.95]
2.7 Incidence of T2DM by sex Show forest plot	6	1395	Risk Ratio (M‐H, Random, 95% CI)	0.40 [0.17, 0.95]

2.7.1 Men and women	5	1239	Risk Ratio (M‐H, Random, 95% CI)	0.47 [0.19, 1.15]
2.7.2 Women only	1	156	Risk Ratio (M‐H, Random, 95% CI)	0.13 [0.02, 0.98]
2.8 Incidence of T2DM after intervention end, assessed at any time after intervention end Show forest plot	3	468	Risk Ratio (M‐H, Random, 95% CI)	0.89 [0.49, 1.60]

2.9 Incidence of T2DM after intervention end, assessed at any time after intervention end by 'prediabetes' criteria Show forest plot	3	468	Risk Ratio (M‐H, Random, 95% CI)	0.89 [0.49, 1.60]

2.9.1 IFG and IGT	2	321	Risk Ratio (M‐H, Random, 95% CI)	0.92 [0.51, 1.68]
2.9.2 IFG or IGT	1	147	Risk Ratio (M‐H, Random, 95% CI)	0.30 [0.01, 7.33]
2.10 Serious adverse events Show forest plot	2	187	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.32, 28.22]

2.11 Cardiovascular mortality Show forest plot	4	1156	Risk Ratio (M‐H, Random, 95% CI)	5.14 [0.25, 106.28]

2.12 Non‐fatal myocardial infarction Show forest plot	3	789	Risk Ratio (M‐H, Random, 95% CI)	1.97 [0.18, 21.65]

2.13 Congestive heart failure Show forest plot	3	789	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.93 [0.20, 18.60]

2.14 Congestive heart failure by 'prediabetes' criteria Show forest plot	3	789	Risk Ratio (M‐H, Random, 95% CI)	1.59 [0.20, 12.86]

2.14.1 IFG and IGT	2	633	Risk Ratio (M‐H, Random, 95% CI)	0.99 [0.06, 15.70]
2.14.2 IFG or IGT	1	156	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.12, 72.53]
2.15 Congestive heart failure by sex Show forest plot	3	789	Risk Ratio (M‐H, Random, 95% CI)	1.59 [0.20, 12.86]

2.15.1 Men and women	2	633	Risk Ratio (M‐H, Random, 95% CI)	0.99 [0.06, 15.70]
2.15.2 Women only	1	156	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.12, 72.53]
2.16 Non‐serious adverse events Show forest plot	2	187	Risk Ratio (M‐H, Random, 95% CI)	1.03 [0.81, 1.31]

2.17 Non‐serious adverse events by 'prediabetes' criterion Show forest plot	2	187	Risk Ratio (M‐H, Random, 95% CI)	1.03 [0.81, 1.31]

2.17.1 IFG and IGT	1	31	Risk Ratio (M‐H, Random, 95% CI)	2.13 [0.22, 21.17]
2.17.2 IFG or IGT	1	156	Risk Ratio (M‐H, Random, 95% CI)	1.02 [0.80, 1.30]
2.18 Time to progression of T2DM Show forest plot	2		Hazard Ratio (IV, Random, 95% CI)	Subtotals only

2.19 Time to progression of T2DM by subgroup Show forest plot	2	1009	Hazard Ratio (IV, Random, 95% CI)	0.56 [0.15, 2.11]

2.19.1 Mainly white ethnicity, from USA, commercial funding, IFG and IGT	1	602	Hazard Ratio (IV, Random, 95% CI)	0.28 [0.16, 0.49]
2.19.2 Asian Indian ethnicity, from India, non‐commercial funding, only IGT	1	407	Hazard Ratio (IV, Random, 95% CI)	1.08 [0.75, 1.56]
2.20 Fasting plasma glucose Show forest plot	5		Mean Difference (IV, Random, 95% CI)	Subtotals only

2.21 Fasting plasma glucose by 'prediabetes' criteria Show forest plot	5	1125	Mean Difference (IV, Random, 95% CI)	‐0.52 [‐0.97, ‐0.07]

2.21.1 IFG and IGT	2	560	Mean Difference (IV, Random, 95% CI)	‐0.86 [‐1.82, 0.11]
2.21.2 IFG or IGT	2	198	Mean Difference (IV, Random, 95% CI)	‐0.48 [‐0.71, ‐0.25]
2.21.3 Only IGT	1	367	Mean Difference (IV, Random, 95% CI)	0.05 [‐0.17, 0.27]
2.22 Fasting plasma glucose by ethnicity Show forest plot	5	1125	Mean Difference (IV, Random, 95% CI)	‐0.52 [‐0.97, ‐0.07]

2.22.1 Mainly white	2	559	Mean Difference (IV, Random, 95% CI)	‐0.39 [‐0.48, ‐0.29]
2.22.2 Asian Indian	1	367	Mean Difference (IV, Random, 95% CI)	0.05 [‐0.17, 0.27]
2.22.3 Chinese (assumed)	2	199	Mean Difference (IV, Random, 95% CI)	‐0.92 [‐1.83, ‐0.02]
2.23 Fasting plasma glucose by sex Show forest plot	5	1125	Mean Difference (IV, Random, 95% CI)	‐0.52 [‐0.97, ‐0.07]

2.23.1 Men and women	4	1007	Mean Difference (IV, Random, 95% CI)	‐0.53 [‐1.10, 0.05]
2.23.2 Women only	1	118	Mean Difference (IV, Random, 95% CI)	‐0.49 [‐0.74, ‐0.24]
2.24 2‐hour blood glucose Show forest plot	5	1119	Mean Difference (IV, Random, 95% CI)	‐0.89 [‐1.43, ‐0.34]

2.25 2‐hour blood glucose by 'prediabetes' criteria Show forest plot	5	1119	Mean Difference (IV, Random, 95% CI)	‐0.89 [‐1.43, ‐0.34]

2.25.1 IFG and IGT	2	560	Mean Difference (IV, Random, 95% CI)	‐1.44 [‐1.82, ‐1.05]
2.25.2 IFG or IGT	2	192	Mean Difference (IV, Random, 95% CI)	‐0.65 [‐1.09, ‐0.21]
2.25.3 Only IGT	1	367	Mean Difference (IV, Random, 95% CI)	‐0.09 [‐0.67, 0.49]
2.26 2‐hour blood glucose by ethnicity Show forest plot	5	1119	Mean Difference (IV, Random, 95% CI)	‐0.89 [‐1.43, ‐0.34]

2.26.1 Mainly white	2	553	Mean Difference (IV, Random, 95% CI)	‐0.94 [‐1.65, ‐0.23]
2.26.2 Asian Indian	1	367	Mean Difference (IV, Random, 95% CI)	‐0.09 [‐0.67, 0.49]
2.26.3 Chinese (assumed)	2	199	Mean Difference (IV, Random, 95% CI)	‐1.23 [‐2.11, ‐0.35]
2.27 2‐hour blood glucose by sex Show forest plot	5	1119	Mean Difference (IV, Random, 95% CI)	‐0.89 [‐1.43, ‐0.34]

2.27.1 Men and women	4	1007	Mean Difference (IV, Random, 95% CI)	‐0.96 [‐1.59, ‐0.33]
2.27.2 Women only	1	112	Mean Difference (IV, Random, 95% CI)	‐0.53 [‐1.18, 0.12]
2.28 HbA1C Show forest plot	5	1001	Mean Difference (IV, Random, 95% CI)	‐0.13 [‐0.20, ‐0.07]

2.29 HbA1C by 'prediabetes' criteria Show forest plot	5	1001	Mean Difference (IV, Random, 95% CI)	‐0.13 [‐0.20, ‐0.07]

2.29.1 IFG and IGT	2	435	Mean Difference (IV, Random, 95% CI)	‐0.20 [‐0.28, ‐0.13]
2.29.2 IFG or IGT	2	199	Mean Difference (IV, Random, 95% CI)	‐0.09 [‐0.16, ‐0.01]
2.29.3 Only IGT	1	367	Mean Difference (IV, Random, 95% CI)	‐0.10 [‐0.26, 0.06]
2.30 HbA1C by ethnicity Show forest plot	5	1001	Mean Difference (IV, Random, 95% CI)	‐0.13 [‐0.20, ‐0.07]

2.30.1 Mainly white	3	554	Mean Difference (IV, Random, 95% CI)	‐0.15 [‐0.24, ‐0.05]
2.30.2 Asian Indian	1	367	Mean Difference (IV, Random, 95% CI)	‐0.10 [‐0.26, 0.06]
2.30.3 Chinese (assumed)	1	80	Mean Difference (IV, Random, 95% CI)	‐0.06 [‐0.25, 0.13]
2.31 HbA1c by sex Show forest plot	5	1001	Mean Difference (IV, Random, 95% CI)	‐0.13 [‐0.20, ‐0.07]

2.31.1 Men and women	4	882	Mean Difference (IV, Random, 95% CI)	‐0.16 [‐0.24, ‐0.09]
2.31.2 Women only	1	119	Mean Difference (IV, Random, 95% CI)	‐0.09 [‐0.17, ‐0.01]

Comparison 2. Pioglitazone versus placebo

Ir a la tabla de la revisión

Comparison 3. Pioglitazone versus no intervention

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 All‐cause mortality Show forest plot	3	866	Risk Ratio (M‐H, Random, 95% CI)	0.85 [0.38, 1.91]

3.2 All‐cause mortality by 'prediabetes' criteria Show forest plot	3	866	Risk Ratio (M‐H, Random, 95% CI)	0.85 [0.38, 1.91]

3.2.1 IFG or IGT	2	766	Risk Ratio (M‐H, Random, 95% CI)	0.91 [0.37, 2.27]
3.2.2 Only IGT	1	100	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.12, 3.82]
3.3 All‐cause mortality by comorbidity Show forest plot	3	866	Risk Ratio (M‐H, Random, 95% CI)	0.85 [0.38, 1.91]

3.3.1 No comorbidity in inclusion criteria	2	261	Risk Ratio (M‐H, Random, 95% CI)	0.89 [0.32, 2.49]
3.3.2 Hypertension	1	605	Risk Ratio (M‐H, Random, 95% CI)	0.80 [0.22, 2.94]
3.4 Incidence of T2DM Show forest plot	16	2053	Risk Ratio (M‐H, Random, 95% CI)	0.31 [0.23, 0.40]

3.5 Incidence of T2DM by 'prediabetes' criteria Show forest plot	16	2053	Risk Ratio (M‐H, Random, 95% CI)	0.31 [0.23, 0.40]

3.5.1 IFG and IGT	5	445	Risk Ratio (M‐H, Random, 95% CI)	0.26 [0.13, 0.50]
3.5.2 IFG or IGT	3	812	Risk Ratio (M‐H, Random, 95% CI)	0.34 [0.22, 0.51]
3.5.3 Only IGT	7	676	Risk Ratio (M‐H, Random, 95% CI)	0.27 [0.17, 0.44]
3.5.4 Mixed (subpopulations of IFG, IGT, IFG + IGT)	1	120	Risk Ratio (M‐H, Random, 95% CI)	0.42 [0.16, 1.11]
3.6 Incidence of T2DM by age Show forest plot	16	2053	Risk Ratio (M‐H, Random, 95% CI)	0.31 [0.23, 0.40]

3.6.1 Age ≤ 65	15	1986	Risk Ratio (M‐H, Random, 95% CI)	0.30 [0.23, 0.40]
3.6.2 Age > 65	1	67	Risk Ratio (M‐H, Random, 95% CI)	0.39 [0.08, 1.86]
3.7 Incidence of T2DM by comorbidity Show forest plot	16	2053	Risk Ratio (M‐H, Random, 95% CI)	0.31 [0.23, 0.40]

3.7.1 No comorbidity in inclusion criteria	12	1657	Risk Ratio (M‐H, Random, 95% CI)	0.29 [0.22, 0.40]
3.7.2 Hypertension	4	396	Risk Ratio (M‐H, Random, 95% CI)	0.37 [0.19, 0.72]
3.8 Serious adverse events Show forest plot	7	1211	Risk Ratio (M‐H, Random, 95% CI)	0.71 [0.38, 1.32]

3.9 Serious adverse events by 'prediabetes' criteria Show forest plot	3	861	Risk Ratio (M‐H, Random, 95% CI)	0.71 [0.38, 1.32]

3.9.1 IFG or IGT	2	761	Risk Ratio (M‐H, Random, 95% CI)	0.72 [0.37, 1.40]
3.9.2 Only IGT	1	100	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.12, 3.82]
3.10 Serious adverse events by comorbidity Show forest plot	3	861	Risk Ratio (M‐H, Random, 95% CI)	0.71 [0.38, 1.32]

3.10.1 No comorbidity in inclusion criteria	1	605	Risk Ratio (M‐H, Random, 95% CI)	0.85 [0.29, 2.51]
3.10.2 Hypertension	2	256	Risk Ratio (M‐H, Random, 95% CI)	0.64 [0.30, 1.39]
3.11 Non‐serious adverse events Show forest plot	11	1546	Risk Ratio (M‐H, Random, 95% CI)	1.26 [0.81, 1.97]

3.12 Non‐serious adverse events by 'prediabetes' criteria Show forest plot	11	1546	Risk Ratio (M‐H, Random, 95% CI)	1.26 [0.81, 1.97]

3.12.1 IFG and IGT	2	135	Risk Ratio (M‐H, Random, 95% CI)	3.59 [0.41, 31.63]
3.12.2 IFG or IGT	4	944	Risk Ratio (M‐H, Random, 95% CI)	1.18 [0.83, 1.66]
3.12.3 Only IGT	5	467	Risk Ratio (M‐H, Random, 95% CI)	2.08 [0.50, 8.72]
3.13 Non‐serious adverse events by comorbidity Show forest plot	11	1546	Risk Ratio (M‐H, Random, 95% CI)	1.26 [0.81, 1.97]

3.13.1 No comorbidity in inclusion criteria	8	1235	Risk Ratio (M‐H, Random, 95% CI)	1.27 [1.07, 1.51]
3.13.2 Hypertension	3	311	Risk Ratio (M‐H, Random, 95% CI)	0.42 [0.17, 1.00]
3.14 Participants with events of hypoglycaemia Show forest plot	4	314	Risk Ratio (M‐H, Random, 95% CI)	2.71 [0.76, 9.75]

3.15 Participants with events of hypoglycaemia by 'prediabetes' criteria Show forest plot	2	179	Risk Ratio (M‐H, Random, 95% CI)	2.71 [0.76, 9.75]

3.15.1 Mixed (subpopulations of IFG, IGT, IFG + IGT)	1	120	Risk Ratio (M‐H, Random, 95% CI)	3.00 [0.32, 28.03]
3.15.2 Only IGT	1	59	Risk Ratio (M‐H, Random, 95% CI)	2.59 [0.54, 12.29]
3.16 Fasting plasma glucose Show forest plot	18	1533	Mean Difference (IV, Random, 95% CI)	‐0.70 [‐0.96, ‐0.44]

3.17 Fasting plasma glucose by age Show forest plot	18	1551	Mean Difference (IV, Random, 95% CI)	‐0.70 [‐0.96, ‐0.44]

3.17.1 Age ≤ 65	15	1360	Mean Difference (IV, Random, 95% CI)	‐0.70 [‐0.98, ‐0.42]
3.17.2 Age < 65	3	191	Mean Difference (IV, Random, 95% CI)	‐0.69 [‐1.39, 0.01]
3.18 Fasting plasma glucose by 'prediabetes' criteria Show forest plot	18	1551	Mean Difference (IV, Random, 95% CI)	‐0.70 [‐0.96, ‐0.44]

3.18.1 IFG and IGT	6	560	Mean Difference (IV, Random, 95% CI)	‐0.85 [‐1.39, ‐0.31]
3.18.2 IFG or IGT	2	95	Mean Difference (IV, Random, 95% CI)	‐1.12 [‐1.62, ‐0.62]
3.18.3 Only IGT	8	704	Mean Difference (IV, Random, 95% CI)	‐0.64 [‐1.05, ‐0.23]
3.18.4 Mixed (subpopulations of IFG, IGT, IFG + IGT)	1	120	Mean Difference (IV, Random, 95% CI)	‐0.33 [‐0.51, ‐0.15]
3.18.5 Unspecified	1	72	Mean Difference (IV, Random, 95% CI)	‐0.16 [‐0.41, 0.09]
3.19 Fasting plasma glucose by comorbidity Show forest plot	18	1551	Mean Difference (IV, Random, 95% CI)	‐0.70 [‐0.96, ‐0.44]

3.19.1 No comorbidity in inclusion criteria	15	1368	Mean Difference (IV, Random, 95% CI)	‐0.61 [‐0.89, ‐0.32]
3.19.2 Hypertension	3	183	Mean Difference (IV, Random, 95% CI)	‐1.11 [‐1.26, ‐0.96]
3.20 2‐hour blood glucose Show forest plot	17	1463	Mean Difference (IV, Random, 95% CI)	‐1.58 [‐1.88, ‐1.28]

3.21 2‐hour blood glucose by 'prediabetes' criteria Show forest plot	17	1463	Mean Difference (IV, Random, 95% CI)	‐1.58 [‐1.88, ‐1.28]

3.21.1 IFG and IGT	5	472	Mean Difference (IV, Random, 95% CI)	‐1.76 [‐2.42, ‐1.11]
3.21.2 IFG or IGT	2	95	Mean Difference (IV, Random, 95% CI)	‐2.68 [‐5.10, ‐0.26]
3.21.3 Only IGT	8	704	Mean Difference (IV, Random, 95% CI)	‐1.50 [‐1.69, ‐1.30]
3.21.4 Mixed (subpopulations of IFG, IGT, IFG + IGT)	1	120	Mean Difference (IV, Random, 95% CI)	‐0.84 [‐1.31, ‐0.37]
3.21.5 Unspecified	1	72	Mean Difference (IV, Random, 95% CI)	‐0.05 [‐0.86, 0.76]
3.22 2‐hour blood glucose by 'prediabetes' criteria Show forest plot	15	1271	Mean Difference (IV, Random, 95% CI)	‐1.72 [‐2.01, ‐1.42]

3.22.1 IFG and IGT	5	472	Mean Difference (IV, Random, 95% CI)	‐1.76 [‐2.42, ‐1.11]
3.22.2 IFG or IGT	2	95	Mean Difference (IV, Random, 95% CI)	‐2.68 [‐5.10, ‐0.26]
3.22.3 Only IGT	8	704	Mean Difference (IV, Random, 95% CI)	‐1.50 [‐1.69, ‐1.30]
3.23 2‐hour blood glucose by age Show forest plot	17	1463	Mean Difference (IV, Random, 95% CI)	‐1.58 [‐1.88, ‐1.28]

3.23.1 Age ≤ 65	14	1272	Mean Difference (IV, Random, 95% CI)	‐1.51 [‐1.84, ‐1.18]
3.23.2 Age > 65	3	191	Mean Difference (IV, Random, 95% CI)	‐2.06 [‐2.39, ‐1.72]
3.24 2‐hour blood glucose by comorbidity Show forest plot	17	1463	Mean Difference (IV, Random, 95% CI)	‐1.58 [‐1.88, ‐1.28]

3.24.1 No comorbidity in inclusion criteria	14	1280	Mean Difference (IV, Random, 95% CI)	‐1.38 [‐1.67, ‐1.09]
3.24.2 Hypertension	3	183	Mean Difference (IV, Random, 95% CI)	‐2.52 [‐3.54, ‐1.50]
3.25 HbA1C Show forest plot	9	762	Mean Difference (IV, Random, 95% CI)	‐0.77 [‐1.32, ‐0.23]

3.26 HbA1C by 'prediabetes' criteria Show forest plot	9	762	Mean Difference (IV, Random, 95% CI)	‐0.77 [‐1.32, ‐0.23]

3.26.1 IFG and IGT	2	174	Mean Difference (IV, Random, 95% CI)	‐0.69 [‐1.18, ‐0.20]
3.26.2 Only IGT	6	468	Mean Difference (IV, Random, 95% CI)	‐0.70 [‐1.46, 0.07]
3.26.3 Unspecified	1	120	Mean Difference (IV, Random, 95% CI)	‐1.43 [‐1.98, ‐0.88]
3.27 HbA1C by age Show forest plot	9	762	Mean Difference (IV, Random, 95% CI)	‐0.77 [‐1.32, ‐0.23]

3.27.1 Age ≤ 65	6	571	Mean Difference (IV, Random, 95% CI)	‐0.82 [‐1.55, ‐0.10]
3.27.2 Age < 65	3	191	Mean Difference (IV, Random, 95% CI)	‐0.72 [‐1.07, ‐0.37]
3.28 HbA1c by comorbidity Show forest plot	9	762	Mean Difference (IV, Random, 95% CI)	‐0.77 [‐1.32, ‐0.23]

3.28.1 No comorbidity in inclusion criteria	7	622	Mean Difference (IV, Random, 95% CI)	‐0.80 [‐1.48, ‐0.13]
3.28.2 Hypertension	2	140	Mean Difference (IV, Random, 95% CI)	‐0.63 [‐1.06, ‐0.20]

Comparison 3. Pioglitazone versus no intervention

Ir a la tabla de la revisión

	Idioma de la Revisión Cochrane Escoja su idioma de preferencia para las revisiones Cochrane y otros contenidos. Las secciones sin una traducción aparecerán en inglés.

	Idioma de la web Escoja su idioma de preferencia para la web de la Biblioteca Cochrane.

Idioma de la Revisión Cochrane

Idioma de la web

Resumen

Antecedentes

Objetivos

Métodos de búsqueda

Criterios de selección

Obtención y análisis de los datos

Resultados principales

Conclusiones de los autores

PICO

PICO

Population

Intervention

Comparison

Outcome

Resumen en términos sencillos

¿La pioglitazona previene o retrasa la diabetes tipo 2 y sus complicaciones en personas con riesgo de desarrollar diabetes mellitus de tipo 2?

Resumen visual

Authors' conclusions

Implications for practice

Implications for research

Summary of findings

Background

Description of the condition

Description of the intervention

Adverse effects of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Diagnostic criteria for 'prediabetes'

Types of interventions

Intervention

Comparisons

Minimum duration of intervention

Minimum duration of follow‐up

Summary of specific exclusion criteria

Types of outcome measures

Primary outcomes

Secondary outcomes

Method of outcome measurement

Timing of outcome measurement

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Dealing with duplicate and companion publications

Data from clinical trials registers

Assessment of risk of bias in included studies

Summary assessment of risk of bias

Risk of bias for a study across outcomes

Risk of bias for an outcome within a study and across domains

Measures of treatment effect

Unit of analysis issues

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Summary of findings and assessment of the certainty of the evidence

'Summary of findings' table

Results

Description of studies

Results of the search

Included studies

Source of data

Comparisons

Overview of study populations

Study design

Settings

Participants

Diagnosis

Interventions