Macrolides versus placebo for chronic asthma

Krishna Undela; Lucy Goldsmith; Kayleigh M Kew; Giovanni Ferrara

doi:10.1002/14651858.CD002997.pub5

Macrólidos versus placebo para el asma crónica

Declaraciones de intereses de los autores

Versión publicada: 22 noviembre 2021 Historial de versiones

https://doi.org/10.1002/14651858.CD002997.pub5

Contraer todo Desplegar todo

Resumen

disponible en

Antecedentes

El asma es una enfermedad crónica en la que la inflamación de las vías respiratorias causa sibilancias sintomáticas, tos y dificultad respiratoria. Los macrólidos son antibióticos con actividades antimicrobianas y antiinflamatorias que se han administrado a largo plazo para controlar los síntomas del asma.

Objetivos

Evaluar los efectos de los macrólidos en comparación con placebo para el tratamiento del asma crónica.

Métodos de búsqueda

Se realizaron búsquedas en el registro especializado del Grupo Cochrane de Vías respiratorias (Cochrane Airways Group) hasta marzo de 2021. También se realizaron búsquedas manuales en las bibliografías de revisiones publicadas anteriormente y en resúmenes de congresos y se estableció contacto con los autores de los estudios. Se incluyeron en la búsqueda registros publicados en cualquier idioma.

Criterios de selección

Ensayos clínicos controlados aleatorizados (ECA) que incluyeron a niños y adultos con asma crónica tratados con macrólidos versus placebo durante cuatro o más semanas. Los desenlaces principales fueron las exacerbaciones que requirieron hospitalización, las exacerbaciones graves (exacerbaciones que requirieron visitas al servicio de urgencias [SU] o corticosteroides sistémicos, o ambos), las escalas de síntomas, el cuestionario de control del asma (CCA, con una puntuación de 0, totalmente controlado, a 6, gravemente descontrolado), el cuestionario de calidad de vida del asma (CCVA, con una puntuación de 1 a 7, en el que las puntuaciones más altas indican una mejor calidad de vida), las inhalaciones de medicación de rescate por día, el flujo espiratorio máximo matutino y vespertino (FEM; litros por minuto), el volumen espiratorio forzado en un segundo (VEF ₁; litros), la hiperreactividad bronquial y la dosis de corticoides orales. Los desenlaces secundarios fueron los eventos adversos (incluida la mortalidad), los retiros, los eosinófilos en sangre, los eosinófilos en esputo, la proteína catiónica de eosinófilos (PCE) en suero y la PCE en esputo.

Obtención y análisis de los datos

Dos autores de la revisión, de forma independiente, examinaron todos los registros identificados en las búsquedas y luego revisaron el texto completo de todos los artículos potencialmente pertinentes antes de extraer los datos por duplicado de todos los estudios incluidos. Según el protocolo, se utilizó un modelo de efectos fijos. Se realizó un análisis de sensibilidad para los análisis con alta heterogeneidad (I² mayor del 30%). Para evaluar la certeza del conjunto de evidencia se utilizó el método GRADE.

Resultados principales

Veinticinco estudios cumplieron los criterios de inclusión y asignaron al azar a 1973 participantes a recibir macrólidos o placebo durante al menos cuatro semanas. La mayoría de los estudios incluidos informaron datos de adultos (media de edad de 21 a 61 años) con asma persistente o grave, mientras que cuatro estudios incluyeron niños. Todos los participantes fueron reclutados en ámbitos ambulatorios. Los criterios de inclusión, las intervenciones y los desenlaces fueron muy variables.

La evidencia indica que los macrólidos probablemente proporcionan una reducción de tamaño moderado en las exacerbaciones que requieren hospitalizaciones en comparación con el placebo (odds ratio [OR] 0,47; intervalo de confianza [IC] del 95%: 0,20 a 1,12; estudios = 2, participantes = 529; evidencia de certeza moderada). Los macrólidos probablemente reducen las exacerbaciones que requieren visitas al SU o tratamiento con corticosteroides sistémicos (cociente de tasas [CT] 0,65; IC del 95%: 0,53 a 0,80; estudios = 4, participantes = 640; evidencia de certeza moderada). Los macrólidos podrían reducir los síntomas (medidos en escalas de síntomas) (diferencia de medias estandarizada [DME] ‐0,46; IC del 95%: ‐0,81 a ‐0,11; estudios = 4, participantes = 136; evidencia de certeza muy baja). Los macrólidos podrían dar lugar a una pequeña mejoría en el CCA (DME ‐0,17; IC del 95%: ‐0,31 a ‐0,03; estudios = 5, participantes = 773; evidencia de certeza baja). Los macrólidos podrían tener un efecto escaso o nulo sobre el CCVA (diferencia de medias [DM] 0,24; IC del 95%: 0,12 a 0,35; estudios = 6, participantes = 802; evidencia de certeza muy baja). En el caso del CCA y del CCVA, el efecto indicado de los macrólidos versus placebo no alcanzó una diferencia mínima clínicamente importante (DMCI, 0,5 para el CCA y el CCVA) (CCA: evidencia de certeza baja; CCVA: evidencia de certeza muy baja). Debido a la alta heterogeneidad (I² > 30%), se realizaron análisis de sensibilidad sobre los resultados anteriores, que redujeron el tamaño de los efectos indicados al disminuir la ponderación de los estudios grandes y de calidad alta.

Los macrólidos podrían producir un efecto pequeño en comparación con placebo en la reducción de la necesidad de medicación de rescate (DM ‐0,43 inhalaciones/día; IC del 95%: ‐0,81 a ‐0,04; estudios = 4, participantes = 314; evidencia de certeza baja). Los macrólidos podrían aumentar el VEF₁, pero el efecto está casi con seguridad por debajo de un nivel perceptible para los pacientes (DM 0,04 l; IC del 95%: 0 a 0,08; estudios = 10, participantes = 1046; evidencia de certeza baja). No fue posible agrupar los desenlaces para la hiperreactividad bronquial inespecífica o la dosis más baja tolerada de corticosteroides orales (en personas que requirieron corticosteroides orales al inicio). No hubo evidencia de una diferencia en los eventos adversos graves (incluida la mortalidad), aunque menos de la mitad de los estudios informaron el desenlace (OR 0,80; IC del 95%: 0,49 a 1,31; estudios = 8, participantes = 854; evidencia de certeza baja). El informe de los efectos adversos específicos fue demasiado inconsistente entre los estudios para un análisis significativo.

Conclusiones de los autores

La evidencia existente indica un efecto de los macrólidos en comparación con placebo sobre la tasa de exacerbaciones que requieren hospitalización. Los macrólidos probablemente reducen las exacerbaciones graves (que requieren una visita al SU o tratamiento con corticosteroides sistémicos) y podrían reducir los síntomas. Sin embargo, no es posible descartar la posibilidad de otros efectos beneficiosos o perjudiciales porque la calidad de la evidencia es muy baja debido a la heterogeneidad entre los pacientes y las intervenciones, la imprecisión y los sesgos de notificación. Los resultados se basaron principalmente en un ECA bien diseñado y con un poder estadístico suficiente, e indican que la azitromicina podría reducir la tasa de exacerbación y mejorar las puntuaciones de los síntomas en el asma grave.

La revisión destaca la necesidad de que los investigadores informen sobre los desenlaces con exactitud y de acuerdo con definiciones estándar. Los macrólidos pueden reducir la tasa de exacerbación en personas con asma grave. En los ensayos futuros se podría evaluar si este efecto se mantiene en todos los fenotipos de asma grave, la comparación con los nuevos fármacos biológicos, si los efectos persisten o disminuyen tras el cese del tratamiento y si los efectos se asocian con biomarcadores de infección.

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

¿Se deben administrar macrólidos para el asma crónica?

Punto principal: la evidencia existente indica un efecto beneficioso de los macrólidos en comparación con placebo para reducir las exacerbaciones que requieren hospitalización y las exacerbaciones graves (definidas como exacerbaciones que requieren visita al servicio de urgencias/tratamiento con corticosteroides sistémicos). El efecto de los macrólidos sobre otros desenlaces clínicos relevantes, como las escalas de síntomas y la función pulmonar, aún no está claro.

Antecedentes

El asma es una enfermedad crónica en la cual la inflamación de las vías respiratorias causa tos, sibilancias y problemas respiratorios. Probablemente hay diferentes razones para esta inflamación y motivos por los cuales persiste, y podrían ser necesarios diferentes tratamientos. La infección en los pulmones podría ser una causa, y los macrólidos son un tipo de antibiótico que se podrían utilizar a largo plazo como una forma de mejorar los síntomas en estas personas.

Cómo se respondió la pregunta

Se buscaron estudios sobre adultos o niños con asma a los que se les administró un macrólido o placebo (tratamiento simulado) durante al menos cuatro semanas para determinar si mejoraban los síntomas y disminuía la probabilidad de que tuvieran una crisis de asma, a menudo denominada "exacerbación". La búsqueda más reciente de estudios se realizó en marzo de 2021. Tras encontrar todos los estudios relevantes, se recogió información sobre las crisis de asma que requirieron ingreso hospitalario, las crisis de asma que necesitaron tratamiento con corticosteroides orales, las puntuaciones de los síntomas, el control del asma, la calidad de vida, varias medidas de la función pulmonar, la necesidad de inhaladores de rescate, los efectos secundarios graves y las medidas de la actividad del asma en sangre y esputo (moco).

Datos encontrados

Se encontraron 25 estudios, incluidos dos nuevos que se habían publicado desde la última búsqueda realizada en 2015. En general, poco más de 2000 pacientes recibieron un macrólido o placebo. Hubo muchos problemas en la forma en la que se describieron los estudios y en la forma en la que se informaron los datos, lo que hizo que se considerara que la evidencia en general fue de calidad baja, lo que disminuyó la confianza en la mayoría de los resultados. Los estudios fueron bastante diferentes entre sí, por ejemplo en cuanto a la gravedad del asma de las personas, el tipo de macrólido que se les administró y la duración del periodo de tratamiento.

Esta revisión mostró que los macrólidos fueron mejores que placebo en la reducción de las exacerbaciones y podrían ser beneficiosos en algunas personas para la mejoría de los síntomas del asma, el control del asma, la calidad de vida del asma y algunas medidas de la función pulmonar, pero no se sabe en qué medida ni para quiénes. Según un estudio bien realizado, el macrólido azitromicina podría tener algún beneficio en las personas con asma grave, pero en general los resultados de esta revisión no apoyan el uso de macrólidos para todo el asma de cualquier grado o gravedad. No hubo informes de efectos secundarios graves de los macrólidos, pero 16 estudios no informaron si se produjo alguno.

Authors' conclusions

Implications for practice

Existing evidence suggests an effect of macrolides compared with placebo on the rate of exacerbations requiring hospitalisation. Macrolides probably reduce exacerbations requiring emergency department visit/treatment with systemic steroids and may reduce symptoms. Based on one well‐designed and powered randomised controlled trial, azithromycin may reduce exacerbation rate and improve symptom scales in people with severe asthma but overall we cannot rule out the possibility of other benefits or harms because the evidence is of low certainty due to heterogeneity among participants and interventions, imprecision and reporting biases (Gibson 2017).

Implications for research

The review highlights the need for researchers to report clinically relevant outcomes accurately and completely using guideline definitions of exacerbations, validated symptoms and quality of life scales, as well as international scales for adverse effects. The review and meta‐analysis showed that macrolides can lead to a reduction of exacerbation rate and future trials could evaluate if this effect is sustained across all the severe asthma phenotypes and in comparison with newer biological drugs, whether effects persist or wane after treatment cessation, and are associated with infection biomarkers. Trials with prespecified subgroup analyses by asthma severity or phenotype would usefully contribute to the literature.

Summary of findings

Open in table viewer

Summary of findings 1. Macrolides compared to placebo for chronic asthma

Outcomes^a	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence	Comments
Macrolide versus placebo for chronic asthma
Patient or population: adults and children with chronic asthma Settings: outpatient Intervention: macrolide Drugs used were clarithromycin, azithromycin, roxithromycin and troleandomycin. Macrolide was given once or twice daily in most studies for 4–52 weeks (median 8 weeks). Comparison: placebo Treatment durations were calculated as weighted means of the studies included in each analysis.
	Assumed comparator risk**	Corresponding risk
	Placebo	Macrolide
Exacerbation requiring hospitalisation Weighted mean study duration: 45 weeks	61 per 1000	29 per 1000 (13 to 68)	OR 0.47 (0.20 to 1.12)	529 (2 RCTs)	⊕⊕⊕⊝ Moderate^b	—
Severe exacerbations (requiring ED visits or systemic steroids, or both) Number of people having ≥ 1 exacerbations requiring an ED visit or systemic steroids, or both. Classification varied across studies. Weighted mean study duration: 35 weeks	410 per 1000	311 per 1000 (269 to 357)	Rate ratio 0.65 (0.53 to 0.80)	640 (4 RCTs)	⊕⊕⊕⊝ Moderate^b	—
Asthma symptoms – symptom scales (various scales; lower score = better) Weighted mean study duration: 10 weeks	The mean change in symptom scales in the intervention group compared to the control group was a reduction; 0.46 SD lower (0.81 SD lower to 0.11 SD lower)		—	136 (4 RCTs)	⊕⊝⊝⊝ Very low^e,f,g	The SMD is a Cohen's effect size and can be interpreted as moderate (< 0.4 = small, 0.40–0.70 = moderate, > 0.70 = large).
Asthma control (Asthma Control Questionnaire) Scored 0–6 (lower scores indicate improvement in asthma control) Weighted mean study duration: 34 weeks	The mean change in asthma control in the intervention group compared to the mean change in control group was a greater reduction;0.17 SD lower (0.31 SD lower to 0.03 SD lower)		—	773 (5 RCTs)	⊕⊕⊝⊝ Low^d,h	The SMD is a Cohen's effect size and can be interpreted as small.
Asthma quality of life (AQLQ) Scored 1–7 (higher scores indicate improvement in quality of life) Weighted mean study duration: 35 weeks	The mean change on the AQLQ scale in the control group was an increase from baseline; 0.31 points higher**	The mean change in AQLQ in the intervention group compared to the mean change in the control group was an increase; 0.24 points higher (0.12 higher to 0.35 higher)	—	802 (6 RCTs)	⊕⊝⊝⊝ Very low^c,d,h	MCID: 0.5 points
Rescue medication (puffs/day) Weighted mean study duration: 17 weeks	The mean rescue medication use in the control group was 1.08 puffs/day**	The mean change in rescue medication puffs/day in the intervention group compared to the mean change in the control group was reduction; 0.43 fewer puffs (0.81 fewer to 0.04 fewer)	—	314 (4 RCTs)	⊕⊕⊝⊝ Low^d,h	—
FEV₁ (L) Weighted mean study duration: 28 weeks	The mean FEV₁ in the control group was 2.49 L**	The mean change in FEV₁ (L) the intervention group compared to the control group was an increase; 0.04 L higher (0 L to 0.08 L higher)	—	1046 (10 RCTs)	⊕⊕⊝⊝ Low^h,i,j	Although there is no universally accepted MCID for FEV₁ in asthma, variability within a single testing session can be up to 0.12 L (data from a mixed pool of respiratory patients; Enright 2004).
Serious adverse events (including mortality) Weighted mean study duration: 16 weeks	93 per 1000	76 per 1000 (50 to 118)	OR 0.80 (0.49 to 1.31)	854 (8 RCTs)	⊕⊕⊝⊝ Low^i,k	—
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). Assumed risk for continuous outcomes were calculated as weighted means of the scores in the control group. Note: Sutherland 2010 could not be included in the rescue medication or quality of life calculations because they reported only mean difference between groups; Brusselle 2013 was not included in the FEV₁ calculation because it was the only change score; Cameron 2013 was not included in the asthma control or quality of life calculations because it was the only study reporting absolute endpoint scores rather than change from baseline. AQLQ: Asthma Quality of Life Questionnaire; CI: confidence interval; ED: emergency department; FEV₁: forced expiratory volume in one second; MCID: minimal clinically important difference; OR: odds ratio; RCT: randomised controlled trial; SD:** standard deviation.
GRADE domains: study limitations, consistency of effect, imprecision, indirectness and publication bias. GRADE Working Group grades of evidence (Schünemann 2021). High certainty: further research is very unlikely to change our confidence in the estimate of effect. Moderate certainty: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low certainty: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low certainty: we are very uncertain about the estimate.
^aTwo primary outcomes (morning and evening peak expiratory flow) are not presented. Bronchial hyperresponsiveness could not be pooled in a meta‐analysis and is described narratively in the review. Studies did not report lowest tolerated oral corticosteroid dose well and there were no data to analyse. ^bDowngraded one level for indirectness due to differences in the recruited populations and in the criteria used to define 'severe exacerbations'. ^cDowngraded one level for imprecision as among the included studies, one reported an important benefit of macrolide and another possible benefit of placebo. ^dDowngraded one level due to uncertainties with randomisation procedures and high risk of attrition bias in some studies included in the analysis. ^eDowngraded one level for inconsistency due to high heterogeneity (I² = 70%) mainly due to one cross‐over study. ^fDowngraded one level for indirectness. Symptom scales were often invalidated and highly variable across studies, and we chose not to pool most in a meta‐analysis using standard mean differences, as this would lead to a result that would have been much more difficult to interpret. ^gDowngraded one level for imprecision due to small number of participants in the analysis; it was difficult to judge precision due to different scales. ^hDowngraded one level for indirectness as studies in the analysis recruited different populations with regard to severity of asthma, and one study only recruited smokers. ⁱDowngraded one level for risk of bias as four studies that we were unable to properly assess for risk of bias were included in this analysis, and we were uncertain of how and when the measurement was taken in some cases. ^jNo change in grade: there was uncertainty in several domains across studies, but the two studies carrying most of the weight were well conducted. ^kDowngraded one level for imprecision; eight studies reported the outcomes, but only four studies observed events, leading to very wide confidence intervals that included important benefit and harm of macrolide treatment.

Background

Description of the condition

Asthma is an inflammatory disease of the airways characterised by chronic inflammation, bronchial hyperresponsiveness and paroxysmal attacks of wheezing. It affects people of every age, but frequently the disease occurs in childhood, especially in those who are atopic. It is estimated that asthma may affect between 1% and 18% of the general population, and it represents a significant cause of morbidity and costs for healthcare systems. Furthermore, the control of the disease is difficult to achieve in people with severe asthma, and even in people with milder asthma it may be hampered by poor adherence to treatments and lack of access to healthcare (GINA 2021).

Different phenotypes of the disease are recognised and under investigation. Current guidelines recommend tailoring asthma treatment according to a stepwise approach, considering severity of symptoms and response to treatment (GINA 2021).

Recently, the role of short‐acting bronchodilators in intermittent asthma was revised, and a combination of inhaled corticosteroids (ICS) and formoterol as needed is now recommended in people with mild asthma. Persistent asthma is treated with regular ICS, longer‐acting bronchodilators, or both (GINA 2021). More recent therapies include anti‐leukotrienes in mild‐to‐moderate asthma, humanised antibodies targeting immunoglobulin E (omalizumab), interleukin (IL)‐5 (mepolizumab, reslizumab, benralizumab), and IL‐4/‐13 (dupilumab), which are currently only recommended in severe asthma with markers of a phenotype likely to respond (e.g. raised blood eosinophils for the anti‐IL‐5 agents) (GINA 2021; Olin 2014).

Description of the intervention

Macrolides are a class of antibiotics that are widely used in the treatment of various infectious diseases, including respiratory tract infections (Alvarez‐Elcoro 1999). The first studies on macrolides in people with asthma suggested a steroid‐sparing effect (Nelson 1993), while other reports have demonstrated an anti‐inflammatory effect of this class of antibiotics, whereby macrolides seem to decrease bronchial hyperresponsiveness associated with eosinophilic inflammation (Amayasu 2000). A host‐directed anti‐inflammatory effects was also postulated (Spagnolo 2013). Macrolides are effective in the long‐term treatment of cystic fibrosis, diffuse panbronchiolitis and chronic obstructive pulmonary disease (COPD), and they are not associated with an increased risk of adverse events (Cai 2011; Spagnolo 2013).

However, a potential drawback of longer‐term antibiotic use for asthma is the development of bacterial resistance by strains that normally colonise the airways. Macrolide use in healthy volunteers led to pharyngeal carriage of macrolide‐resistant streptococci (Malhotra‐Kumar 2007), which is of particular concern for the wider community. Similar concerns were raised by studies in COPD (Brill 2015). Furthermore, the potential for arrhythmias due to QTc prolongation, potential ototoxicity and hepatotoxicity from macrolide long‐term use was highlighted in a British Thoracic Society document (BTS 2020).

How the intervention might work

Macrolides have anti‐inflammatory and antimicrobial properties that may improve asthma symptoms in two ways: by reducing airways inflammation directly and by controlling intracellular infection, which may trigger and maintain inflammation (Black 1997; Black 2000; Kawasaki 1998). Their anti‐inflammatory potential has been linked to their action on pro‐inflammatory cytokines and chemokines causing inflammation, which was highlighted by the results of the previous versions of this systematic review (Richeldi 2002; Richeldi 2005; Kew 2015). In vivo and in vitro studies of human and animal models have demonstrated that macrolides suppress the production of cytokines such as ILs and inhibit neutrophil adhesion to epithelial cells, the respiratory burst of neutrophils and the secretion of mucous from human airways (Adachi 1996; Hinks 2021; Konno 1994; Koyama 1998).

Older macrolides such as troleandomycin were investigated for a steroid‐sparing effect, related to reduced hepatic glucocorticoid metabolism (Nelson 1993).

The potential benefit of their antimicrobial action for people with asthma was suggested after observational studies identified intracellular bacterial infection (i.e. Chlamydophila pneumoniae or Mycoplasma pneumoniae) as a possible trigger of bronchial inflammation (Kraft 1998). Gencay 2001 subsequently demonstrated that people with asthma had a higher frequency of C pneumoniae antibodies than matched controls. Longitudinal studies showed no clear effect of infection with C pneumoniae on the incidence of asthma, but people who had an infection and developed asthma showed a faster decline in lung function (Pasternack 2005). Furthermore, in children with asthma, M pneumoniae detection in respiratory samples was associated with poorer asthma control (Wood 2013). Studies in animal models seem to point to an important role of the infection with C pneumoniae in the early phases of life in the pathogenesis of severe asthma (Essilfie 2015; Hansbro 2014).

Why it is important to do this review

Macrolides represent a relatively inexpensive intervention that may improve control of inflammation and clinical outcomes in people with chronic asthma.

Objectives

To assess the effects of macrolides compared with placebo for managing chronic asthma.

Methods

Criteria for considering studies for this review

Types of studies

We included parallel and cross‐over randomised controlled trials (RCTs).

Types of participants

Children and adults with chronic asthma.

Types of interventions

Macrolides, administered for four or more weeks versus placebo. We pooled data from studies comparing different macrolide therapies.

Types of outcome measures

Primary outcomes

Exacerbations requiring hospitalisation.
Severe exacerbations (defined as requiring an emergency department (ED) visits or short‐course of systemic steroids, or both).
Asthma symptoms, control and quality of life scores.
Asthma medication requirements (need for rescue medications).
Lung function, including morning and evening peak expiratory flow (PEF) and forced expiratory volume in one second (FEV₁).
Non‐specific bronchial hyperresponsiveness (to histamine or methacholine).
Lowest tolerated oral corticosteroid dose (in people requiring oral corticosteroids at baseline).

Secondary outcomes

Number and type of serious adverse events (including mortality).
Number of study withdrawals.
Eosinophil count in peripheral blood samples, sputum samples or both.
Eosinophilic cationic protein (ECP) measurements in serum and sputum.

Search methods for identification of studies

Electronic searches

Search methods used in the previous version of this review are detailed in Appendix 1. The previously published version included searches up to April 2015. The search period for this update was April 2015 to 31 March 2021.

We searched the Cochrane Airways Trials Register on 31 March 2021. At that time the Register contained studies identified from:

monthly searches of the Cochrane Central Register of Controlled Trials (CENTRAL), through the Cochrane Register of Studies Online (crso.cochrane.org), all years to Issue 3, 2021;
weekly searches of MEDLINE (OvidSP), 1946 to 26 March 2021;
weekly searches of Embase (OvidSP), 1974 to week 12 2021;
monthly searches of PsycINFO (OvidSP), 1967 to March week 4 2021;
monthly searches of CINAHL (EBSCO) (Cumulative Index to Nursing and Allied Health Literature), 1937 to 15 March 2021;
monthly searches of AMED (EBSCO) (Allied and Complementary Medicine), all years to 11 March 2021;
handsearches of the proceedings of major respiratory conferences.

Studies contained in the Trials Register are identified through search strategies based on the scope of Cochrane Airways. Details of these strategies, as well as a list of handsearched conference proceedings are in Appendix 2. See Appendix 3 for search terms used to identify studies for this review. We did not restrict our search by language or type of publication.

We also searched the following trials registries on 31 March 2021:

US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (www.clinicaltrials.gov);
World Health Organization International Clinical Trials Registry Platform (apps.who.int/trialsearch).

Searching other resources

We surveyed review articles and bibliographies identified from the primary papers for additional references and RCTs.

Data collection and analysis

Selection of studies

Two review authors (KU and GF) independently screened the abstracts of articles identified using the search strategy above, retrieving the full text for articles that appeared to fulfil the inclusion criteria. Two review authors (KU and GF) independently reviewed and categorised each article identified as included or excluded. When there was disagreement or doubt, a third review author (KK) assessed the article and helped to reach a consensus. We presented a PRISMA diagram to illustrate the flow of studies through the selection process (Moher 2009).

Data extraction and management

We used a data collection form to collect study characteristics and outcome data. We piloted the form on at least one study in the review. Two review authors (KU and GF) extracted the following study characteristics from included studies, when available.

Methods: study design, total duration of study, details of any run‐in period, number of study centres and location, study setting, withdrawals and date of study.
Participants: number, mean age, age range, gender, severity of condition, diagnostic criteria, baseline lung function, smoking history, inclusion criteria and exclusion criteria.
Interventions: intervention, comparison, concomitant medications and excluded medications.
Outcomes: primary and secondary outcomes specified and collected, and time points reported.
Notes: funding for trial, and notable conflicts of interest of trial authors.

Two review authors (KU and GF) independently extracted outcome data from included studies. We noted in the Characteristics of included studies table if outcome data were not reported in a usable way. We resolved disagreements by involving a third review author (KK). Two review authors (KK and KU) transferred data into the Review Manager 5 (Review Manager 2014). Three review authors (KU, LG and GF) double‐checked that data were entered correctly by comparing the data presented in the systematic review with the study reports.

Assessment of risk of bias in included studies

Two review authors (KK and GF) independently assessed risk of bias for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We resolved disagreements by discussion and by involving another review author (KU). We assessed risk of bias according to the following domains.

Random sequence generation.
Allocation concealment.
Blinding of participants and personnel.
Blinding of outcome assessment.
Incomplete outcome data.
Selective outcome reporting.
Other bias.

Each potential source of bias was graded as high, low or unclear and justified with a quote from the study report in the risk of bias table. We summarised the risk of bias judgements across different studies for each of the domains listed. We considered blinding separately for different key outcomes where necessary (e.g. for unblinded outcome assessment, risk of bias for all‐cause mortality may be very different from for a participant‐reported pain scale). Where information on risk of bias related to unpublished data or correspondence with a trial list, we noted this in the risk of bias table.

For treatment effects, we considered the risk of bias for the studies that contributed to that outcome.

Measures of treatment effect

We analysed dichotomous data as odds ratios (OR), incidence rate data as rate ratio (RaR) and continuous data as mean difference (MD; where studies used the same scales) or standardised mean difference (SMD; where studies used different scales). We entered data presented as a scale with a consistent direction of effect. We narratively described skewed data reported as medians and interquartile ranges. We analysed data from cross‐over trials using generic inverse variance (GIV). We pooled results from cross‐over trials and parallel trials. Where studies presented raw data and adjusted analyses (e.g. accounting for baseline differences), we used the adjusted analyses.

We undertook meta‐analyses only where meaningful, that is, if the treatments, participants and the underlying clinical question were similar enough for pooling to make sense and made decisions about this by consensus among review authors.

Where a trial reported multiple arms, we included only the relevant arms but reported all arms in the Characteristics of included studies table. If two comparisons (e.g. drug A versus placebo and drug B versus placebo) were combined in the same meta‐analysis, we halved the control group to avoid double‐counting.

If change from baseline and endpoint scores were available for continuous data, we used change from baseline unless most studies reported endpoint scores. If a study reported outcomes at multiple time points, we used the end‐of‐study treatment measurement.

When both an analysis using only participants who completed the trial and an analysis that imputed data for participants who were randomised but did not provide endpoint data (e.g. last observation carried forward) were available, we used the latter.

Unit of analysis issues

We combined events data using ORs or RaRs (number of participants or number of events) according to which measure would allow us to include the most studies. We used ORs for exacerbations requiring hospitalisation, serious adverse events (including mortality) (Peto OR) and withdrawal. We used RaRs for severe exacerbations (defined as exacerbations requiring ED visits or systemic steroids, or both). For continuous data in cross‐over trials, we entered data using GIV from suitable adjusted analyses to account for the trial's design.

Dealing with missing data

We assessed the potential for bias in each trial as a result of participants dropping out of the intervention prematurely. Where this was thought to introduce serious bias, we removed the studies in a sensitivity analysis.

Assessment of heterogeneity

We used the I² statistic to measure heterogeneity among the trials in each analysis. If we identified high heterogeneity (e.g. I² greater than 30%), we reported it and performed a sensitivity analysis with a random‐effects model.

Assessment of reporting biases

We were unable to pool more than 10 trials for any of the primary outcomes, so were unable to examine a funnel plot to explore possible small‐study and publication biases.

Data synthesis

We used a fixed‐effect model for all analyses, as we expected limited variation in effects due to differences in study populations and methods.

Subgroup analysis and investigation of heterogeneity

We planned subgroup analyses based on serological response or positivity to polymerase chain reaction (PCR) for C pneumoniae.

Sensitivity analysis

We performed a sensitivity analysis with a random‐effects model in case of high heterogeneity (I² greater than 30%).

Summary of findings and assessment of the certainty of the evidence

The summary of findings table included the following outcomes: number of exacerbations requiring hospitalisation; severe exacerbations (requiring ED visits or short‐course systemic steroids, or both); asthma symptoms (including symptom scores, asthma control and Asthma Quality of Life Questionnaire (AQLQ)); asthma medication requirements (as reliever); lung function (including morning and evening PEF and FEV₁); non‐specific bronchial hyperresponsiveness; serious adverse events; withdrawal; blood and sputum eosinophils; and ECP in serum and sputum.

We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the certainty of the body of evidence as it related to the studies that contributed data to the meta‐analyses for the prespecified outcomes. Except for serious adverse events, we did not perform GRADE ratings on the secondary outcomes. We used methods and recommendations described in Section 8.5 and Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) using GRADEpro software (GRADEpro GDT). We justified all decisions to downgrade or upgrade the certainty of the evidence using footnotes and made comments to aid reader's understanding of the review where necessary.

Results

Description of studies

Results of the search

The literature search of previous versions of this review up to April 2015 identified 137 citations. Out of the 38 references identified in the search between 2007 and 2015, duplicate sifting of the titles and abstracts alone identified 33 potentially eligible studies for inclusion in the systematic review. Among them, review authors were concordant in identifying 16 RCTs to be included in the previous update of the systematic review (Kew 2015). Six of the 16 RCTs were identified in a previous meta‐analysis (Tong 2015); these Chinese trials were only listed in China's biomedical databases. The authors of this review were able to confirm key study characteristics in order to include them, although risk of bias could not be properly assessed.

For this updated review, the search extended to March 2021. We identified 55 new references. Seven titles referring to abstracts presented at congresses were duplicates of other studies already included in the review, while one title (Gibson 2019) was a subgroup analysis of another RCT (Gibson 2017) included in this update, leaving 20 references for screening. We excluded four based on the abstract alone and 13 after screening the full‐text of the original manuscripts, leaving two new studies eligible for inclusion in the systematic review and meta‐analyses (Wan 2016; Gibson 2017). One study is ongoing, no results available. See Characteristics of included studies, Characteristics of excluded studies, and Characteristics of ongoing studies tables. The study flow of the new included studies is presented in Figure 1.

Figure 1

Study flow diagram.

Included studies

We included 25 RCTs (Amayasu 2000; Belotserkovskaya 2007; Black 2001; Brusselle 2013; Cameron 2013; Gibson 2017; Hahn 2006; Hahn 2012; He 2009; Kamada 1993; Kapoor 2010; Kostadima 2004; Kraft 2002; Nelson 1993; Piacentini 2007; Shoji 1999; Simpson 2008; Strunk 2008; Sutherland 2010; Wan 2016; Wang 2012; Wang 2014; Xiao 2013; Yan 2008; Zhang 2013). For brief descriptions of the included studies, refer to the Characteristics of included studies table. For a summary of study characteristics and a narrative on the main results of each study, see Table 1 and Appendix 4.

Open in table viewer

Table 1. Summary characteristics of included studies at baseline

Study ID	Country	Number of participants	Design	Duration (weeks)	Macrolide dose and schedule	Mean age (years)	% Male	% on ICS	% Predicted FEV₁
Amayasu 2000	Japan and USA	17	C, R, DB, PC	8	Clarithromycin 200 mg twice daily	38.5	52.9	0.0	76.2
Belotserkovskaya 2007	Russia	51	P, R	8	Azithromycin (unknown dose)	NR	NR	NR	NR
Black 2001	Multinational	219	P, R, DB, PC	6	Roxithromycin 150 mg twice daily	41.0	47.5	80.8	77.1
Brusselle 2013	Belgium	109	P, R, DB, PC	26	Azithromycin 250 mg once daily for 5 days then 3 times per week	53.0 (median)	38.5	100^a	82.5
Cameron 2013	UK	77	P, R, DB, PC	12	Azithromycin 250 mg once daily	44.6	48.1	85.7	78.3
Gibson 2017	Australia	420	P, R, DB, PC	48	Azithromycin 500 mg 3 times per week	60.5	39.3	99.8	72.9
Hahn 2006	USA, Canada	45	P, R, DB, PC	6	Azithromycin 600 mg once daily for 3 days then weekly	47.7	48.9	80.0	NR
Hahn 2012	USA	75	P, R, DB, PC	12	Azithromycin 600 mg once daily for 3 days then weekly	46.6	32.0	72.0	NR
He 2009	China	40	P, R, PC	12	Azithromycin 250 mg twice weekly	34.5	NR	NR	NR
Kamada 1993	USA	19	P, R, DB, PC	12	Troleandomycin 250 µg once daily + OCS	12.5	63.2	100	NR
Kapoor 2010	India	40	C, R, DB, PC	6	Roxithromycin 150 mg once daily	NR	NR	NR	NR
Kostadima 2004	Greece	75	P, R, DB, PC	8	Clarithromycin 250 mg twice daily or 3 times daily	43.7	47.1	100	85.3
Kraft 2002	USA	55	P, R, DB, PC	6	Clarithromycin 500 mg twice daily	33.4	49.1	32.7	69.3
Nelson 1993	USA	75	P, R, DB, PC	52	Troleandomycin 250 mg once daily + OCS	NR	33.3	0	NR
Piacentini 2007	Italy	16	P, R, DB, PC	8	Azithromycin 10 mg/kg once daily 3 days per week	13.4	75	100	78.9
Shoji 1999	Japan	14	C, R, DB, PC	8	Roxithromycin 150 mg twice daily	39.6	42.9	0.0	75.0
Simpson 2008	Australia	45	P, R, DB, PC	8	Clarithromycin 500 mg twice daily	57.6	48.9	NR^b	70.7
Strunk 2008	USA	55^c	P, R, DB, PC	30	Azithromycin 250 mg or 500 mg once daily	11.2	58.2	100^a	101.9
Sutherland 2010	USA	92	P, R, DB, PC	16	Clarithromycin 500 mg twice daily	39.4	43.5	NR	76.0
Wan 2016	Taiwan	58	P, R, PC	4	Clarithromycin 5 mg/kg daily	10.1	60.3	100	79.7
Wang 2012	China	45	P, R, PC	8	Clarithromycin 500 mg twice daily	NR	NR	NR	NR
Wang 2014	China	58	P, R, PC	52	Azithromycin 250 mg twice weekly	29.0	NR	NR	NR
Xiao 2013	China	210	P, R, PC	12	Roxithromycin 150 mg twice daily	34.1	NR	NR	NR
Yan 2008	China	40	P, R, PC	4	Roxithromycin 150 mg twice daily	38.5	NR	NR	NR
Zhang 2013	China	60	P, R, PC	9	Azithromycin 100 mg once daily	NR	NR	NR	NR

C: cross‐over; DB: double‐blind; FEV₁ : forced expiratory volume in one second; ICS: inhaled corticosteroid; NR: not reported; OCS: oral corticosteroids; P: parallel; PC: placebo‐controlled; R: randomised.

^aAll participants were taking long‐acting beta₂‐agonist + ICS combination.
^b82.2% were taking long‐acting beta₂‐agonist + ICS combination.
^c19 participants were not included in the review because they were randomised to a third group who received montelukast.

The 25 included studies reported a great variability in type of participants (ranging from intermittent aspirin‐induced asthma to severe asthma), interventions (different type of macrolides, administration scheme and doses in most of the studies) and outcomes recorded.

Design

All studies were RCTs using placebo controls, and most were described as double‐blind. There were 22 parallel group studies and three cross‐over studies (Amayasu 2000; Kapoor 2010; Shoji 1999). Median study duration was 15.2 weeks (range four to 52 weeks). Two studies were reported in the form of abstracts from congresses, with a very limited amount of data available (Belotserkovskaya 2007; Kapoor 2010); the results of one new included study could not be used due to very poor reporting (Wan 2016), and only basic information was available from six Chinese studies (He 2009; Wang 2012; Wang 2014; Xiao 2013; Yan 2008; Zhang 2013) included in the previous version of this review (Kew 2015).

Participants

The studies included 1991 participants of whom 1973 were relevant to this review (Strunk 2008 included a third group of 18 people receiving a treatment that was outside the protocol of this review). All participants had an asthma diagnosis, which was generally established according to the guidelines in use at the time of the studies (ATS 1987; GINA 1995; GINA 2002; GINA 2007; GINA 2010; GINA 2014); these are similar to the current international guidelines for what concerns main diagnosis and grading of severity of asthma (GINA 2021).

Four studies assessed the effects of macrolide treatment in children with asthma (Kamada 1993; Piacentini 2007; Strunk 2008; Wan 2016), and the rest recruited adults with asthma (aged over 18 years). The studies varied according to GINA 2014/GINA 2020/GINA 2021 criteria for asthma severity, and often there was very little information about baseline severity.

Most studies included participants with persistent mild‐to‐severe asthma, while one included participants with mild asthma (step 1 according to GINA 2021) (Amayasu 2000), and one used people with aspirin‐induced intermittent asthma (Shoji 1999).

Five studies investigated the role of macrolides in people with evidence of C pneumoniae or M pneumoniae infection, based on serological (Black 2001; Hahn 2006; Wan 2016) or molecular (Kraft 2002; Sutherland 2010) methods. The remaining studies did not investigate the presence of these co‐infections or of other concomitant co‐infections, although we could not confirm this in the non‐English language papers. Cameron 2013 investigated the effect of macrolides in adult smokers with persistent asthma, while Brusselle 2013, Simpson 2008, and Gibson 2017 considered the effect of macrolides in people with eosinophilic and non‐eosinophilic asthma.

Interventions

Five studies compared roxithromycin with placebo (Black 2001; Kapoor 2010; Shoji 1999; Xiao 2013; Yan 2008); seven studies compared clarithromycin with placebo (Amayasu 2000; Kostadima 2004; Kraft 2002; Simpson 2008; Sutherland 2010; Wang 2014; Wan 2016); 11 studies investigated the effect of azithromycin (Belotserkovskaya 2007; Brusselle 2013; Cameron 2013; Gibson 2017; Hahn 2006; Hahn 2012; He 2009; Piacentini 2007; Strunk 2008; Wang 2012; Zhang 2013), and two studies assessed the effects of troleandomycin in addition to oral steroid therapy as part of a steroid‐tapering protocol (Kamada 1993; Nelson 1993).

Outcomes

Six studies did not appear in any of the quantitative syntheses (Belotserkovskaya 2007; Black 2001; Kapoor 2010; Wan 2016; Wang 2012; Zhang 2013), and two more only contributed to the bronchial hyperresponsiveness summary of results and withdrawal (Piacentini 2007; Simpson 2008).

Two studies reported data on exacerbations requiring hospitalisation (Brusselle 2013; Gibson 2017), and four on severe exacerbations (defined as exacerbations requiring ED visits or systemic steroids, or both) as an outcome, but the definition of 'severe exacerbation' used in the different studies was variable and sometimes unclear (Brusselle 2013; Gibson 2017; Kostadima 2004; Strunk 2008). The meta‐analysis of Tong 2015 did not include exacerbations as an outcome but explicitly confirmed that the Chinese studies did not report this outcome either. Data from these studies only contributed to one meta‐analysis (FEV₁). We narratively summarised data that could not be meta‐analysed for the relevant outcomes.

Most studies reported measures of symptoms, asthma control or quality of life, but the analyses were limited by the way data were reported and by the use of different scales. We did not consider a meta‐analysis of all these measures to be valid or the subsequent results to be interpretable in any meaningful way, so we chose only to meta‐analyse those that we knew would be similar. We used SMDs for the 'symptom scale' meta‐analysis, which still made the effect and its precision difficult to interpret.

Four studies reported data about change in rescue medication as puffs per day in a way that could be included in meta‐analysis (Brusselle 2013; Cameron 2013; Hahn 2006; Sutherland 2010).

Most of the studies reported measures of lung function such as FEV₁ or PEF, but only 10 reported data for FEV₁ (Amayasu 2000; Cameron 2013; Gibson 2017; He 2009; Kraft 2002; Shoji 1999; Sutherland 2010; Wang 2014; Xiao 2013; Yan 2008), four reported morning PEF (Brusselle 2013; Cameron 2013; Kamada 1993; Sutherland 2010), and three reported for evening PEF (Brusselle 2013; Kamada 1993; Sutherland 2010) that could be pooled. There were some issues with selective reporting that prevented studies from being included in the analyses, such as data only being presented graphically or without a measure of variance (e.g. Black 2001; Wan 2016). It was often unclear when the measures were taken (i.e. pre‐ or post‐bronchodilator), but when the information was available, we recorded it in the analysis footnotes. Brusselle 2013 reported percentage FEV₁, but their data could not be combined with the other studies, which reported the outcome in litres. We combined the data made available to us from Tong 2015 for He 2009, Wang 2014, Xiao 2013, and Yan 2008. We were also provided with data for peak flow for Wang 2014, Xiao 2013, and Yan 2008, but the data were a different order of magnitude to the other studies, and it did not make sense to pool them.

Nine studies considered bronchial hyperresponsiveness, but there was variation in the measures used and the way the data were reported, which meant it was not possible to meta‐analyse the data (Amayasu 2000; Cameron 2013; Kamada 1993; Kostadima 2004; Nelson 1993; Piacentini 2007; Shoji 1999; Simpson 2008; Sutherland 2010).

Most studies considered adverse events, but only eight explicitly reported serious adverse events (Amayasu 2000; Brusselle 2013; Cameron 2013; Gibson 2017; Hahn 2006; Hahn 2012; Kamada 1993; Sutherland 2010). While it was not ideal to include the dichotomous cross‐over data without adjusting them to account for matched pairs, there were no events in Amayasu 2000, so it did not contribute to the pooled effect.

Ten studies reported study withdrawal (Brusselle 2013; Gibson 2017; Hahn 2006; Hahn 2012; Kamada 1993; Kostadima 2004; Nelson 1993; Simpson 2008; Strunk 2008; Sutherland 2010).

Eight studies reported the effect of macrolides on markers of inflammation related to asthma activity, but they used different measures, which could not be pooled in one analysis (Amayasu 2000; Cameron 2013; Kraft 2002; Nelson 1993; Piacentini 2007; Shoji 1999; Simpson 2008; Yan 2008). There were also some issues with data accuracy or incomplete reporting that reduced our confidence in the reliability of the data. The separate analyses include very small participant numbers, mostly from the two cross‐over studies.

Two studies considered the steroid‐sparing effect of macrolides (Kamada 1993; Nelson 1993).

Excluded studies

We excluded 14 studies from the review after viewing the full papers. Reasons for exclusion are reported in the Characteristics of excluded studies table and Figure 1.

Previous versions of this systematic review excluded 17 studies after reading the full papers (Kew 2015; Richeldi 2005), so a total of 31 studies are excluded.

Studies awaiting classification

There are no studies awaiting classification.

Ongoing studies

We found one ongoing study (NCT02517099).

Risk of bias in included studies

There was considerable uncertainty relating to study methodology due to insufficient reporting in the published reports. This was particularly true for the selection bias domains, but also for blinding of outcome assessment and attrition bias. We had concerns about incomplete and selective reporting of the results for most of the studies and generally considered there to be a high risk for bias because only a few studies reported data in a way that could be pooled in meta‐analysis. Summaries of the risk of bias judgements for each study are presented in Figure 2 and Figure 3.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Figure 3

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Pharmaceutical industries financed at least five included studies (Amayasu 2000; Cameron 2013; Hahn 2006; Hahn 2012; Kamada 1993); which could raise the risk of publication bias; this could not be ascertained in the non‐English language papers. The authors of Tong 2015 provided us with information about study quality for the studies that were not available in English.

Allocation

We deemed 13 studies at low risk of bias for random sequence generation, including seven of the English language studies (Brusselle 2013; Gibson 2017; Hahn 2006; Hahn 2012; Piacentini 2007; Simpson 2008; Wan 2016) and six in Tong 2015 (He 2009; Wang 2012; Wang 2014; Xiao 2013; Yan 2008; Zhang 2013). For the rest, the methods for random sequence generation was unclear.

Only five studies were at low risk for adequate allocation concealment (Brusselle 2013; Gibson 2017; Hahn 2006; Hahn 2012; Wan 2016); the studies from the Tong 2015 review were not assessed for this criterion because it is not considered in the Jadad 1996 system, so we had to rate those studies as unclear, and the 14 other studies did not adequately describe the methods used.

Blinding

Most studies described as double‐blind and placebo‐controlled contained adequate descriptions of the blinding of participants and personnel, but methods were unclear in nine studies (He 2009; Wan 2016; Wang 2012; Wang 2014; Xiao 2013; Yan 2008; Zhang 2013, including two in an abstract form: Belotserkovskaya 2007; Kapoor 2010). Blinding of outcome assessment was adequate in seven studies (Cameron 2013; Gibson 2017; Hahn 2006; Hahn 2012; Kraft 2002; Piacentini 2007; Sutherland 2010).

The same study that we rated high for performance bias also carried a high risk for detection bias (Strunk 2008).

Incomplete outcome data

Six studies had a high risk of attrition bias (Gibson 2017; Hahn 2006; Hahn 2012; Kostadima 2004; Nelson 1993; Sutherland 2010), five carried a low risk (Black 2001; Brusselle 2013; Kraft 2002; Simpson 2008; Strunk 2008), and the risk was unclear for the other 14 studies.

Selective reporting

We considered three studies at low risk of bias (Brusselle 2013; Hahn 2006; Hahn 2012). We judged eight studies at high risk of selective reporting (Belotserkovskaya 2007; Black 2001; Cameron 2013; Kamada 1993; Kapoor 2010; Simpson 2008; Strunk 2008; Sutherland 2010). This was mostly due to insufficient reporting of numerical data, which meant they could not be pooled in meta‐analysis. We rated 14 studies as unclear for this domain, including the six non‐English language studies, which we could not assess fully.

Overall, it is likely that reporting biases had a significant effect on the completeness of the meta‐analyses in this systematic review.

Other potential sources of bias

We judged Kamada 1993 to be at high risk of bias because the report showed significant baseline imbalances between groups, and this may have been an issue in some of the other trials that included very small number of participants. We judged eight studies to be at unclear risk of bias (Hahn 2012; He 2009; Kapoor 2010; Kostadima 2004; Wang 2012; Wang 2014; Xiao 2013; Yan 2008; Zhang 2013).

Effects of interventions

See: Summary of findings 1 Macrolides compared to placebo for chronic asthma

We present the data with the order of outcomes listed in the methods. Evidence certainty was varied. We downgraded most for indirectness due to differences in the study populations and the way outcomes were defined, and for risk of bias due to uncertainty of randomisation procedures and high risk of attrition bias. Appendix 4 presents a narrative on each study, except for the six that we could not assess fully (He 2009; Wang 2012; Wang 2014; Xiao 2013; Yan 2008; Zhang 2013).