Métodos hormonales e intrauterinos para la anticoncepción de mujeres de hasta 25 años de edad
Resumen
Antecedentes
Las mujeres de entre 15 y 24 años de edad tienen altas tasas de embarazos no deseados; más de la mitad de las mujeres de este grupo de edad desean evitar el embarazo. Sin embargo, las mujeres menores de 25 años tienen tasas típicas de fracaso anticonceptivo más altas en los primeros 12 meses de uso que las mujeres de más edad. Las tasas altas de interrupción también pueden ser un problema en esta población. La preocupación acerca de que las adolescentes y las mujeres jóvenes no encontrarán anticonceptivos hormonales o intrauterinos aceptables o efectivos podría disuadir a los profesionales sanitarios de recomendar estos métodos anticonceptivos.
Objetivos
Comparar las tasas de fracaso anticonceptivo (embarazo) y examinar las tasas de continuación de la anticoncepción hormonal e intrauterina entre las mujeres jóvenes de 25 años o menos.
Métodos de búsqueda
Se realizaron búsquedas hasta el 4 de agosto de 2015 de ensayos controlados aleatorizados (ECA) que compararan métodos anticonceptivos hormonales o intrauterinos en mujeres de 25 años o menos. Las bases de datos informatizadas incluyeron el Registro Cochrane de Ensayos Controlados (Cochrane Register of Controlled Trials, CENTRAL), MEDLINE, EMBASE, POPLINE, CINAHL y LILACS. También se realizaron búsquedas de ensayos en curso a través de ClinicalTrials.gov y de la Plataforma Internacional de Registros de Ensayos Clínicos (International Clinical Trials Registry Platform, ICTRP).
Criterios de selección
Se consideraron los ECA en cualquier idioma que informaran las tasas de fracaso anticonceptivo de los métodos anticonceptivos hormonales o intrauterinos, comparados con otro método anticonceptivo, para mujeres de 25 años de edad o menos. El otro método anticonceptivo podría haber sido otro anticonceptivo intrauterino, otro anticonceptivo hormonal o una dosis diferente del mismo método, o un anticonceptivo no hormonal. La duración del tratamiento debía haber sido de, al menos, tres meses. Los ensayos elegibles debían incluir la medida de resultado primario de la tasa de fracaso anticonceptivo (embarazo). La medida de resultado secundaria fue la tasa de continuación de los anticonceptivos.
Obtención y análisis de los datos
Un autor realizó la extracción de datos primarios e introdujo la información en Review Manager. Otro revisor realizó la extracción de los datos de forma independiente y verificó la introducción inicial. Para los resultados dicotómicos, se calculó el odds ratio (OR) de Mantel‐Haenszel con un intervalo de confianza (IC) del 95%. Debido a las intervenciones y medidas de resultado dispares, no se realizó un metanálisis.
Resultados principales
Cinco ensayos cumplieron los criterios de inclusión. Los estudios incluyeron un total de 1 503 mujeres, con una media de 301 participantes. Los ensayos compararon los siguientes anticonceptivos: anticonceptivo oral combinado (AOC) versus parche anticonceptivo transdérmico, anillo anticonceptivo vaginal o sistema intrauterino de levonorgestrel 20 µg/día (SIU‐LNG 20); SIU‐LNG 12 µg/día (SIU‐LNG 12) versus SIU‐LNG16 µg/día (SIU‐LNG 16); y SIU‐LNG 20 versus el dispositivo intrauterino (DIU) de cobre T380A. En los ensayos que compararon dos tipos diferentes de métodos, los brazos del estudio no difirieron significativamente en cuanto a la eficacia anticonceptiva o la continuación. Los tamaños de la muestra fueron pequeños para dos de esos estudios. El único resultado significativo fue que un grupo de AOC tuvo una mayor proporción de mujeres que descontinuaron por "otras razones personales" en comparación con el grupo asignado al SIU‐LNG 20 (OR 0,27; IC del 95%: 0,09 a 0,85), que puede tener poca relevancia clínica. El ensayo que comparó el SIU‐LNG 12 versus el SIU‐LNG 16 mostró una eficacia similar durante uno y tres años. En tres ensayos que examinaron diferentes SIU‐LNG, la continuación fue al menos del 75% de los seis a 36 meses.
Conclusiones de los autores
La calidad general de la evidencia se consideró moderada a baja. Las limitaciones se debieron al diseño del ensayo o al informe limitado. Las diferentes dosis en el SIU‐LNG no parecieron influir en la eficacia durante tres años. En otro estudio, la continuación del SIU‐LNG pareció ser al menos tan alta como la del AOC. La evidencia actual es insuficiente para comparar las tasas de eficacia y de continuación de los métodos anticonceptivos hormonales e intrauterinos en mujeres de 25 años o menos.
PICO
Resumen en términos sencillos
Métodos hormonales e intrauterinos para la anticoncepción de mujeres de hasta 25 años de edad
Antecedentes
Las mujeres de 25 años o menos tienen más probabilidades de dejar de usar métodos anticonceptivos que las mujeres mayores de 25 años. Tienen también más probabilidades de quedar embarazadas mientras utilizan métodos anticonceptivos que las mujeres mayores. No se conoce qué métodos anticonceptivos tienen las tasas de embarazo más bajas y el uso continuado más alto en mujeres jóvenes.
Características de los estudios
Se buscaron ensayos aleatorizados de métodos anticonceptivos hasta agosto de 2015. Los ensayos aleatorizados son estudios clínicos en los que las personas son colocadas aleatoriamente en uno de dos o más grupos de tratamiento. Las mujeres en estos estudios tenían 25 años o menos. Los métodos anticonceptivos podían ser hormonales o un dispositivo no hormonal colocado en el útero. Los métodos hormonales incluían píldoras, anillos vaginales o implantes. Los métodos que se colocan en el útero incluyen el dispositivo intrauterino (DIU) sin hormonas y el sistema intrauterino que tiene la hormona levonorgestrel (SIU‐LNG). Los DIU y el SIU‐LNG a veces se denominan anticonceptivos intrauterinos (AIU).
Resultados clave y calidad de la evidencia
Se encontraron cinco ensayos que reclutaron a un total de 1 503 mujeres. Algunos estudios analizaron diferentes tipos de AIU, mientras que otros compararon píldoras versus un anillo vaginal, un parche cutáneo o un AIU. Ningún estudio mostró diferencias importantes entre los grupos durante el embarazo o el uso continuado. Algunos de los ensayos fueron demasiado pequeños para encontrar una diferencia. Las mujeres siguieron usando un AIU por lo menos tanto tiempo como las píldoras en un estudio. Los AIU pueden ser útiles para las mujeres de este grupo de edad. Los estudios de diferentes métodos anticonceptivos con más mujeres ayudarían a determinar qué métodos funcionan mejor para las mujeres jóvenes. En general, la calidad de los resultados fue de moderada a baja.
Authors' conclusions
Summary of findings
| Contraceptive failure rate (pregnancy) | ||||
| Patient or population: women aged ≤ 25 years with desire for contraception Settings: clinic or community Intervention: hormonal or intrauterine method Comparison: other hormonal or intrauterine method | ||||
| Outcomes | Relative effect | Participants | Quality of the evidence | Interventions |
| Pregnancy (12 months) | Not estimable (none) | 193 | ⊕⊝⊝⊝ | LNG‐IUS 20 vs. COC (EE 30 µg + desogestrel 150 µg) |
| Estimated cumulative failure rate (3 years) | Reported Kaplan‐Meier: 0.010 (0.004 to 0.027) vs. 0.005 (0.001 to 0.019) | 1130 | ⊕⊕⊝⊝ | LNG‐IUS 12 vs. LNG‐IUS 16 |
| Pregnancy (6 months) | OR 1.00 (0.05 to 18.57) | 20 | ⊕⊕⊕⊝ | COC (EE 30 µg + norgestimate 250 µg) vs. transdermal contraceptive patch |
| GRADE Working Group grades of evidence | ||||
| CI: confidence interval; LNG‐IUS 20: levonorgestrel‐releasing intrauterine system (20 µg/day); COC: combined oral contraceptive; EE = ethinyl estradiol; OR: odds ratio | ||||
| Continuation or discontinuation of contraceptive method | ||||
| Patient or population: women aged ≤ 25 years with desire for contraception Settings: clinic or community Intervention: hormonal or intrauterine method Comparison: other hormonal or intrauterine method | ||||
| Outcomes | Relative effect | Participants | Quality of the evidence | Interventions |
| Continuation (12 months) | OR 1.48 (0.76 to 2.89) | 193 | ⊕⊝⊝⊝ | LNG‐IUS 20 vs. COC (EE 30 µg + desogestrel 150 µg) |
| Continuation (1 month; 6 months) | OR 2.44 (0.19 to 31.53); | 23 | ⊕⊕⊕⊝ moderate | Cu T380A IUD vs. LNG‐IUS 20 |
| Continuation (cycles 1 to 3; cycles 4 to 6) | OR 1.09 (0.49 to 2.42); | 130 | ⊕⊕⊕⊝ | COC (EE 35 µg + norgestimate 250 µg) vs. vaginal contraceptive ring |
| Continuation (6 months) | OR 0.38 (0.05 to 2.77) | 20 | ⊕⊕⊕⊝ | COC (EE 30 µg + norgestimate 250 µg) vs. transdermal contraceptive patch |
| Discontinuation: other personal reasons (12 months) | OR 0.27 (0.09 to 0.85) | 193 | ⊕⊝⊝⊝ | LNG‐IUS 20 vs. COC (EE 30 µg + desogestrel 150 µg) |
| Discontinuation: pain, hormonal, bleeding, spotting, other medical, or planning pregnancy (12 months) | No significant difference | 193 | ⊕⊝⊝⊝ | LNG‐IUS 20 vs. COC (EE 30 µg + desogestrel 150 µg) |
| GRADE Working Group grades of evidence | ||||
| CI: confidence interval; OR: Odds ratio; LNG‐IUS: levonorgestrel‐releasing intrauterine system; COC: combined oral contraceptive; EE = ethinyl estradiol | ||||
Background
Description of the condition
Women between the ages of 15 and 24 years have a high demand for contraception, and Millennium Development Goal 5 includes increased access to contraception among young women (United Nations 2013). In sub‐Saharan Africa, south central Asia, and Southeast Asia, 68% of women between the ages of 15 and 19 years, and 51% of women between 20 and 24 years old want to avoid pregnancy (Darroch 2011). However, in low‐income countries, approximately 33 million women aged 15 to 24 years old have an unmet need for contraception (MacQuarrie 2014). Unmet need in these age groups is not limited to low‐income countries. In the United States (US), over half of the pregnancies among women under age 25 years are unintended (Finer 2014). In the US, 30% of women 15 to 19 years old use contraception, and the percentage increases as women age and use female sterilization as a permanent method of contraception (Jones 2012).
Women under the age of 25 account for a large proportion of maternal mortality worldwide. Maternal sepsis and hemorrhage are major causes of death among women under 25 years of age in low‐income countries (Patton 2009). Over eight million unsafe abortions, which are the third major cause of maternal mortality, occurred in women aged 15 to 24 years in 2008 (Shah 2012). Third‐quarters of these unsafe abortions, as well as other causes of maternal mortality, could be prevented by improved access to modern contraception (WHO 2014).
The age range for the transition from adolescence to young adulthood is poorly defined in the scientific literature and among policymakers. Healthcare policy organizations have defined the lower age cutoff for adolescence as 10 to 15 years of age and the upper limit as 18 to 24 years of age (DHHS 2008; Gavin 2009; UNPF 2010; WHO 2011). Given the inconsistency of definitions for adolescence, this review will focus on women aged 25 years and younger.
Description of the intervention
Among reversible forms of contraception, the contraceptive implant and intrauterine contraception have the lowest contraceptive failure rates and the highest continuation rates in the general population. These forms of contraception are also likely to have the highest continuation rates and lowest pregnancy rates for women aged 25 years and younger. First‐year pregnancy rates for typical use range from 9% for oral contraceptives, the patch, and the ring to less than 1% with the contraceptive implant and intrauterine contraception (Trussell 2011). Many factors may contribute to the gap between perfect use and typical use, including missed doses and discontinuation due to inconvenience, forgetfulness, or side effects.
Among women using modern contraceptives, those under the age of 25 years have higher typical contraceptive failure rates within the first 12 months compared with women aged 25 years and older. For example, the first‐year contraceptive failure rates are 13% for women younger than 20 years, 14% for women aged 20 to 24 years and 8% for women aged 30 years and older. These findings were particularly pronounced for women who were using condoms or oral contraceptives (Kost 2008). In addition, high discontinuation rates may be a problem in this population. Among married women aged 15 to 24 years with an unmet need for contraception, only 8% cited lack of access or high cost as the reason for not using a method, but 15% to 19% reported concern about health or side effects (Guttmacher 2010). Data from 2006 to 2008 indicated that typical discontinuation rates at 12 months were 30% for oral contraceptives, 43% for the contraceptive injection, and 50% for the contraceptive patch (Mosher 2010). A 12‐month cohort study of 1387 US women aged 15 to 24 years showed that continuation rates were lower for the contraceptive patch (10.9 per 100 person‐years) and injection (12.1 per 100 person‐years) when compared with the contraceptive ring (29.4 per 100 person‐years) and pills (32.7 per 100 person‐years). Discontinuation was independently associated with the method initiated and younger age (Raine 2011). In one prospective study that emphasized counseling, free devices, and immediate access, age less than 20 years was associated with higher discontinuation of contraception over 24 months compared with that among older women. However, young women were less likely to discontinue long‐acting reversible methods of contraception (i.e. intrauterine contraception and implants) compared with other types of contraception (O'Neil‐Callahan 2013).
Why it is important to do this review
The comparative efficacy and acceptability of contraceptive methods for adolescents and young women are important public health issues. Concern that adolescents and young women will not find hormonal or intrauterine contraceptives acceptable or effective might deter healthcare providers from recommending them. When we conducted the initial review, a systematic review had not been done with randomized controlled trials (RCTs) on hormonal and intrauterine contraception in this population. The findings of this update could help clinicians and policymakers improve their recommendations for contraceptive counseling and use, and decrease the high unintended pregnancy rate in this age group.
Objectives
To compare the contraceptive failure (pregnancy) rates and to examine the continuation rates for hormonal and intrauterine contraception among women aged 25 years and younger.
Methods
Criteria for considering studies for this review
Types of studies
We considered RCTs that reported the contraceptive failure rates for hormonal and intrauterine contraceptives when compared with another contraceptive method. The other contraceptive could be another intrauterine contraceptive, another hormonal contraceptive or different dose of the same method, or a non‐hormonal contraceptive.
Types of participants
We only included studies that specified that their primary study population was adolescents or young women and in which data were presented for women aged 25 years or younger. We excluded studies that focused on special populations of young women such as women with diabetes, polycystic ovarian syndrome, or anorexia.
Types of interventions
Any hormonal or intrauterine contraceptive was examined, such as the combined oral contraceptive (COC), progestin‐only pill, transdermal patch, vaginal ring, contraceptive injection (both the combined injectable and the progestin‐only injectable), contraceptive implant, copper intrauterine device (IUD), or levonorgestrel intrauterine system (LNG‐IUS). The comparison group could be another intrauterine contraceptive, another hormonal contraceptive or different dose of the same method, or a non‐hormonal contraceptive. Treatment duration must have been at least three months. The contraceptive method must have been used primarily for pregnancy prevention in all participants.
Types of outcome measures
Primary outcomes
Contraceptive failure rate (pregnancy)
Secondary outcomes
Contraceptive continuation rate
Search methods for identification of studies
Electronic searches
We searched Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, CINAHL, POPLINE, and LILACS until 4 August 2015. We also searched for ongoing trials via ClinicalTrials.gov and the search portal of the International Clinical Trials Registry Platform (ICTRP). Appendix 1 shows the search strategies. Appendix 2 has the strategies for the initial review.
Searching other resources
We examined the reference lists of relevant articles and contacted investigators in the field to seek additional published trials or unpublished trials.
Data collection and analysis
Selection of studies
We assessed for inclusion all titles and abstracts identified during the literature search, with no language limitations. Two authors independently reviewed the search results and identified reports for inclusion or exclusion.
Data extraction and management
Two authors conducted the data extraction. One author entered the data into Review Manager 5 (RevMan 2014), and a second author checked accuracy (Contributions of authors). These data included the study characteristics, risk of bias, and outcome data. We resolved any discrepancies by discussion.
Assessment of risk of bias in included studies
We examined the RCTs for methodological quality according to recommended principles in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). The methodology considered included randomization sequence generation, allocation concealment, blinding, loss to follow‐up, and early discontinuation. Adequate methods for allocation concealment included a centralized system and the use of sequentially numbered, opaque, sealed envelopes (Schulz 2002). We presented limitations in design in the Risk of bias in included studies and considered these when interpreting the results.
Measures of treatment effect
For dichotomous outcomes, we computed the Mantel‐Haenszel odds ratio (OR) with 95% confidence interval (CI). For the contraceptive failure rate, we compared the proportions of women who became pregnant in each study arm at the reported time intervals. For the continuation rate, we compared the proportions of women who were using the hormonal or intrauterine contraceptive at the reported time intervals. We also compared the proportion of women who discontinued by specific reason (if available). When data were not available for analysis, we presented the results as reported by the investigators. This includes the Pearl Index with 95% CI. We could not perform life‐table analysis.
Dealing with missing data
We contacted four investigators to obtain supplemental data (Suhonen 2004; Stuart 2005; Stewart 2007; Kaunitz 2013). We obtained additional information for three trials (Stuart 2005; Stewart 2007; Kaunitz 2013).
Assessment of heterogeneity
Study populations, designs, and interventions were heterogeneous. We did not pool data from studies that used different contraceptive methods or had different lengths of follow‐up.
Data synthesis
We applied principles from GRADE to assess the quality of evidence and address confidence in the effect estimates (Balshem 2011; Higgins 2011). Our assessment of the body of evidence was based on the quality of evidence from the studies. When a meta‐analysis is not viable because of varied interventions or outcome measures, a formal GRADE assessment is not feasible, that is, with an evidence profile and 'Summary of findings' table (Guyatt 2011). However, we did provide 'Summary of findings' tables.
To assess the quality of evidence from each trial, we used a rating system. We considered RCTs to be high quality then downgraded for the following: each risk of bias criteria judged as 'high'; risk of bias 'unclear' for two or more criteria due to missing or incomplete information.
Results
Description of studies
Results of the search
For the initial review in 2012, two authors independently reviewed 786 abstracts. We considered 15 studies for inclusion. Four trials met our inclusion criteria. The included trials enrolled a total of 468 women, and the sample sizes ranged from 20 to 200 women. Trial locations included two European countries (Sweden and Finland) and the US. All trials were published within the previous 10 years and included contraceptive methods currently available.
For the 2015 update, the database searches produced 342 citations. After we removed 73 duplicates electronically or by hand, we had 269 unduplicated references. We considered 17 studies for inclusion. One trial met our eligibility criteria and 16 were excluded.
Searches of clinical trials databases yielded 17 unduplicated listings. We listed one as excluded; the others did not appear to be eligible for this review. In most of those trials with young women, the interventions were educational rather than clinical.
Included studies
Five trials met our eligibility criteria (see Characteristics of included studies). The studies included a total of 1503 women, ranging from 20 to 1130; the mean was 301. Two studies compared different methods of intrauterine contraception.
-
Godfrey 2010 was a multicenter pilot study of 23 women, aged 14 to 18 years, assigned to either the copper T380A IUD or the levonorgestrel intrauterine system with 20 µg/day initial release (LNG‐IUS 20). The study was completed in urban family planning and family medicine clinics in Chicago, Illinois and enrolled mostly African‐American high‐school students. Eleven of the 23 women had previously given birth, although none were recently postpartum. The study compared the rates of pregnancy, expulsion, continuation, infection, side effects, bleeding, and satisfaction for the two methods at six months.
-
Kaunitz 2013 is a subanalysis from a multicenter Phase III study of 2884 women, aged 18 to 35 years, which had a full report (Nelson 2013). Kaunitz 2013 was an abstract from a conference poster. Women were assigned to one of two LNG‐IUS: 12 µg/day initial release (LNG‐IUS 12) or 16 µg/day initial release (LNG‐IUS 16). Most women were parous. Of the 2884 original participants, 1130 were included in a sub‐analysis of women aged 18 to 25 years. The study was completed in 138 centers in 11 countries. The Phase III study compared the rates of pregnancy at one and three years, bleeding over three years, ectopic pregnancies, and bone mineral density. The primary report described side effects and continuation over three years (Nelson 2013). This abstract provides rates of pregnancy and ectopic pregnancy over one and three years according to the women's ages. The abstract also describes discontinuation because of side effects for both groups combined. The LNG‐IUS 12 is approved by the US Food and Drug Administration (FDA) for use over three years. The investigators provided additional data on the number of women aged 18 to 25 years who were randomized to each arm.
Three trials compared a COC with another hormonal method: vaginal ring, transdermal patch, or LNG‐IUS 20.
-
Suhonen 2004 was a multicenter trial of 200 women, aged 18 to 25 years. Women were assigned to either COC (ethinyl estradiol 30 µg plus desogestrel 150 µg) or the LNG‐IUS 20. The study was completed in urban family planning clinics located in the cities of Linköping, Sweden and Helsinki, Finland. All women except one were white, and all women were nulliparous. The primary outcome was the continuation rate, but the study also assessed pregnancy, infection, patterns of bleeding, and acceptability. In the LNG‐IUS group, evaluation by both the women and the doctors of the insertion as well as perforation and expulsion rates were recorded and analyzed. Three‐month packages of the COCs were given at each three‐month visit.
-
Stuart 2005 was a pilot study of 20 women, aged 15 to 19 years. Stuart 2005 was an abstract from a conference poster; no full report was published. An author of this review (JT) was involved in the study and provided data for the outcome analyses. Women were assigned to either COC (ethinyl estradiol 30 µg plus norgestimate 250 µg) or the transdermal contraceptive patch (daily dose of ethinyl estradiol 20 µg plus norelgestromin 150 µg). The study was completed in urban family planning clinics for low‐income people in Dallas, Texas. Seventy‐five per cent of women were Latina, and 20% were African‐American. Parity of the women was not reported, although less than half had previously had an abortion. The primary outcome was pregnancy rate, but the study also assessed continuation, compliance, and satisfaction for the methods at three and six months. A three‐month supply of the randomized contraceptive method was provided at the enrollment visit and after three cycles. Women were allowed to switch methods at the three‐month visit and were considered to have discontinued their assigned method if they switched.
-
Stewart 2007 was a cross‐over pilot study of 130 women, aged 15 to 21 years. The women were assigned to either COC (ethinyl estradiol 35 µg plus norgestimate 250 µg) or the vaginal contraceptive ring (daily dose of ethinyl estradiol 15 µg and etonogestrel 120 µg) for three cycles, followed by three cycles of the alternate method. The study was completed in an urban family planning clinic for low‐income young people in San Francisco, California. The majority of women were either African‐American or Latina and had not completed high school. Only 6% had previously given birth. The primary outcome was acceptability of the two methods, but also reported were rates of compliance, side effects, continuation, and pregnancy for each three‐cycle period. A co‐investigator shared the database with a review author (JT) for use in the initial review.
Excluded studies
We excluded 28 studies. The reasons included contraceptive method not used primarily for pregnancy prevention in all women, different contraceptive methods not compared, participants not limited to women aged 25 years or younger, no hormonal or intrauterine method used for comparison, or continuation measured for less than three months. In addition, the full text for one study indicated it was not an RCT and several studies included special populations of women. See Characteristics of excluded studies for more details.
Risk of bias in included studies
Allocation
Two studies had adequate information on the randomization sequence generation and allocation concealment. Stuart 2005 had blocks of four and six. Godfrey 2010 had a computer‐generated random number sequence with block sizes of six. Both studies used sequentially numbered, sealed, opaque envelopes to conceal allocation.
Three studies had adequate information on the sequence generation but no mention of allocation concealment. Suhonen 2004 randomized into two equal‐sized groups in blocks of eight prior to study initiation and carried out separately in both centers. Stewart 2007 used a random number generator in blocks (size unspecified). In the main study from which Kaunitz 2013 was derived, the randomization sequence was developed using cards in ascending order and lists were prepared for each study site (Nelson 2013).
Blinding
In Godfrey 2010 and Kaunitz 2013, the women were blinded to which intrauterine method was inserted until the study ended (Godfrey 2010) or three years (Kaunitz 2013). None of the studies used blinding of the research staff or investigators since the different contraceptive methods either required different routes for use (oral, transdermal, vaginal, intrauterine) or required intrauterine insertion by a research staff member or investigator (copper T380A IUD versus the LNG‐IUS 20 and different reservoir sizes for the LNG‐IUS 12 versus LNG‐IUS 16).
Incomplete outcome data
Losses greater than 20% threaten trial validity (Strauss 2005). In three studies, the information on loss to follow‐up was incomplete (Suhonen 2004; Stewart 2007; Kaunitz 2013).
-
Suhonen 2004 had overall discontinuation greater than 20% (LNG‐IUS 20%; COC 27%). Loss to follow‐up was not specified by method. The investigators excluded from the outcome analyses seven randomized women who never began the study intervention. They excluded five women randomized to the LNG‐IUS 20 for the following reasons: uterus too small at insertion, chlamydia infection detected at randomization, genital infection, and lost to follow‐up. Two women randomized to the COC were excluded because the participant moved abroad or never started the study medication. However, the investigators did include two women in the COC group who had a positive chlamydia test but entered the study after receiving treatment. Suhonen 2004 listed loss to follow‐up as one of the 'personal reasons' for discontinuation of the assigned method. COC users had a higher proportion of women reporting 'personal reasons' for discontinuation (P = 0.02). They may have had pregnancies not identified in the study.
-
Stewart 2007 included all randomized women in the analysis for continuation at three months. The discontinuation rate at three months was high (25% for the COC, 24% for the vaginal contraceptive ring). Only 51% of women completed the six months of the study. The reasons for discontinuation, including loss to follow‐up, were not available by assigned method. An attrition analysis showed no difference between women lost to follow‐up and women who completed the six‐month study visit for age, race or ethnicity, previous pregnancy, prior use of hormones, condoms use at last intercourse, or age of sexual debut.
-
Kaunitz 2013 included all randomized women in the analysis for pregnancy and continuation rates. However, the report did not specify whether discontinued women requested device removal or were lost to follow‐up.
Two trials had low losses to follow‐up. Godfrey 2010 included all randomized women in the pregnancy and continuation rates. The six‐month loss to follow‐up was 9% for the copper T380A IUD and 18% for the LNG‐IUS 20. Stuart 2005 included all randomized women in the pregnancy and continuation rates. The six‐month loss to follow‐up was 10% for the COC and 20% for the transdermal contraceptive patch.
Selective reporting
Four trials did not have complete reporting for all outcomes. Two studies did not report how pregnancy was assessed at the study visits (Suhonen 2004; Godfrey 2010). Kaunitz 2013 used pregnancy tests at screening, baseline, the three‐year study visit, and when clinically indicated. Three studies did not include information on how continuation was determined (Suhonen 2004; Stuart 2005; Stewart 2007). Stewart 2007 did not include reasons for discontinuation and loss to follow‐up by assigned method.
Effects of interventions
See: Summary of findings for the main comparison Contraceptive failure (pregnancy); Summary of findings 2 Continuation of contraceptive method
Godfrey 2010 compared the copper T380A IUD with the LNG‐IUS 20. Only one pregnancy occurred among 23 women. This small trial was not informative regarding contraceptive efficacy. The one pregnancy was diagnosed 37 days after copper T380A IUD removal for prolonged bleeding. One‐month continuation rates were 92% (11 of 12) for the LNG‐IUS 20 and 82% (9 of 11) for the copper T380A IUD (OR 2.44, 95% CI 0.19 to 31.53; Analysis 1.1). Six‐month continuation rates were 75% (9 of 12) for the LNG‐IUS 20 and 45% (5 of 11) for the copper T380A IUD (OR 3.60, 95% CI 0.62 to 21.03; Analysis 1.2). Bleeding problems were mentioned as the reason for discontinuation by one woman in each group. In the copper T380A IUD group, three additional women reported reasons for discontinuation: one reported excessive cramping and two had expulsion.
Kaunitz 2013 compared LNG‐IUS with two different dosages: 12 µg/day (LNG‐IUS 12) versus 16 µg/day (LNG‐IUS 16). Unadjusted Pearl Indices were similar for the two interventions: 0.22 (95% CI 0.01 to 1.22) for LNG‐IUS 12 and 0.21 (95% CI 0.01 to 1.18) for LNG‐IUS 16 at one year (Analysis 5.1). At three years, the unadjusted Pearl Indices were 0.36 (0.10 to 0.92) for the LNG‐IUS 12 and 0.17 (0.02 to 0.60) for the LNG‐IUS 16 (Analysis 5.2). Of four ectopic pregnancies, two occurred with each devices over three years: 0.17 per 100 women‐years for the LNG‐IUS 12; 0.18 per 100 women‐years for the LNG‐IUS 16. The risk of expulsion was 4.78% overall, and two cases (0.2%) of pelvic inflammatory disease were reported overall. Twenty‐two percent of women discontinued an IUS due to adverse events; however, overall continuation was not stated.
The cross‐over trial of Stewart 2007 compared the COC with the vaginal contraceptive ring. Overall, 13 pregnancies occurred: four were associated with an interval of vaginal contraceptive ring use and nine related to an interval of COC use. Because the report did not specify whether the pregnancies occurred with cycles one to three or with cycles four to six, we could not calculate contraceptive efficacy rates. However, continuation rates did not differ significantly between the intervention groups for the first three months (OR 1.09, 95% CI 0.49 to 2.42) or the second three months (OR 0.61, 95% CI 0.26 to 1.43) (Analysis 2.1). For the COC, 50 of 67 (75%) women completed three cycles compared with 48 of 63 (76%) women for the vaginal contraceptive ring. Reasons for study discontinuation included pregnancy (n = 9), side effects (n = 1), desire to switch contraceptive method (n = 1), lost to follow‐up (n = 4), medical or personal reasons (n = 5), and other or unknown (n = 12). A co‐investigator provided additional data for this review.
Stuart 2005 compared the COC with the transdermal contraceptive patch. The trial was underpowered to show important differences in pregnancy or continuation rates. One of 10 women in each intervention group became pregnant during the study (OR 1.00, 95% CI 0.05 to 18.57; Analysis 3.1). Eight of 10 women continued the COC at six months, whereas six women continued the transdermal contraceptive patch at six months (OR 0.38, 95% CI 0.05, 2.77; Analysis 3.2). Reasons for discontinuation for the COC group included pregnancy (n = 1) and loss to follow‐up (n = 1), whereas reasons for discontinuation for the transdermal contraceptive patch group included loss to follow‐up (n = 2) and concern about the patch falling off (n = 2).
Finally, Suhonen 2004 compared the LNG‐IUS 20 versus the COC. This trial also showed no important differences in pregnancy rates or continuation rates. No pregnancies occurred in either intervention group over 12 months (Analysis 4.1). Twelve‐month continuation rates were 80% (75 of 94) in the LNG‐IUS group and 73% (72 of 99) in the COC group (OR 1.48, 95% CI 0.76 to 2.89; Analysis 4.2). Women in the LNG‐IUS 20 group were more likely than women in the COC group to discontinue their method because of pain (OR 14.62, 95% CI 0.81 to 263.16) and less likely to discontinue because of personal reasons (OR 0.27, 95% CI 0.09 to 0.85; Analysis 4.3). In the LNG‐IUS 20 group, four out of six discontinuations because of pain occurred within the first three months after insertion. For discontinuation due to pain, the study arms differed somewhat but the confidence interval was wide (Analysis 4.3).
Discussion
Summary of main results
Four studies compared different contraceptive methods and one trial compared two levonorgestrel‐releasing intrauterine systems (LNG‐IUS). Of those comparing different methods, two were pilot studies and, therefore, were underpowered to find important differences (Stuart 2005; Godfrey 2010). The other two did not provide a power calculation (Suhonen 2004; Stewart 2007). However, Suhonen 2004 noted the sample size was based on 80% continuation in the LNG‐IUS group, and for Stewart 2007 the primary outcome of the trial (acceptability) differed from the outcomes for this review.
The trials that assessed pregnancy did not show any significant difference between study groups (summary of findings Table for the main comparison). In addition, the study groups did not differ significantly for method continuation (summary of findings Table 2). Among reasons for discontinuation, one study showed a significant difference between the study arms for 'other personal reasons' but not for the various other reasons. Kaunitz 2013 compared two LNG‐IUS doses in a secondary analysis from a larger trial. The study arms did not differ significantly in contraceptive effectiveness, and the overall incidence of adverse events in either group was rare. High loss to follow‐up or discontinuation may have played a role in the studies. None of the trials reported loss to follow‐up by assigned method; we obtained additional data for three studies (Stuart 2005; Stewart 2007; Kaunitz 2013).
Overall completeness and applicability of evidence
Four studies reported pregnancy and continuation data for the interventions in the study (Suhonen 2004; Stuart 2005; Stewart 2007; Godfrey 2010). One study reported pregnancy data (Kaunitz 2013). None of the studies assessed the same comparisons, precluding meta‐analysis. In addition, we did not find any trials of other commonly used contraceptives in this age group, such as the contraceptive injection or the contraceptive implant. Four studies were completed in urban settings, although they had very different participant populations (Suhonen 2004; Stuart 2005; Stewart 2007; Godfrey 2010). One study in Europe enrolled only white or Asian women (Suhonen 2004), whereas three other studies enrolled mostly low‐income Latina or African‐American women in the US (Stuart 2005; Stewart 2007; Godfrey 2010). These studies may not be applicable to populations with different racial and socioeconomic characteristics. The fifth study was conducted across Europe and the Americas (Kaunitz 2013). While that trial may have had a diverse participant population, neither the abstract nor the main report had demographic details.
Three of the five studies noted enrollment difficulties, which decreased the generalizability of the studies. Godfrey 2010 approached 37 women but randomized only 62% of these women. Stewart 2007 screened 230 women but only enrolled and randomized 56% of screened women. The most common reasons for not enrolling in the study were recent use of hormonal contraceptives, lack of interest in potential participants after receiving study information, ineligible age, and not sexually active. Stuart 2005 screened 72 women but only randomized 20% of those screened. The most commons reasons cited for non‐enrollment in the three studies were ineligibility, desire by the women to choose their contraceptive method, desire to choose another contraceptive method, and disinterest in the study. Suhonen 2004 did not report how many women were screened to recruit 200 women for the study. The main trial that served as the source for Kaunitz 2013 randomized 79% of screened women in all age groups, but the percentage of screened women who were randomized in the 18‐ to 25‐year‐old group is unknown.
Quality of the evidence
We consider the overall quality of evidence to be moderate to low. Figure 1 summarizes the risk of bias for the review overall. Figure 2 shows the risk of bias for individual trials. According to our grading system (Data synthesis), the evidence was moderate quality for three trials, low quality for one, and very low quality for one. Incomplete reporting of the methods, especially allocation concealment and loss to follow‐up, were problems in three studies (Suhonen 2004; Stewart 2007; Kaunitz 2013). Two studies blinded the participants to the intervention group assignment (Godfrey 2010; Kaunitz 2013). None of the trials blinded the study staff or investigators. The trials were all published since the first CONSORT guidelines (Consolidated Standards of Reporting Trials), which have been updated (Schulz 2010), and would be expected to have adequate reporting.
Risk of bias graph: review authors' judgments about each risk of bias item presented as percentages across all included studies.
Risk of bias summary: review authors' judgments about each risk of bias item for each included study.
Potential biases in the review process
One author (JT) was involved in Stuart 2005. JT also had access to the database for Stewart 2007. Additional data and information were more easily obtained for those studies compared with the other studies.
Agreements and disagreements with other studies or reviews
Prior studies have found that women under the age of 25 years have high rates of contraceptive failure and discontinuation within the first year of use when using oral contraceptive pills, patch or ring, or depo‐medroxyprogesterone acetate (Kost 2008; Raine 2011). The two trials included in this review that addressed contraceptive pills, ring, or patch found a slightly higher proportion of contraceptive failure compared with the general population; however, they were not powered to address this outcome (Stuart 2005; Stewart 2007; Trussell 2011). The studies examining continuation and effectiveness of intrauterine contraception in this review showed over 75% continuation of the different doses of LNG‐IUS at one year and 45% continuation of the copper IUD at one year and few pregnancies (Suhonen 2004; Godfrey 2010; Kaunitz 2013). The low continuation for the copper IUD occurred in a study underpowered to assess this outcome (Godfrey 2010), and was not consistent with data from other studies. In one large cohort study, over 75% of women aged 15 to 25 years continued both methods of intrauterine contraception and the contraceptive implant at one year (Rosenstock 2012). The low contraceptive failure with intrauterine contraception in this review is consistent with that reported elsewhere for young women (Secura 2014).
Risk of bias graph: review authors' judgments about each risk of bias item presented as percentages across all included studies.
Risk of bias summary: review authors' judgments about each risk of bias item for each included study.
Comparison 1 Copper T380A IUD versus LNG‐IUS 20 µg/day), Outcome 1 Continuation at 1 month.
Comparison 1 Copper T380A IUD versus LNG‐IUS 20 µg/day), Outcome 2 Continuation at 6 months.
Comparison 2 Combined oral contraception versus vaginal contraceptive ring, Outcome 1 Continuation.
Comparison 3 Combined oral contraception versus transdermal contraceptive patch, Outcome 1 Pregnancy at 6 months.
Comparison 3 Combined oral contraception versus transdermal contraceptive patch, Outcome 2 Continuation at 6 months.
Comparison 4 LNG‐IUS 20 µg/day versus combined oral contraception, Outcome 1 Pregnancy at 12 months.
Comparison 4 LNG‐IUS 20 µg/day versus combined oral contraception, Outcome 2 Continuation at 12 months.
Comparison 4 LNG‐IUS 20 µg/day versus combined oral contraception, Outcome 3 Discontinuation at 12 months by reason.
| Study | IUC | N | Reported unadjusted | Reported |
| Kaunitz 2013 | LNG‐IUS 12 | 566 | 0.22 | 0.01 to 1.22 |
| Kaunitz 2013 | LNG‐IUS 16 | 564 | 0.21 | 0.01 to 1.18 |
Comparison 5 LNG‐IUS 12 µg/day versus LNG‐IUS 16 µg/day, Outcome 1 Reported unadjusted Pearl Index at year 1.
| Study | IUC | N | Reported unadjusted | Reported |
| Kaunitz 2013 | LNG‐IUS 12 | 566 | 0.36 | 0.10 to 0.92 |
| Kaunitz 2013 | LNG‐IUS 16 | 564 | 0.17 | 0.02 to 0.60 |
Comparison 5 LNG‐IUS 12 µg/day versus LNG‐IUS 16 µg/day, Outcome 2 Reported unadjusted Pearl Index at year 3.
| Study | IUC | N | Reported 3‐year Kaplan‐Meier | Reported |
| Kaunitz 2013 | LNG‐IUS 12 | 566 | 0.010 | 0.004 to 0.027 |
| Kaunitz 2013 | LNG‐IUS 16 | 564 | 0.005 | 0.001 to 0.019 |
Comparison 5 LNG‐IUS 12 µg/day versus LNG‐IUS 16 µg/day, Outcome 3 Reported 3‐year Kaplan‐Meier estimated cumulative failure rates.
| Contraceptive failure rate (pregnancy) | ||||
| Patient or population: women aged ≤ 25 years with desire for contraception Settings: clinic or community Intervention: hormonal or intrauterine method Comparison: other hormonal or intrauterine method | ||||
| Outcomes | Relative effect | Participants | Quality of the evidence | Interventions |
| Pregnancy (12 months) | Not estimable (none) | 193 | ⊕⊝⊝⊝ | LNG‐IUS 20 vs. COC (EE 30 µg + desogestrel 150 µg) |
| Estimated cumulative failure rate (3 years) | Reported Kaplan‐Meier: 0.010 (0.004 to 0.027) vs. 0.005 (0.001 to 0.019) | 1130 | ⊕⊕⊝⊝ | LNG‐IUS 12 vs. LNG‐IUS 16 |
| Pregnancy (6 months) | OR 1.00 (0.05 to 18.57) | 20 | ⊕⊕⊕⊝ | COC (EE 30 µg + norgestimate 250 µg) vs. transdermal contraceptive patch |
| GRADE Working Group grades of evidence | ||||
| CI: confidence interval; LNG‐IUS 20: levonorgestrel‐releasing intrauterine system (20 µg/day); COC: combined oral contraceptive; EE = ethinyl estradiol; OR: odds ratio | ||||
| Continuation or discontinuation of contraceptive method | ||||
| Patient or population: women aged ≤ 25 years with desire for contraception Settings: clinic or community Intervention: hormonal or intrauterine method Comparison: other hormonal or intrauterine method | ||||
| Outcomes | Relative effect | Participants | Quality of the evidence | Interventions |
| Continuation (12 months) | OR 1.48 (0.76 to 2.89) | 193 | ⊕⊝⊝⊝ | LNG‐IUS 20 vs. COC (EE 30 µg + desogestrel 150 µg) |
| Continuation (1 month; 6 months) | OR 2.44 (0.19 to 31.53); | 23 | ⊕⊕⊕⊝ moderate | Cu T380A IUD vs. LNG‐IUS 20 |
| Continuation (cycles 1 to 3; cycles 4 to 6) | OR 1.09 (0.49 to 2.42); | 130 | ⊕⊕⊕⊝ | COC (EE 35 µg + norgestimate 250 µg) vs. vaginal contraceptive ring |
| Continuation (6 months) | OR 0.38 (0.05 to 2.77) | 20 | ⊕⊕⊕⊝ | COC (EE 30 µg + norgestimate 250 µg) vs. transdermal contraceptive patch |
| Discontinuation: other personal reasons (12 months) | OR 0.27 (0.09 to 0.85) | 193 | ⊕⊝⊝⊝ | LNG‐IUS 20 vs. COC (EE 30 µg + desogestrel 150 µg) |
| Discontinuation: pain, hormonal, bleeding, spotting, other medical, or planning pregnancy (12 months) | No significant difference | 193 | ⊕⊝⊝⊝ | LNG‐IUS 20 vs. COC (EE 30 µg + desogestrel 150 µg) |
| GRADE Working Group grades of evidence | ||||
| CI: confidence interval; OR: Odds ratio; LNG‐IUS: levonorgestrel‐releasing intrauterine system; COC: combined oral contraceptive; EE = ethinyl estradiol | ||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Continuation at 1 month Show forest plot | 1 | 23 | Odds Ratio (M‐H, Fixed, 95% CI) | 2.44 [0.19, 31.53] |
| 2 Continuation at 6 months Show forest plot | 1 | 23 | Odds Ratio (M‐H, Fixed, 95% CI) | 3.6 [0.62, 21.03] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Continuation Show forest plot | 1 | Odds Ratio (M‐H, Fixed, 95% CI) | Subtotals only | |
| 1.1 Cycles 1 to 3 | 1 | 130 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.09 [0.49, 2.42] |
| 1.2 Cycles 4 to 6 (cross‐over) | 1 | 98 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.61 [0.26, 1.43] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Pregnancy at 6 months Show forest plot | 1 | 20 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.0 [0.05, 18.57] |
| 2 Continuation at 6 months Show forest plot | 1 | 20 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.38 [0.05, 2.77] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Pregnancy at 12 months Show forest plot | 1 | 193 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.0 [0.0, 0.0] |
| 2 Continuation at 12 months Show forest plot | 1 | 193 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.48 [0.76, 2.89] |
| 3 Discontinuation at 12 months by reason Show forest plot | 1 | Odds Ratio (M‐H, Fixed, 95% CI) | Subtotals only | |
| 3.1 Pain | 1 | 193 | Odds Ratio (M‐H, Fixed, 95% CI) | 14.62 [0.81, 263.16] |
| 3.2 Hormonal | 1 | 193 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.44 [0.13, 1.50] |
| 3.3 Bleeding | 1 | 193 | Odds Ratio (M‐H, Fixed, 95% CI) | 5.38 [0.25, 113.51] |
| 3.4 Spotting | 1 | 193 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.35 [0.01, 8.64] |
| 3.5 Other medical | 1 | 193 | Odds Ratio (M‐H, Fixed, 95% CI) | 2.13 [0.19, 23.89] |
| 3.6 Planning pregnancy | 1 | 193 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.21 [0.01, 4.36] |
| 3.7 Other personal | 1 | 193 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.27 [0.09, 0.85] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Reported unadjusted Pearl Index at year 1 Show forest plot | Other data | No numeric data | ||
| 2 Reported unadjusted Pearl Index at year 3 Show forest plot | Other data | No numeric data | ||
| 3 Reported 3‐year Kaplan‐Meier estimated cumulative failure rates Show forest plot | Other data | No numeric data | ||
