Combination contraceptives: effects on weight

Maria F Gallo; Laureen M Lopez; David A Grimes; Florence Carayon; Kenneth F Schulz; Frans M Helmerhorst

doi:10.1002/14651858.CD003987.pub5

. 2014 Jan 29;2014(1):CD003987. doi: 10.1002/14651858.CD003987.pub5

Combination contraceptives: effects on weight

Maria F Gallo ¹, Laureen M Lopez ^2,^✉, David A Grimes ³, Florence Carayon ⁴, Kenneth F Schulz ⁵, Frans M Helmerhorst ⁶

Editor: Cochrane Fertility Regulation Group

PMCID: PMC10640873 PMID: 24477630

Abstract

Background

Weight gain is often considered a side effect of combination hormonal contraceptives, and many women and clinicians believe that an association exists. Concern about weight gain can limit the use of this highly effective method of contraception by deterring the initiation of its use and causing early discontinuation among users. However, a causal relationship between combination contraceptives and weight gain has not been established.

Objectives

The aim of the review was to evaluate the potential association between combination contraceptive use and changes in weight.

Search methods

In November 2013, we searched the computerized databases CENTRAL (The Cochrane Library), MEDLINE, POPLINE, EMBASE, and LILACS for studies of combination contraceptives, as well as ClinicalTrials.gov and International Clinical Trials Registry Platform (ICTRP). For the initial review, we also wrote to known investigators and manufacturers to request information about other published or unpublished trials not discovered in our search.

Selection criteria

All English‐language, randomized controlled trials were eligible if they had at least three treatment cycles and compared a combination contraceptive to a placebo or to a combination contraceptive that differed in drug, dosage, regimen, or study length.

Data collection and analysis

All titles and abstracts located in the literature searches were assessed. Data were entered and analyzed with RevMan. A second author verified the data entered. For continuous data, we calculated the mean difference and 95% confidence interval (CI) for the mean change in weight between baseline and post‐treatment measurements using a fixed‐effect model. For categorical data, such as the proportion of women who gained or lost more than a specified amount of weight, the Peto odds ratio with 95% CI was calculated.

Main results

We found 49 trials that met our inclusion criteria. The trials included 85 weight change comparisons for 52 distinct contraceptive pairs (or placebos). The four trials with a placebo or no intervention group did not find evidence supporting a causal association between combination oral contraceptives or a combination skin patch and weight change. Most comparisons of different combination contraceptives showed no substantial difference in weight. In addition, discontinuation of combination contraceptives because of weight change did not differ between groups where this was studied.

Authors' conclusions

Available evidence was insufficient to determine the effect of combination contraceptives on weight, but no large effect was evident. Trials to evaluate the link between combination contraceptives and weight change require a placebo or non‐hormonal group to control for other factors, including changes in weight over time.

Plain language summary

Effect of birth control pills and patches on weight

Weight gain is thought to be a side effect of birth control methods. Many women and healthcare providers believe that pills and patches cause weight gain. Concern about weight gain can limit the use of these effective birth control methods. Fear of weight gain keeps some women from starting the pill or patch. Women may stop using the pill because they think it caused weight gain. This review looked at trials of birth control pills or patches where the woman's weight was measured.

In November 2013, we did a computer search for studies of pills or patches containing two types of hormones. For the initial review, we also wrote to researchers and manufacturers to find other trials. We included randomized trials in the English language if they had at least three treatment cycles. The studies also had to compare two types of birth control methods or one type with a 'dummy' method.

We found 49 trials. These trials compared 52 different pairs of birth control methods, or a birth control method and a 'dummy' method. The four trials with a dummy or no method group did not show that these pills or patches led to weight change. Most studies of different birth control methods showed no large weight difference. Also, women did not stop using the pill or patch because of weight change. The evidence was not strong enough to be sure that these methods did not cause some weight change. However, we found no major effect on weight. To look at the link between these birth control methods and weight change, studies should have a 'dummy' method or a group not using hormones. Having that type of control group would help remove other factors, such as weight change over time.

Background

Weight gain is often considered a side effect of using combination contraceptives (that is, an estrogen plus a progestin) (IOM 1996; Nelson 2007), and many women and clinicians believe that an association exists. Almost three‐quarters of women in a random survey conducted in the United Kingdom reported believing that weight gain was related to oral contraceptive use (Turner 1994). In a Canadian survey of women filling an oral contraceptive (OC) prescription (Gaudet 2004), 68% had counseling from their physician on weight gain and the pill. Of those who had counseling, 36% said their weight would stay the same while on the pill compared to 50% of those who had no counseling. In the United States, 45% of adolescents starting OC use believed that oral contraceptive use increased the risk of weight gain (Emans 1987). Also, in a large German convenience sample, about 27% of ever users reported gaining weight from oral contraceptive use (Oddens 1999). In a representative sample of 3600 females in France, aged 15 to 45 years, 1665 were taking OCs (Le 2003). Of these women using the pill, 30% claimed to have gained weight on their most recent pill.

Concern about weight gain can deter the initiation of combination contraceptives and cause early discontinuation among users. Weight gain was the most frequently cited reason for oral contraceptive discontinuation in a national study of adult women in the United States (Rosenberg 1998). A second survey found that about 20% of women claimed that weight gain was a reason for oral contraceptive discontinuation or failure to initiate use (Wysocki 2000). In a convenience sample of oral contraceptive users in five European nations, women who reported weight gain had a relative risk of 1.4 (95% CI 1.2 to 1.6) of method discontinuation before two years of use compared to those who did not report a gain in weight (Rosenberg 1995). Furthermore, even the perception of weight gain can lead to contraceptive discontinuation. A United States study found that women who stopped using OCs were more likely to report weight gain than those who continued using the method, even with no significant difference in measured weight gain between the two groups (Emans 1987). Thus, concern about weight gain limits the use of a highly effective method of contraception.

Nevertheless, a causal relationship between combination contraceptives and weight gain has not been established. Several mechanisms by which combination contraceptives could lead to weight gain have been hypothesized. In general, weight gain is due to an increase in one or more factors of fluid retention, muscle mass, and fat deposition. Fluid retention could be induced by the mineralocorticoid activity that occurs when ethinyl estradiol, the estrogen in combination oral contraceptives, enters the renin‐angiotensin‐aldosterone system (Corvol 1983). Estrogen has been associated with increased subcutaneous fat, especially in the breasts, hips, and thighs (Nelson 2007). The anabolic properties of combination contraceptives could result in increased food intake through a physiological effect on satiety and appetite. Androgens may stimulate nitrogen retention and increase muscle mass, although it is unlikely that oral contraceptives would cause such weight gain (Nelson 2007).

The possible causal association between combination contraceptives and weight gain is difficult to study for several reasons. During adolescence, some weight gain is developmentally normal and appropriate. Also, women tend to gain weight over time (Flegal 2000). A contemporaneous control group is needed, but a randomized controlled trial comparing a combination contraceptive method with a placebo or non‐hormonal method for contraception raises ethical issues. Few such studies have been conducted. Comparing combination contraceptive products is complicated by the variety of formulations and regimens. In addition, most combination contraceptive studies have been of short duration (that is, six cycles or fewer); more time might be required for the weight gain to become evident. Finally, no consensus exists regarding what constitutes excessive weight gain. Ideally, studies would set an a priori definition of clinically important weight gain, but this is rarely specified, perhaps because weight change is not a primary outcome in most comparison trials of combination contraception. Most studies that present a dichotomous classification for weight gain selected either 2.0 or 2.3 kilograms as the cut point; however, the justification for this decision is not apparent. Even if clinically important weight gain was well specified, any gain in weight could still be relevant since the mere perception of weight gain is associated with discontinuation of oral contraception.

Objectives

The aim of the review was to evaluate the potential association between combination contraceptive use and changes in weight. The primary hypothesis was that combination contraceptives do not result in weight changes greater than that of a placebo. The secondary hypothesis was that different formulations and regimens of combination contraceptives are not associated with differences in weight changes.

Methods

Criteria for considering studies for this review

Types of studies

All randomized controlled trials reported in English (Juni 2002; Moher 2000) that compared a combination contraceptive to a placebo, no intervention, or a combination contraceptive that differed in drug, dosage, regimen, or study length.

Types of participants

Women of reproductive age without medical contraindications to combination contraceptives.

Types of interventions

Any combination contraceptive compared to a placebo, no intervention, or another combination contraceptive. Trial drug interventions must have included at least three consecutive cycles to be eligible.

Types of outcome measures

Trials must have collected data on change in body weight to be eligible for inclusion. Weight change could have been measured as either the change in the study group's mean weight or as the proportion of the study group who lost or gained more than a specified amount.

Search methods for identification of studies

Electronic searches

In November 2013, we searched the computerized databases Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library), MEDLINE, POPLINE, EMBASE, and LILACS for studies of combination contraceptives. We also searched for trials via ClinicalTrials.gov and the search portal of the International Clinical Trials Registry Platform (ICTRP). The strategies are given below.

Searching other resources

For the initial review, we wrote to known investigators and manufacturers to request information about other published or unpublished trials not discovered in our search.

Data collection and analysis

Selection of studies

All titles and abstracts identified during the literature searches were assessed for inclusion, and all potentially relevant articles were photocopied. For the initial review, we wrote to the manufacturers of combination contraception and authors of the included trials to seek other published or unpublished trials.

Data extraction and management

The abstracted data were entered into RevMan and were double‐checked by a second author. The analysis depended on the data available. For the mean change in weight between baseline and post‐treatment measurements, the mean difference (MD) with 95% confidence interval (CI) was calculated using a fixed‐effect model. Alternatively, the Peto odds ratio (OR) with 95% CI was calculated using the proportion of women who gained or lost more than a specified amount of weight.

Significant weight change could be considered a negative side effect of contraceptive use. We used a consistent direction for the graph labels even though the outcomes differed. Therefore, the intervention ‘favors treatment’ if the change is greater in the control group. The intervention ‘favors control’ if the change is greater in the treatment group.

Assessment of risk of bias in included studies

The validity of trials was critically appraised by assessing potential biases; however, summary quality scores were not calculated since the available evidence does not support their use (Higgins 2011). The appraisal of potential biases concentrated on the study design, blinding, randomization method, group allocation concealment, and loss to follow up and early discontinuation.

Measures of treatment effect

Despite some trials reporting weight change data for multiple cycles, only one weight change measure was abstracted for each intervention group. We chose the cycle 12, cycle 6, or last treatment cycle data (in this order of preference) to facilitate comparisons between trials. For trials that included more than two interventions, comparisons were made between the control group and the other groups. If the authors did not identify a control group, all possible combinations were included in the review. We did not use any technique to control for multiple testing.

Significance testing has been criticized for forcing the decision to recognize a difference between two interventions to rest on an arbitrary alpha‐level. Rather than determine differences solely based on a dichotomous classification of the P value, analyses using interval estimation consider both the location of the point estimate and the spread of the CI (Rothman 1998). While this process introduces more subjectivity, borderline 'statistically significant' results are not overlooked and clinically insignificant results are not ascribed undue importance. Although interval estimation is a preferable method, given the reliance on significance testing, we presented results based on both considerations.

Dealing with missing data

For the initial review, we wrote to the authors of 94 trials that either did not report weight data or reported insufficient details. At that point we revised and broadened the search strategy and decided that continuing to contact the authors of trials that appeared to be eligible except for a lack of weight data was no longer feasible. The 14 trials with additional, unpublished data supplied by the authors contacted before the protocol change were included in the present review.

Since point estimates that are not accompanied by a measure of sampling variation have limited interpretation, trial reports that did not include the standard error, standard deviation, or CI for the mean change in weight were not included in the review. For trials that did not report the denominators used to calculate the mean weight difference, or the percentage of women experiencing weight change, we estimated these numbers based on denominators used for other outcomes or the number of women who completed the trial.

Data synthesis

Studies were combined for a meta‐analysis only when identical drugs, dosages, regimens, and delivery systems were compared. No sensitivity analyses were planned since few trials were anticipated to be eligible for meta‐analysis. The data abstracted for the review were dependent on the analytic method used in the trial report (for example, intent to treat, per protocol, or a modification of either type).

Results

Description of studies

Results of the search

Initial review

The initial search strategy in 2002 yielded 570 reports of randomized controlled trials that compared a combination contraceptive to a placebo or to another combination contraceptive. Of those, 476 articles were not eligible for inclusion due to a study length less than three treatment cycles in duration or a lack of reported weight change. An additional 53 articles with weight data were excluded since they lacked an estimate of the sampling variability for the mean difference in weight. The 41 eligible articles, including the 14 articles with additional unpublished data from the authors, reported on 44 trials. One article (Oelkers 2000) described two trials, and two articles (Coney 2001; Kaunitz 2000) each reported pooled results from two eligible trials with similar or identical protocols. Since they could not be disaggregated, we treated the pooled results as if they were from one larger trial, for a total of 42 trials.

Updates

2005: 12 new RCTs of combination contraceptives also included a weight measurement. Only two trials were of sufficient duration and had reported weight change data with sampling variability, which yielded 44 trials.
2008: 13 trials had weight measurements, but only three had sufficient data for inclusion, for a new total of 47 trials.
2011: Of five possible trials, two met our inclusion criteria. The new total was 49 trials.
2013: The search yielded 134 unduplicated citations from the electronic databases. We reviewed the full text of eight articles; none met our inclusion criteria. In addition, we found 32 unduplicated listings in ClinicalTrials.gov and ICTRP. We identified one ongoing trial that was relevant (Mahidol 2013).

Included studies

Of the 49 eligible trials, four included a placebo group or no hormonal method. Three of these trials evaluated oral contraceptives (Coney 2001; Goldzieher 1971; Procter‐Gray 2008) and one studied a contraceptive skin patch (Sibai 2001). The products evaluated in the 49 trials included 18 progestins and three estrogens. Trials examined combination oral contraceptives except for the following comparisons: two combination injectables (Sang 1995); two combination vaginal rings (Weisberg 1999); a combination skin patch (Sibai 2001; Stewart 2005); and a combination ring with an oral contraceptive (Milsom 2006; Oddsson 2005). Seven trials included more than two intervention groups; three of these trials did not specify a control group.

The sample sizes for the trials ranged from 20 to 5654 randomized participants with a median of 196 participants. The study location was not described for 13 trials; the other studies were conducted in locations worldwide. The number of trial sites ranged from a single site (12 trials) to 131 sites, except for 10 trials that did not specify the number of sites. The duration of the trials ranged from 3 to 24 treatment cycles with most trials designed to be either 6 or 12 treatment cycles in length. The eligibility criteria for the participants varied among the trials with most trials recruiting healthy women of reproductive age without contraindications to hormonal contraceptive use. However, six of the articles did not describe any inclusion or exclusion criteria.

Risk of bias in included studies

The quality of the reporting of the trials on this topic was generally poor, and poor quality is associated with empirical evidence of bias (Schulz 1995).

Allocation

The method of generating the randomization sequence was not reported for 31 trials. The remaining 18 trials included at least some detail of the process (for example, use of a random numbers table, pre‐distributed lists, or computer‐generated sequence). Most trial reports (N = 45) did not describe a method of allocation concealment (Schulz 2002a). Cachrimanidou 1993 and Kashanian 2010 reported the use of sealed envelopes but did not provide details on whether the envelopes were impervious to deciphering (for example, use of opaque, sequentially‐numbered envelopes). Only three articles reported adequate allocation concealment: Miller 2001 used sealed, sequentially‐numbered, opaque envelopes containing carbon paper, which allowed the participant to sign the allocation card before study staff opened the envelope and learned the group assignment; Milsom 2006 and Oddsson 2005 had interactive voice response systems for the randomization process.

Only four articles reported the number of women recruited for the trial (Cachrimanidou 1993; Kashanian 2010; Oddsson 2005; Wiegratz 2002). One article stated that not all randomized women were included in the study results but did not specify the number of randomized women (Worsley 1980). A second trial report included sample sizes for the weight outcome but did not state whether these data included all randomized women (Sibai 2001). Although they reported the total number of randomized women, seven trial reports did not provide the number of randomized women stratified by study group.

Blinding

About half of the eligible trials were open; two were single‐blinded, 10 were double‐blinded, and one was triple‐blinded. Blinding was not mentioned in 15 trials. Participants, investigators, and outcome assessors were blinded as to group assignment in the triple‐blinded study (Oelkers 1995), and participants appeared to have been blinded in two of the double‐blinded trials (Coney 2001; Goldzieher 1971) using active and placebo pills that were identical in appearance. Since the Goldzieher 1971 trial reported that the randomization code was not broken during the study, the blinding of the investigators can be inferred. Two trials did not inform the assessors of group assignment, but the investigators and participants were not blinded (Kashanian 2010; Procter‐Gray 2008). The remaining blinded trials were unclear about who was blinded, and none of the trials included details regarding whether blinding appeared to have been implemented successfully (Schulz 2002b).

Incomplete outcome data

Eleven trials did not report the denominators used to derive the mean weight difference (Coenen 1996) or the percentage of women with weight change greater than a specified amount (Brill 1991; Dionne 1974; Endrikat 1999; Endrikat 2001a; Goldzieher 1971; Halbe 1998; Koetsawang 1995; Lachnit‐Fixson 1984; Oelkers 1995; Rosenbaum 2000). Discrepancies existed in at least 13 trials, since more women were missing in the weight estimate than could be explained with the possible reasons in the article (for example, non‐starters, early discontinuation, those lost to follow up, or exclusion).

Deducing which participants were included in the reported weight change estimate was hindered for most trials due to the lack of details regarding the method of analysis. An intention‐to‐treat analysis was described for three trials (Oddsson 2005; Procter‐Gray 2008; Wiegratz 2002); a per‐protocol analysis (also called valid‐case) was reported for three trials (Endrikat 1999; Serfaty 1998; Winkler 1996); and an analysis based on all participants who started treatment was reported for five trials (Coney 2001; Endrikat 2001b; Gruber 2006; Kaunitz 2000; Milsom 2006). Two trials (Endrikat 2001a; Stewart 2005) described intention‐to‐treat or valid‐case analyses but did not specify which was used for the weight change data. Two trials (Spellacy 1970; Van der Does 1995) had complete study participation and based the weight estimates on measurements from all participants. Kashanian 2010 excluded from the analysis two women who discontinued the intervention. The remaining 33 trials did not specify the analytic method used for the weight change data.

Other potential sources of bias

Change in body weight was a primary outcome for only one trial (Sibai 2001). Most trials either recorded weight at the baseline and follow‐up clinic visits or did not describe the method used for measuring weight. Liukko 1987 reported that the clinic visits, in which weight was measured, were scheduled around day 22 of the cycle. Milsom 2006 and Procter‐Gray 2008 described the methods used to standardize weight measurements. Four trials instructed participants to weigh and record their weight at home (Rosenbaum 2000), at home without clothing every second day (Oelkers 2000, Study 2) and in a fasting state (Oelkers 1995), or at home without clothing on a weekly basis (Oelkers 2000, Study 3).

Most trial reports either did not include data on loss to follow up (19 trials) or reported it combined with early discontinuation (nine trials). Loss to follow up ranged from zero to 17% for the 20 trials with loss to follow up reported separately from early discontinuation and from 10% to 35% for those that only reported losses from all causes. Early discontinuation ranged from zero to 39% for the 31 trials that reported this study factor separately. Eight trial reports did not include information on early discontinuation. Furthermore, 18 trial reports described excluding randomized women from study participation or analysis. Exclusions after randomization are not consistent with an intention‐to‐treat analysis and could have led to biased results (Weiss 1998).

Effects of interventions

The trials included 85 weight change comparisons for 52 distinct contraceptive pairs (or a placebo) that were eligible for the present review. We combined data from two or more trials for four of the 84 comparisons since they were identical in drug, dosage, regimen, and study length. The comparisons of a combination contraceptive with a placebo or no hormonal method showed no significant differences in weight change. These included five comparisons between an oral contraceptive and placebo (Coney 2001; Goldzieher 1971) or no intervention (Procter‐Gray 2008) and one comparison between a combination skin patch and placebo (Sibai 2001).

The study groups differed significantly in six comparisons of two pills and one comparison of the vaginal ring with an OC.

Three studies showed differences in the numbers of women with a weight change of more than 2 kg.
- For gaining more than 2 kg, the OR was 3.29 (95% CI 1.84 to 5.88) for women assigned to oral desogestrel 150 μg and ethinyl estradiol (EE) 30 μg compared to those with oral levonorgestrel 50‐75‐125 μg and EE 30‐40‐30 μg (Lachnit‐Fixson 1984) (Analysis 11.1).
- For losing more than 2 kg, the OR was 9.22 (95% CI 1.79 to 55.04) for oral drospirenone 3 mg and EE 15 μg compared to oral levonorgestrel 150 μg and EE 30 μg (Oelkers 1995) (Analysis 20.2).
- The OR for losing more than 2 kg was 1.65 (95% CI 1.13 to 2.41) for the oral desogestrel 150 μg and EE 20 μg group compared to the oral gestodene 75 μg and EE 20 μg group (Serfaty 1998) (Analysis 9.2).

Four studies showed differences in the mean weight change between groups. One group had a slight increase while the other group showed a small decrease or no change.
- The mean difference in Kaunitz 2000 was 0.26 kg (95% CI 0.12 to 0.40) (Analysis 46.1). Women assigned to oral norethindrone 500‐750‐1000 μg and EE 35 μg group had a slight mean increase, and those in the oral desogestrel 100‐125‐150 μg and EE 25 μg group had a slight decrease.
- In Loudon 1990, the mean difference was 0.70 kg (95% CI 0.14 to 1.26) (Analysis 35.1). The group with oral levonorgestrel 150 μg and EE 30 μg gained, on average, while those with oral gestodene 75 μg and EE 30 μg lost a little weight.
- The mean difference was ‐0.67 kg (95% CI ‐1.16 to ‐0.18) in Gruber 2006 (Analysis 23.1). On average, the women assigned to drospirenone 3 mg and EE 20 μg had a slight decrease, and the group with desogestrel 150 μg and EE 20 μg had a small increase.
- Tn Milsom 2006, the mean difference was 0.40 kg (95% CI 0.03 to 0.77) (Analysis 52.1). Women assigned to the vaginal ring gained on average while the group with the combination oral contraceptive (COC) containing drospirenone 3 mg and EE 30 µg had no change overall.

11.1 — Comparison 11 Desogestrel 150 µg and EE 30 µg versus levonorgestrel 50‐75‐125 µg and EE 30‐40‐30 µg, Outcome 1 Gained >2 kg (cycle 6).

20.2 — Comparison 20 Drospirenone 3 mg and EE 15 µg versus levonorgestrel 150 µg and EE 30 µg, Outcome 2 Lost >2 kg (cycle 6).

9.2 — Comparison 9 Desogestrel 150 µg and EE 20 µg versus gestodene 75 µg and EE 20 µg, Outcome 2 Lost >2 kg (cycle 6).

46.1 — Comparison 46 Norethindrone 500‐750‐1000 µg and EE 35 µg versus desogestrel 100‐125‐150 µg and EE 25 µg, Outcome 1 Mean weight change in kg (cycle 6).

35.1 — Comparison 35 Levonorgestrel 150 µg and EE 30 µg versus gestodene 75 µg and EE 30 µg, Outcome 1 Mean weight change in kg (cycle 6).

23.1 — Comparison 23 Drospirenone 3 mg and EE 20 μg versus desogestrel 150 μg and EE 20 μg, Outcome 1 Mean weight change in kg (cycle 7).

52.1 — Comparison 52 Vaginal ring etonogestrel 120 µg and EE 15 µg versus drospirenone 3 mg and EE 30 µg, Outcome 1 Mean weight change in kg (cycle 13 or last assessment).

Six additional comparisons had a point estimate and 95% CI that were consistent with at least a minimal difference between the two groups. Odds ratios for three of the six studies were as follows:

OR 1.54 (95% CI 0.92 to 2.60) for gaining more than 2 kg with oral gestodene 75 μg and EE 30 μg versus oral norgestimate 250 μg and EE 35 μg (Brill 1991) (Analysis 29.1);
OR 1.75 (95% CI 0.98 to 3.11) for gaining more than 2.5 kg with oral lynestrenol 2 mg and EE 40 μg versus oral lynestrenol 1 mg and EE 40 μg (Koetsawang 1977) (Analysis 42.1);
OR 1.69 (95% CI 0.89 to 3.20) of losing more than 2 kg with oral norethisterone 500 μg and EE 20 μg versus oral levonorgestrel 150 μg and EE 30 μg (Endrikat 2001b) (Analysis 45.2).

29.1 — Comparison 29 Gestodene 75 µg and EE 30 µg versus norgestimate 250 µg and EE 35 µg, Outcome 1 Gained >2 kg (cycle 6).

42.1 — Comparison 42 Lynestrenol 2 mg and EE 40 µg versus lynestrenol 1 mg and EE 40 µg, Outcome 1 Gained >2.5 kg (cycle 12).

45.2 — Comparison 45 Norethisterone 500 µg and EE 20 µg versus levonorgestrel 150 µg and EE 30 µg, Outcome 2 Lost >2 kg (cycle 6).

Mean differences in kg were as follows for the other three studies:

MD 1.30 (95% CI ‐0.32 to 2.92) between oral levonorgestrel 50, 75, 125 μg and EE 30, 40, 30 μg group and oral desogestrel 50‐100‐150 μg and EE 35‐30‐30 μg group (Van der Does 1995) (Analysis 38.2);
MD 1.80 (95% CI ‐0.73 to 4.33) between a standard versus prolonged oral norgestrel and EE regimen (Miller 2001) (Analysis 48.1);
MD 1.14 (95% CI ‐0.54 to 2.82) between oral dl‐norgestrel 500 μg and EE 50 μg group and oral norethisterone acetate 1 mg and EE 50 μg group (Worsley 1980) (Analysis 18.1).

38.2 — Comparison 38 Levonorgestrel 50‐75‐125 µg and EE 30‐40‐30 µg versus desogestrel 50‐100‐150 µg and EE 35‐30‐30 µg, Outcome 2 Mean weight change in kg (cycle 6).

48.1 — Comparison 48 Standard norgestrel and EE regimen versus prolonged norgestrel and EE regimen, Outcome 1 Mean weight change in kg (cycle 12).

18.1 — Comparison 18 Dl‐norgestrel 500 µg and EE 50 µg versus norethisterone acetate 1 mg and EE 50 µg, Outcome 1 Mean weight change in kg (cycle 3).

Twenty‐one trial reports provided quantitative data on the primary reasons for early discontinuation. Ten of these trial reports included weight change as a distinct category for early discontinuation from trial participation (Table 1). The proportions of women who discontinued due to weight changes were small (zero to 5%) and did not differ between the study groups. Sang 1995 included the mean weight gain for the women who cited weight gain as the primary or secondary reason for early discontinuation. In the group with injectable medroxyprogesterone acetate 25 mg and estradiol cypionate 5 mg, the 18 women who attributed weight gain to the contraceptive had a mean gain of 2.8 kg compared to 3.1 kg for the 16 women in the group assigned to the injectable norethisterone enanthate 50 mg and estradiol valerate 5 mg (Sang 1995).

1. Discontinuation due to weight change.

Study ID	Intervention group	n	N (randomized women)
Cachrimanidou 1993	Prolonged desogestrel / ethinyl estradiol (EE) regimen	10	200
	Standard desogestrel / EE regimen	1	100
Coenen 1996	Norgestimate 250 μg / EE 35 μg	1	25
	Gestodene 75 μg / EE 30 μg	0	25
	Desogestrel 150 μg / EE 30 μg	0	25
	Desogestrel 150 μg / EE 20 μg	0	25
Coney 2001	Levonorgestrel 100 μg / EE 20 μg	2	359
	Placebo	0	362
Dionne 1974	Levonorgestrel 250 μg / EE 50 μg	3	73
	Levonorgestrel 150 μg / EE 30 μg	1	77
Halbe 1998	Gestodene 75 μg / EE 30 μg	4	279
	Desogestrel 150 μg / EE 30 μg	0	316
Kirkman 1994	Gestodene 75 μg / EE 30 μg	4	505
	Desogestrel 150 μg / EE 20 μg	2	501
Miller 2001	Standard norgestrel / EE regimen	0	44
	Prolonged norgestrel / EE regimen	1	46
Oddsson 2005	Vaginal ring etonogestrel 120 µg / EE 15 µg	2	512
	Levonorgestrel 150 µg / EE 30 µg	6	518
Sang 1995	Injectable medroxyprogesterone acetate 25 mg / estradiol cypionatge (EC) 5 mg	14	1955
	Injectable norethisterone enanthate 50 mg / estradiol valerate (EV) 5 mg	10	1960
Wiik 1993	Norethisterone 500‐1000 μg / EE 35 μg	3	100
	Levonorgestrel 50‐75‐125 μg / EE 30‐40 μg	1	96

Date	Event	Description
1 January 2014	New citation required but conclusions have not changed	Searches updated
18 December 2013	New search has been performed	No new trials met inclusion criteria.   Added one ongoing trial (Mahidol 2013).  Added Figure 1 and Figure 2 to summarize risk of bias.

Date	Event	Description
31 May 2011	New search has been performed	MEDLINE search was updated. No additional trials found.
28 February 2011	New citation required but conclusions have not changed	Two new trials included (Kashanian 2010; Procter‐Gray 2008)
25 February 2011	New search has been performed	Searches were updated. Searches added for ClinicalTrials.gov and ICTRP.
17 June 2008	New citation required but conclusions have not changed	Three new trials were added (Gruber 2006; Burkman 2007; Milsom 2006). Several recent trials were excluded.
16 June 2008	New search has been performed	Searches were updated in May and June 2008.
9 May 2008	Amended	Converted to new review format.
29 September 2005	New citation required and conclusions have changed	Substantive amendment

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >5% baseline weight (cycle 9)	1	136	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.95 [0.30, 2.98]
2 Lost >5% baseline weight (cycle 9)	1	136	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.27 [0.04, 1.82]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 6)	1	801	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.84 [0.58, 1.22]
2 Lost >2 kg (cycle 6)	1	801	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.65 [1.13, 2.41]
3 Gained >2 kg (cycle 12)	1	1476	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.13 [0.85, 1.49]
4 Lost >2 kg (cycle 12)	1	1476	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.95 [0.68, 1.33]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >5% baseline weight (cycle 3)	1	49	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.20 [0.32, 4.51]
2 Lost >5% baseline weight (cycle 3)	1	49	Peto Odds Ratio (Peto, Fixed, 95% CI)	7.10 [0.14, 358.08]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >5% baseline weight (cycle 3)	1	49	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.95 [0.24, 3.76]
2 Lost >5% baseline weight (cycle 3)	1	49	Peto Odds Ratio (Peto, Fixed, 95% CI)	7.40 [0.45, 121.93]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >5% baseline weight (cycle 3)	1	48	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.77 [0.18, 3.21]
2 Lost >5% baseline weight (cycle 3)	1	48	Peto Odds Ratio (Peto, Fixed, 95% CI)	7.39 [0.15, 372.38]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 3)	2	112	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.14 [0.00, 6.82]
2 Lost >2 kg (cycle 3)	1	66	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.99 [0.20, 19.82]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 6)	1	40	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.50 [0.05, 5.06]
2 Lost >2 kg (cycle 6)	1	40	Peto Odds Ratio (Peto, Fixed, 95% CI)	9.92 [1.79, 55.04]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 6)	1	40	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.13 [0.01, 2.13]
2 Lost >2 kg (cycle 6)	1	40	Peto Odds Ratio (Peto, Fixed, 95% CI)	7.39 [0.15, 372.38]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 13)	1	56	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.86 [0.29, 2.56]
2 Lost >2 kg (cycle 13)	1	56	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.38 [0.29, 6.62]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 6)	1	39	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.05 [0.06, 17.51]
2 Lost >2 kg (cycle 6)	1	39	Peto Odds Ratio (Peto, Fixed, 95% CI)	7.79 [0.15, 393.02]
3 Gained >2 kg (cycle 12)	1	452	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.06 [0.63, 1.81]
4 Lost >2 kg (cycle 12)	1	452	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.13 [0.63, 2.03]
5 Gained >2 kg (cycle 13)	1	40	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.71 [0.14, 3.57]
6 Lost >2 kg (cycle 13)	1	40	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.14 [0.00, 6.82]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mean body mass percentage change (cycle 6)	1	43	Mean Difference (IV, Fixed, 95% CI)	0.7 [‐1.32, 2.72]
2 Mean weight change in kg (cycle 6)	1	805	Mean Difference (IV, Fixed, 95% CI)	0.20 [0.00, 0.40]
3 Mean weight change in kg (cycle 12)	2	462	Mean Difference (IV, Fixed, 95% CI)	0.01 [‐0.50, 0.51]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 6)	3	1524	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.18 [0.87, 1.60]
2 Lost >2 kg ( cycle 6)	2	1172	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.34 [0.90, 2.00]
3 Mean body mass percentage change (cycle 6)	1	46	Mean Difference (IV, Fixed, 95% CI)	0.8 [‐1.18, 2.78]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 6)	1	357	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.54 [0.92, 2.60]
2 Mean body mass percentage change (cycle 6)	1	43	Mean Difference (IV, Fixed, 95% CI)	1.3 [‐1.03, 3.63]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 6)	1	418	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.26 [0.74, 2.15]
2 Lost >2 kg (cycle 6)	1	418	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.31 [0.70, 2.44]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2.5 kg (cycle 24)	1	17	Peto Odds Ratio (Peto, Fixed, 95% CI)	8.66 [0.77, 97.74]
2 Lost >2.5 kg (cycle 24)	1	17	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.88 [0.17, 21.18]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mean BMI change (cycle 6)	1	57	Mean Difference (IV, Fixed, 95% CI)	0.35 [‐0.13, 0.83]
2 Mean weight change in kg (cycle 6)	1	31	Mean Difference (IV, Fixed, 95% CI)	1.30 [‐0.32, 2.92]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2.5 kg (cycle 12)	1	228	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.75 [0.98, 3.11]
2 Lost >2.5 kg (cycle 12)	1	228	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.66 [0.25, 1.77]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 6)	1	349	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.15 [0.65, 2.06]
2 Mean body mass percentage change (cycle 6)	1	45	Mean Difference (IV, Fixed, 95% CI)	‐0.5 [‐2.51, 1.51]

PERMALINK

Combination contraceptives: effects on weight

Maria F Gallo

Laureen M Lopez

David A Grimes

Florence Carayon

Kenneth F Schulz

Frans M Helmerhorst

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Background

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Dealing with missing data

Data synthesis

Results

Description of studies

Results of the search

Initial review

Updates

Included studies

Risk of bias in included studies

Allocation

Blinding

Incomplete outcome data

Other potential sources of bias

Effects of interventions

11.1. Analysis.

20.2. Analysis.

9.2. Analysis.

46.1. Analysis.

35.1. Analysis.

23.1. Analysis.

52.1. Analysis.

29.1. Analysis.

42.1. Analysis.

45.2. Analysis.

38.2. Analysis.

48.1. Analysis.

18.1. Analysis.

1. Discontinuation due to weight change.

Discussion

Summary of main results

Overall completeness and applicability of evidence

Quality of the evidence

1.

2.

Authors' conclusions

Implications for practice.

Implications for research.

What's new

History

Acknowledgements

Appendices

Appendix 1. Search 2013

MEDLINE via PubMed (01 Jan 2011 to 01 Jan 2014)

CENTRAL (01 Jan 2011 to 11 Nov 2013)

POPLINE (01 Jan 2011 to 06 Nov 2013)

EMBASE (01 Jan 2011 to 11 Nov 2013)

LILACS (01 Jan 2011 to 06 Nov 2013)

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gained >2 kg (cycle 6)	1	292	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.09 [0.60, 1.99]
2 Lost >2 kg (cycle 6)	1	292	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.69 [0.89, 3.20]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Weight gain >=5% (cycle 6)	1	2157	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.15 [0.91, 1.45]
2 Weight gain >=5% (cycle 13)	1	453	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.09 [0.69, 1.74]
3 Weight loss >=5% (cycle 6)	1	2157	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.99 [0.72, 1.37]
4 Weight loss >=5% (cycle 13)	1	453	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.32 [0.74, 2.34]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gain >2 kg (cycle 4)	1	51	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.69 [0.13, 3.58]
2 Lost >2 kg (cycle 4)	1	51	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.69 [0.35, 8.07]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Gain >=7% body weight (cycle 13)	1	1030	Peto Odds Ratio (Peto, Fixed, 95% CI)	0.84 [0.55, 1.28]
2 Lost >=7% body weight (cycle 13)	1	1030	Peto Odds Ratio (Peto, Fixed, 95% CI)	1.39 [0.83, 2.32]

Methods	Study location not described.  Cross‐over trial but weight data available from first treatment period only.  Two pre‐treatment 'washout' cycles and three treatment cycles.
Participants	Healthy women age 21 to 32 years with regular menses. Excluded recent hormonal contraceptive use; recent use of certain drugs.
Interventions	Levonorgestrel 50‐75‐150 µg and EE 30‐40‐30 µg versus levonorgestrel 50‐75‐150 µg and EE 30‐40‐30 µg.  29 women randomized; initial number assigned to each study group not reported.
Outcomes	Adverse events, hormonal measurements.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomization scheme.
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	No information
Incomplete outcome data (attrition bias)   All outcomes	Unclear risk	Three women discontinued early. Primary reasons for discontinuation described and did not include weight gain.  Loss to follow up not reported.

Methods	One site in Indonesia.  12 treatment cycles.
Participants	Inclusion and exclusion criteria not described.
Interventions	Gestodene 50‐70‐100 µg and EE 30‐40‐30 µg (N=13) versus desogestrel 150 µg and EE 30 µg (N=17).
Outcomes	Lipoprotein, liver function, blood coagulation, adverse events, body weight, blood pressure.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomization scheme.
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Low risk	"Double‐blinded" but did not report who was blinded.
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Three women discontinued early or were lost to follow up. Primary reasons for discontinuation not described.

Methods	Multicenter trial in Germany.  Six treatment cycles.
Participants	Healthy, sexually‐active women age 16 to 45 years with regular menses.  Excluded contraindications to oral contraceptive use; recent oral contraceptive use; certain drug use; abnormal Pap smear.
Interventions	Gestodene 75 µg and EE 30 µg (N=209) versus desogestrel 150 µg and EE 30 µg (N=201) versus norgestimate 250 µg and EE 35 µg (N=195).
Outcomes	Contraceptive efficacy, cycle control, body weight, blood pressure, adverse events.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	No information
Incomplete outcome data (attrition bias)   All outcomes	Low risk	29 women in the gestodene, 29 women in the desogestrel and 18 women in the norgestimate group discontinued early or were lost to follow up. Primary reasons for discontinuation described and did not include weight gain.

Methods	One site.  Three pill‐free pretreatment cycles and 13 treatment cycles.
Participants	Women age 18 to 35 years with regular menses.  Excluded smokers over age 30 years; pregnancy; liver disease; vascular disease; tumors; certain other diseases; obesity; heavy alcohol use; other hormone preparations or intrauterine device use.
Interventions	Gestodene 75 µg and EE 20 µg (N=32) versus gestodene 75 µg and EE 30 µg (N=32).
Outcomes	Lipid levels, hormone levels, efficacy, cycle control, safety.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	High risk	Unblinded
Incomplete outcome data (attrition bias)   All outcomes	Unclear risk	1 woman in EE 20 µg and 5 in EE 30 µg group withdrew before starting treatment. 4 in EE 20 µg and 1 in EE 30 µg group discontinued early due to adverse events. Primary reasons for discontinuation not described.  3 women in EE 20 µg group were excluded for protocol violations.  Lost to follow up not reported.

Methods	100 sites in USA and 10 in Canada.  First 1/3 of participants were to have 13 treatment cycles and the remaining 2/3 were to have 6 treatment cycles.
Participants	Sexually active, healthy women aged 18 to 45 years at risk for pregnancy with regular menstrual cycles, blood pressure <140/90.  Excluded recent pregnancy or lactation; contraindications to OCs; certain diseases; smokers aged 35 or more years; certain drugs or devices; recent DMPA use; and recent alcohol or substance abuse.
Interventions	Norethindrone acetate (NETA) 1.0 mg plus EE 20 µg, with 75 mg ferrous fumarate on days 22‐28 (N=853 for 6 cycles, 318 for 13 cycles) versus norgestimate (NGM) 180‐215‐250 µg plus EE 25 µg (N=1236 for 6 cycles, 487 for 13 cycles).
Outcomes	Weight change was primary outcome; contraceptive efficacy, cycle control, and safety were in earlier report (Hampton 2001).  'Breakthrough' bleeding or spotting defined as bleeding or spotting that occurred during the active pill days unless it was contiguous with menses.  'Amenorrhea' defined as two consecutive cycles without any bleeding or spotting.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Allocated with block sizes of 11
Allocation concealment (selection bias)	Unclear risk	No information
Blinding (performance bias and detection bias)   All outcomes	Low risk	Blinded (participants and at least the assessors at cycle 3).
Incomplete outcome data (attrition bias)   All outcomes	High risk	Lost to follow up reportedly 6.5% in NGM/EE group and 5.8% in NETA/EE group. Noncompleters were 21% of 6‐cycle groups; 42% to 40% of 13‐cycle groups, respectively.

Methods	Three sites in Sweden.  12 treatment cycles.
Participants	Healthy women age 18 to 39 years at risk of pregnancy.  Excluded "generally accepted" contraindications of OC use.
Interventions	Prolonged regimen (desogestrel 150 µg and EE 30 µg; nine pill weeks and one pill‐free week; N=198) versus standard regimen (desogestrel 96 µg and EE 30 µg; three pill weeks and one pill‐free week; N=96).
Outcomes	Lipoprotein, liver function, blood coagulation, adverse events, body weight, blood pressure.
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information
Allocation concealment (selection bias)	Unclear risk	Allocated with sealed envelopes.
Blinding (performance bias and detection bias)   All outcomes	Unclear risk	No information
Incomplete outcome data (attrition bias)   All outcomes	High risk	83 women in the prolonged regimen and 32 women in the standard regimen group discontinued early.  Primary reasons for discontinuation described; 10 women in the prolonged and one woman in the standard regimen group cited weight gain.  Loss to follow up not reported.