Effect of the Combined Oral Contraceptive Pill and/or Metformin in the Management of Polycystic Ovary Syndrome

A Systematic Review With Meta-Analyses

Helena Teede; Eliza C. Tassone; Terhi Piltonen; Jaideep Malhotra; Ben W. Mol; Alexia Peña; Selma F. Witchel; Anju Joham; Veryan McAllister; Daniela Romualdi; Mala Thondan; Michael Costello; Marie L. Misso

Disclosures

Clin Endocrinol. 2019;91(4):479-489. 

In This Article

Results

A systematic search of electronic databases was conducted on 11 January 2017, with a total of 3205 studies relevant to medical treatments for PCOS identified (MEDLINE = 2266, MEDLINE Epub = 16, MEDLINE IP = 72, all EBM = 181, PsycINFO = 18, EMBASE = 439, CINAHL = 213), of which 470 duplicates were excluded. A title and abstract review was performed on the remaining 2735 studies, with 2492 excluded at this first pass stage. A total of 243 articles relevant to medical treatments for PCOS were eligible for full-text screening; however, one full text could not be retrieved. Of the 242 full-text articles screened, 56 studies met the inclusion criteria for the Metformin and/or OCP comparisons presented in this review, whilst 120 full-text articles were excluded for not meeting the PICO criteria for the Metformin and/or OCP question. A table of the excluded studies with reasons for their exclusion can be found in the guideline technical report.[19] Of the 56 studies which met the inclusion criteria for the Metformin and/or OCP comparisons investigated in this review, 46 studies are presented here in meta-analyses, and 10 studies are presented narratively (Figure 1).

Figure 1.

PRISMA flow diagram. *1 additional full text could not be retrieved [Colour figure can be viewed at wileyonlinelibrary.com]

Characteristics and Quality of Included RCTs

Key characteristics and detailed data of all 56 included studies are presented in Table S2. Treatment duration varied between 3 months and 24 months across the included studies. Sample sizes ranged from 10 participants (in a crossover study) to 253 participants. The majority of included studies were in adults (n = 6 in adolescents, n = 2 did not report age, n = 48 in adults), and the mean BMI of participants varied across studies from the normal range (18.50-24.99 kg/m2) to obese class III (≥40.00 kg/m2). In general, side effects were not adequately reported.

Detailed critical appraisals, including risk of bias and GRADE components, are presented in the technical report of the guideline.[19] Studies presented in this systematic review were found to be at low, moderate or high risk of bias. Common reasons for these ratings included the following: not reporting whether participants, investigators and/or outcome assessors were blinded to the treatment group; not reporting whether allocation to the intervention group was concealed; not reporting whether there was a published study protocol, thus rendering it impossible to know whether the paper was free of selective outcome reporting; not reporting whether the study was sufficiently powered; not reporting whether the groups were similar at baseline; not reporting whether the study was funded; and not declaring whether there were any conflicts of interest.

COCP vs Placebo—Adolescents With PCOS

One 6-month RCT[21] with 20 participants with moderate risk of bias was identified to address this comparison in adolescents (it was also included in a systematic review[22] which has been used here for the risk of bias appraisal).[19] A statistically significant difference between groups in favour of the COCP was found for high-density lipoprotein cholesterol (HDL-C; P < 0.05), whilst no statistically significant differences were found for BMI, waist circumference (WC), total testosterone, sex hormone binding globulin (SHBG), free androgen index (FAI), hirsutism, total cholesterol, low-density lipoprotein cholesterol (LDL-C), triglycerides, fasting insulin, fasting glucose, C-reactive protein (CRP) and plasminogen activator inhibitor 1 (PAI-1). Side effects were not reported.

COCP vs Lifestyle—Adolescents With PCOS

One 6-month RCT[21] with 18 participants at moderate risk of bias was identified to address this comparison in adolescents (it was also included in a systematic review[22] which has been used here for the risk of bias appraisal).[19] A statistically significant difference between groups in favour of the lifestyle intervention was found for LDL-C (P < 0.05), whilst no statistically significant differences were found for BMI, total testosterone, SHBG, FAI, hirsutism, total cholesterol, HDL, triglycerides, fasting insulin, fasting glucose, CRP and PAI-1. Side effects were not reported.

COCP vs COCP—Adults With PCOS

Eight RCTs comparing different COCPs ranged in duration from 3 to 12 months, had sample sizes between 20 and 150 participants and were rated as low-to-high risk of bias (Table S2). Due to the heterogeneity across studies in regard to the type, dose and duration of COCP use, the results could not be combined in meta-analyses and are thus presented individually.

COCP vs Metformin—Adolescents With PCOS

One systematic review including four small RCTs of 6-24 months of duration that address this comparison in adolescents was identified (Table S3). Metformin doses ranged from 1700 to 2000 mg a day. Additional analysis of outcomes (weight and fasting insulin) not addressed in the systematic review was conducted by our evidence team. Whilst a statistically significant improvement was found in BMI (−4.02 [−5.23, −2.81], P < 0.001), dysglycaemia (oral glucose tolerance test [OGTT] (0.41 [0.19, 0.86], P = 0.02) and LDL-C (−35.50 [−57.45, −13.55], P = 0.002) with use of metformin over COCP, and a statistically significant improvement found in menstrual regulation (−0.19 [−0.25, −0.13], P < 0.00001) with use of COCP over metformin, we advise caution when interpreting results due to the low quality of evidence. No statistically significant differences were found for hirsutism, total testosterone, triglyceride, total cholesterol, HDL-C, weight, fasting insulin, SHBG, FAI, fasting blood glucose, CRP and PAI-1 (Table S3). Side effects included weight gain with COCP and gastrointestinal side effects with metformin. Side effects were not consistently reported.

COCP vs Metformin—Adults With PCOS

A Cochrane systematic review was published by Costello et al in 2007 which addressed this comparison (Table S4). Four of the studies included in Costello et al 2007 were identified by our search, but were excluded due to insufficient information in the full text (marked with #). Costello 2007 obtained additional information from the authors of the RCTs. This additional information, which was not listed in the full text of the RCTs, was shared with the evidence team and allowed them to be included here. Risk of bias and data extraction were not performed on these four studies and instead were adopted from Costello 2007, whilst risk of bias and data extraction were performed on all RCTs identified by our search that were published since Costello et al 2007.

There were statistically significant improvements with metformin (compared with COCP) for fasting insulin (4.00 [2.59, 5.41], P = 0.00001), including for both BMI subgroups (<25 kg/m2 and ≥25 kg/m2). There were minor differences in impact on lipids between BMI subgroups (Table S4).

There were statistically significant improvements with COCP (compared with metformin) for SHBG (118.25 [89.65, 146.85], P = 0.00001), FAI (−6.61 [−9.75, −3.48], P = 0.0001), total testosterone (−0.38 [−0.64, −0.12], P = 0.004) and irregular cycles (12.49 [1.34, 116.62], P = 0.03), including all BMI subgroups. No statistically significant differences were found for clamp, HOMA (change from baseline), BMI, waist-to-hip ratio (WHR), hirsutism, fasting glucose and total cholesterol, whilst a P value was not reported in the one study which measured weight[23] (Table S4). Gastrointestinal (GI)-related side effects were reported by both metformin and COCP groups, along with other adverse events. The majority of included studies for this comparison were of moderate quality, and therefore, findings should be interpreted with some degree of caution.

COCP vs COCP + Metformin—Adults With PCOS

Six RCTs of moderate quality were identified to address this comparison (Table S5). The only parameter which had a statistically significant difference in favour of COCP alone was triglycerides (−0.20 [−0.39, −0.01], P = 0.04).

There were statistically significant improvements with COCP plus metformin (compared with COCP alone) for FAI (0.60 [0.35, 0.85], P = 0.00001), total testosterone (0.23 [0.01, 0.44], P = 0.04) and fasting glucose (0.38 [0.22, 0.54], P = 0.00001) when all participants were combined, whilst there were only small differences in outcomes between treatment groups for BMI subgroups (Table S5).

No statistically significant differences were found for weight, BMI, WHR, HOMA, HDL or LDL when all participants were combined (Table S5); however, sample sizes were small, and thus, results should be interpreted with caution.

GI-related events were reported by two studies, with minor GI-related events reported by 20% of participants in the combined COCP and metformin group in one study, and severe GI-related events reported by one participant in the combined COCP and metformin group in another study. No GI-related events were reported for the COCP alone group in any of the six RCTs identified.

COCP vs COCP + Anti-androgen—Adults With PCOS

Four RCTs were identified to address this comparison in adults (Table S6). There was a statistically significant improvement with COCP alone (compared with COCP plus anti-androgen) for BMI (−3.04 [−5.45, −0.64], P = 0.01) and LDL (−13.76 [−27.66, 0.14], P = 0.05). No statistically significant differences were found in meta-analyses for weight, SHBG, fasting glucose, total cholesterol and HDL, and there were no extractable data for hirsutism. There were also no differences for side effects, including headache, breast-related side effects and nausea/vomiting between the two intervention groups (Table S6). P values were not reported for single RCTs reporting on WHR, FAI or testosterone. The studies in these meta-analyses were either at moderate or high risk of bias, and therefore, all findings should be interpreted with caution.

Metformin vs Placebo—Adults and Adolescents With PCOS

Twenty RCTs that address outcomes for this comparison were identified, of which 19 RCTs were in adults and 1 was in adolescents (Table S7). Metformin doses ranged from 1500 to 1700 mg a day. As per Table S7, clinically important differences include that metformin was better than placebo for BMI (−0.48 [−0.94, −0.02], P = 0.04), testosterone (−13.87 [−27.91, 0.17], P = 0.05), total cholesterol (−10.93 [−20.53, −1.33], P = 0.03) and triglycerides (−11.85 [−21.87, −1.83], P = 0.02) when all participants were combined. In BMI ≥ 25 kg/m2 subgroup analyses, it was found that metformin offered additive benefits for weight, BMI, total cholesterol and LDL, whilst there were differences in WHR in favour of metformin in the BMI < 25 kg/m2 subgroup (Table S7).

Gastrointestinal side effects were more prevalent in the metformin groups, but only 4 out of 20 studies including in total 330 women and metformin doses of 1500-1700 mg/d reported on side effects without specific details. Ten to 62% of women taking metformin reported side effects. The majority of gastrointestinal side effects were mild to moderate and were self-limiting. The side effects reported included nausea, vomiting, diarrhoea, abdominal pain or nonspecific gastrointestinal disturbance. Only one study reported higher dropout in the metformin treated due to unacceptable gastrointestinal side effects and suggested using a lower initial metformin dose (500 mg/d). There were no reports on vitamin B12 levels (Table S7).

The majority of studies included for this comparison were at moderate risk of bias; thus, it is important to remain cautious in effect estimates and the quality of evidence across all outcomes.

Metformin vs Metformin (Dose)

One study was identified to address this compareson. Age was not reported. There was no difference in weight between the two interventions in this very low-quality study (P = 0.35).[19] Other relevant outcomes were mentioned in this study; however, no useable data were reported. The highest metformin dose used was 850 mg three times per day.

Metformin vs Metformin + COCP—Adults With PCOS

Three RCTs that address this comparison in adults were identified (Table S8). When all participants were combined, no statistically significant differences were found for weight, BMI, FAI, testosterone, fasting insulin, fasting glucose-insulin ratio, total cholesterol, HDL and LDL. P values were not reported for fasting glucose, HOMA or OGTT. Side effects were not reported. Whilst a statistically significant improvement was found in WHR (−0.03 [−0.06, −0.01], P = 0.002) and triglycerides (−26.30 [−42.99, −9.60], P = 0.002) with use of metformin over metformin plus COCP, regardless of BMI, we advise caution when interpreting results due to moderate risk of bias across all studies.

Metformin + Lifestyle vs Lifestyle ± Placebo – BMI ≥ 25—Adolescents and Adults With PCOS

A systematic review including seven relevant RCTs that address this comparison in adults and adolescents was identified (Table S9). The evidence team conducted additional analysis of outcomes not addressed in the systematic review. No statistically significant differences were found for any of the outcomes in this body of evidence which was at low-to-moderate risk of bias. Side effects were not reported.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....