Metformin and Combined Oral Contraceptive Pills in the Management of Polycystic Ovary Syndrome: A Systematic Review and Meta-analysis.

Materials and Methods

This systematic review and meta-analysis was conducted as part of the 2023 update of the International Evidence-based Guidelines for the Assessment and Management of PCOS (18). We addressed the efficacy of (i) metformin compared to COCP; (ii) COCP monotherapy compared with metformin and COCP; and (iii) metformin monotherapy compared with metformin and COCP in women with PCOS for improving anthropometric, biochemical, clinical, and psychological outcomes.

The Population, Intervention, Comparison and Outcome (PICO) framework for this systematic review is outlined in Table 1 and was determined by an experienced, multidisciplinary, clinical research team (A.M., T.P., D.R., P.M.S., C.T.T., A.P., S.W., H.T.). Core outcomes were based on a Delphi process involving 700 clinicians, academic opinion leaders, and consumers (19). Outcomes included anthropometric, metabolic, androgenicity, and psychological outcomes, as well as adverse events (detailed in Table 1).

Population, intervention, comparison and outcome (PICO) of the systematic review and meta-analysis

Abbreviations: BMI, body mass index; COCP, combined oral contraceptive pills; CPA, cyproterone acetate; CRP, C-reactive protein; DHEAS, dehydroepiandrosterone sulfate; FAI, free androgen index; FG score, Ferriman Gallwey score; HDL, high-density lipoprotein; HOMA-IR, homeostatic model assessment for insulin resistance; LDL, low-density lipoprotein; OGTT, oral glucose tolerance test; PAI-1, plasminogen activator inhibitor; PCOS, polycystic ovary syndrome; QoL, quality of life; WHR, waist to hip ratio; SHBG, sex hormone binding globulin.

This systematic review and meta-analysis provide an update of a previous systematic review (12) conducted in 2017 to inform the International Evidence-based Guidelines for the Assessment and Management of PCOS (20). This review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (21, 22). The study protocol was registered prior to full-text screening in PROSPERO (Registration number CRD42022345640).

Ovid Medline, Embase, PsycINFO, All EBM, and CINAHL were searched to identify relevant literature from 2017 until July 7, 2022. We also re-evaluated all articles, which were found in the search performed during the previous systematic review ranging from 1946 to 2017, as well as additional references identified from relevant systematic reviews. Details of the search strategy are presented in Supplementary Fig. S1 (23).

Two authors (J.M., M.F., or S.A.) independently screened each potential study on title and abstract with the use of COVIDENCE. The same authors performed the full-text screening to determine the final included studies. If there was any doubt about inclusion, the study was reviewed and discussed in a larger group (J.M., M.F., S.A., and C.T.T.).

Quality appraisal of the included studies, in terms of risk of bias (RoB), was performed by J.M. and M.F. using RoB2 (24). Individual quality items included methods of randomization and group allocation; blinding of patients, carers, investigators, or outcome assessors; methods of outcome assessment and reporting; methods of data analysis; statistical issues; and trial pre-registration. Disagreements were resolved by discussion and re-inspection of the full-text article. Each study was adjudged as having an overall low RoB, some concerns, or high RoB. We used the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) method to estimate the certainty of the evidence (25). GRADE assessments were conducted in duplicate by J.M. and M.F.

Two independent reviewers (J.M. and M.F.) extracted data from included studies, using a specially developed data extraction form according to selection criteria. Extracted data included a description of the study (authors, country, year of publication, setting, diagnostic criteria for PCOS), participants (mean age and BMI), intervention (dose and duration of metformin and COCP) and study results according to outcome. All units were recalculated to SI units and SD.

Meta-analyses were performed using Review Manager 5.4.1. For continuous outcomes we calculated mean differences (MD) and 95% CI and for dichotomous outcomes odds ratios (ORs) and 95% CI. Due to clinical heterogeneity from differences in metformin dose, type of COCP, and duration of treatment, random effects models were used for the meta-analyses.

We performed subgroup analyses according to BMI, age (adults and adolescents), and type of COCP. Regarding BMI, studies were sorted into 3 categories; those with participants with a BMI < 25 kg/m2, those with a BMI ≥ 25 kg/m2, and studies using other BMI cutoffs (BMI not classified). Participants aged 10 to 19 years were classified as adolescents. Subgroup analyses according to type of COCP were classified into COCP containing CPA and other COCP.

Funnel plots were inspected for the assessment of publication bias.

Results

From 1660 search results, 450 articles were chosen for full-text review. In addition, we screened 176 full texts from the previous search in 2017 (12). After full-text review, 36 randomized controlled trials (RCTs) in 46 publications were identified (Fig. 1). Of these, 25 RCTs compared metformin with COCP, 17 compared COCP with metformin and COCP, and 6 compared metformin with metformin and COCP. Study characteristics are presented in Table 2. The RCTs were performed in Europe (n = 9) (26-40) North America (n = 4) (41-44), the Middle East (n = 14) (45-58), Asia (n = 8) (59-66), and Australia (n = 1) (67-70) between 2000 and 2021. Intervention durations ranged from 3 to 24 months, with 6 months of follow-up being the most commonly reported data. Metformin doses ranged from 500 mg to 2000 mg daily, with 1500 to 2000 mg daily being the most common dose (in 28 of the 36 studies). Regarding COCP, all studies used 30 to 35 µg ethinylestradiol (EE), 20 studies used CPA as the progestin compound, 14 studies used other progestins, and 2 studies had one study arm with participants using EE/drospirenone and another arm with those using EE/CPA (26, 40). Four studies involved adolescents (41, 42, 44, 49) and the remainder focused on adults.

Included studies are shown in the PRISMA flowchart. The search was performed for 3 medical treatments (combined oral contraceptive pills [COCP], metformin, and anti-androgens) as part of the update of the PCOS guidelines. Results comparing COCP, metformin, and combined COCP and metformin were included in this systematic review.

Abbreviations: AES, Androgen Excess Society; BMI, body mass index; C, Oral contraceptive pill; CPA, cyproterone acetate; DRSP, drospirenone; DSG, desogestrel; EE, ethinylestradiol; LVG, levonorgestrel; M, metformin; NIH, National Institutes of Health; NOR, norgestimate; NR, Not reported, PCOS, polycystic ovary syndrome; RoB, risk of bias; Rott, Rotterdam; SPL, spironolactone.

In our systematic review, 2 studies fulfilled the criteria for low RoB, 14 had some concerns, and 20 had a high RoB. All studies adequately described the PCOS diagnostic criteria and inclusion and exclusion criteria. The most common reason for a high RoB was bias arising from the randomization process. The second most common reason was missing outcome data. RoB assessments of the included studies are shown in Fig. 2.

Risk of bias assessments of the included studies.

Results on metformin compared with COCP are shown in Table 3. Subgroup analyses are shown in Supplementary Tables S1 and S2 (23), also illustrated in Supplementary Fig. S2 (subgroups according to BMI) (23), Supplementary Fig. S3 (subgroups according to adults/adolescents) (23), and Supplementary Fig. S4 (subgroups according to COCP with or without CPA) (23).

Grade assessments and evidence profile on outcomes comparing metformin with combined oral contraceptive pills

Bolding indicates a statistically significant MD (P < .05).

Abbreviations: COCP, combined oral contraceptive pills; MD, mean difference; RCT; randomized controlled trials; RoB; risk of bias.

a Downgraded once as all studies high or moderate RoB.

b Downgraded twice as all studies or all but one are high RoB.

c Downgraded once as studies high to moderate RoB.

d Downgraded once as I2 is close to or >50% but CI partly overlapping.

e Downgraded twice as I2 very high and CI not overlapping.

f Downgraded once as no adolescents in the overall group.

g Downgraded twice as there are very few studies.

Combination treatment (COCP and metformin) compared with COCP monotherapy is presented in Table 4, with subgroup analysis reported in Supplementary Table S3 and Supplementary Fig. S5 (23). Results on combination treatment vs metformin monotherapy are presented in Table 5, with subgroup analysis in Supplementary Table S4 and Supplementary Fig. S6 (23).

Grade assessments and evidence profile of outcomes in PCOS comparing COCP with metformin and COCP

Bolding indicates a statistically significant MD (P < .05).

Abbreviations: COCP, combined oral contraceptive pills; MD, mean difference; Met, metformin; PCOS, polycystic ovary syndrome; RCT; randomized controlled trials; RoB; risk of bias.

a Downgraded once as the majority of evidence is at moderate or high RoB.

d Downgraded twice due to I2 > 50% and CI not overlapping.

e Downgraded twice due to very few patients.

Grade assessments and evidence profile of outcomes in PCOS comparing metformin with metformin and combined oral contraceptive pills

Bolding indicates a statistically significant MD (P < .05).

Abbreviations: COCP, combined oral contraceptive pills; MD, mean difference; Met, metformin; PCOS, polycystic ovary syndrome; RCT; randomized controlled trials; RoB; risk of bias.

a Downgraded once as the studies are at moderate or high RoB.

b Downgraded once as I2 is close to or >50% but CI partly overlapping.

c Downgraded twice as I2 very high and CI not overlapping.

d Downgraded once as there are only a few studies.

e Downgraded twice as there are very few participants.

We found no evidence of publication bias in the funnel plots.

Regarding overall results on weight, BMI, and waist to hip ratio (WHR), no differences were found (Table 3).

In subgroup analyses according to BMI and adults/adolescents no differences were seen on weight, BMI and WHR (Supplementary Table S1; Supplementary Figs. S2 and S3) (23).

In subgroup analyses comparing metformin with COCP with and without CPA (Supplementary Table S2; Supplementary Fig. S4) (23) there were no differences in weight. Nevertheless, BMI was lower with metformin when compared with EE/CPA (MD −0.99 kg/m2; 95% CI, −1.74 to −0.23), but not when compared with COCP without CPA.

COCP alone compared with combination treatment with COCP and metformin showed no differences in weight, WHR, or BMI (Table 4).

Metformin was superior in lowering WHR (MD −0.03; 95% CI, −0.06 to −0.01) compared with metformin, with no difference in BMI (Table 5).

No difference was seen in hirsutism in the overall analysis but regarding biochemical hyperandrogenism, metformin was inferior to COCP on free androgen index (FAI) (MD 7.08; 95% CI, 4.81 to 9.36), sex hormone binding globulin (SHBG) (MD −118.61 nmol/L; 95% CI, −174.46 to −62.75), and testosterone (MD 0.48 nmol/L; 95% CI, 0.32 to 0.64) (Table 3).

Subgroup analysis categorized by BMI (Supplementary Table S1; Supplementary Fig. S2) (23) showed that in women with PCOS and a BMI < 25 kg/m2, metformin was less effective than COCP in treating hirsutism (MD 1.73; 95% CI, 0.07 to 3.40) and improving FAI (MD 5.78; 95% CI, 2.82 to 8.73), total testosterone (MD 0.56 nmol/L; 95% CI, 0.29 to 0.83), and SHBG (MD −168 nmol/L; 95% CI, −211 to −124).

In women with a BMI ≥25 kg/m2, there was no difference in hirsutism when comparing metformin with COCP. For biochemical hyperandrogenism, metformin was inferior for FAI (MD 9.05; 95% CI, 6.44 to 11.66); total testosterone (MD 0.40 nmol/L; 95% CI, 0.15 to 0.66), and SHBG (MD −96 nmol/L; 95% CI, −121 to −72).

The type of COCP did not influence the overall results regarding clinical and biochemical hyperandrogenism (Supplementary Table S2; Supplementary Fig. S4) (23).

No differences were noted in hirsutism between COCP alone and combination treatment. COCP alone was inferior to combination treatment for FAI (MD 0.58; 95% CI, 0.36 to 0.80), SHBG (MD −16.61 nmol/L; 95% CI, −28.51 to −4.71), dehydroepiandrosterone sulfate (DHEAS) (MD 0.93 μmol/L; 95% CI, 0.54 to 1.31), and androstenedione (MD 1.37 nmol/L; 95% CI, 0.14 to 2.60), whereas no differences were noted regarding total and free testosterone (Table 4; Supplementary Fig. S5 (23)).

Metformin was inferior for DHEAS (MD 82.38 μmol/L; 95% CI, 15.43 to 149.36) and total testosterone (MD 0.64 nmol/L; 95% CI, 0.26 to 1.02). None of the identified studies reported hirsutism as an outcome for this comparison (Table 5; Supplementary Fig. S6 (23)).

Metformin was more effective in lowering overall fasting insulin levels (MD −27.12 pmol/L; 95% CI, −40.65 to −13.59), total cholesterol (MD −0.40 mmol/L; 95% CI, −0.66 to −0.14), triglycerides (MD −0.15 mmol/L; 95% CI, −0.29 to −0.01), and C-reactive protein (CRP) levels (MD −11.31 nmol/L; 95% CI, −19.78 to −2.85) compared with COCP (Table 3). For fasting glucose, homeostatic model assessment for insulin resistance (HOMA-IR), high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol and plasminogen activator inhibitor 1 (PAI-1), there were no differences.

In subgroup analyses on women with a BMI ≥ 25 kg/m2 (Supplementary Fig. S2; Supplementary Table S1) (23), the metformin-treated group had lower CRP (MD −33.09 nmol/L; 95% CI, −47.33 to −18.84) and HDL (MD −0.24 mmol/L; 95% CI, −0.38 to −0.09) compared with COCP. Fasting insulin was lower with metformin regardless of BMI (Supplementary Table S1; Supplementary Fig. S2) (23).

Metformin was superior to COCP for fasting insulin, both in adults (MD −22.17 pmol/L; 95% CI, −29.93 to −14.42) and adolescents (MD −30.03 pmol/L; 95% CI, −78.63 to −18.56). Metformin was also superior for total cholesterol, both in adults (MD −0.34 mmol/L; 95% CI, −0.60 to −0.08) and for adolescents (MD −1.12 mmol/L; 95% CI, −1.74 to −0.50) (Supplementary Table S1) (23). The type of COCP showed no major differences compared with the overall result (Supplementary Table S2) (23).

COCP alone increased fasting insulin (MD 17.03 pmol/L; 95% CI, 7.79 to 26.26), HOMA-IR (MD 0.44; 95% CI, 0.17 to 0.70), and CRP (MD 1.94 nmol/L; 95% CI, 0.05 to 3.84) compared with combination treatment (Table 4; Supplementary Fig. S5 (23)). In women with a BMI < 25 kg/m2, COCP alone resulted in marginally higher glucose (MD 0.25 mmol/L; 95% CI, 0.07 to 0.43) compared with combination treatment.

Metformin alone showed marginally lower fasting glucose (MD −0.33 mmol/L; 95% CI, −0.64 to −0.01) and CRP (MD −4.08 nmol/L; 95% CI, −6.01 to −2.16), compared with a combination treatment (Table 5; Supplementary Fig. S6 (23)).

Metformin was inferior to COCP on restoring regular menses overall (OR 0.17; 95% CI, 0.05 to 0.57), as well as in adults (OR 0.19; 95% CI, 0.05 to 0.72) and adolescents (OR 0.10; 95% CI, 0.01 to 1.92).

We were not able to perform a meta-analysis on health-related quality of life. Results from 3 identified studies (32, 42, 50) showed conflicting results and our overall assessment is that there is no difference in quality of life.

Due to lack of systematic reporting, where many studies do not report adverse effects at all or do not report in a similar manner, we were not able to perform a meta-analysis. However, the reports suggest more gastrointestinal side effects with metformin.

Discussion

This extensive systematic review and meta-analysis, including 36 RCTs, was performed to directly inform recommendations on the use of metformin and COCP in women with PCOS, as part of the 2023 update of the International Evidence-based Guidelines on the Assessment and Treatment of PCOS. Our findings showed that metformin was superior to COCP for metabolic outcomes, especially in women with PCOS and a BMI ≥ 25 kg/m2, whereas COCP was superior for improving cycle regularity and, in women with a BMI < 25 kg/m2, for improving hirsutism. The combination of metformin and COCP was more effective for improving biochemical hyperandrogenism, insulin levels, and insulin resistance than COCP monotherapy.

For many patients with PCOS, medical treatment is indicated to treat clinical hyperandrogenism, including acne and hirsutism. For these symptoms, COCPs have been recommended as first-line treatment. In addition, COCPs regulate menstrual cycles and provide contraception (12, 15). In this study we confirm that COCP was superior to metformin for treatment of hirsutism in women with BMI < 25 kg/m2, whereas evidence for other BMI groups was of very low quality. However, we found that COCP was superior in improving biochemical hyperandrogenism compared with metformin. A systematic review on COCP treatment in women with PCOS, including both RCTs and non-RCTs, suggested that COCP containing CPA might be more effective in improving hirsutism. However, no direct comparisons between different COCP were made, limiting the conclusions (15). General population studies have shown that CPA increases the risk of venous thromboembolism compared with other COCP (71, 72); hence, COCPs containing CPA are currently not recommended as first-line treatments for PCOS (20).

No differences were found regarding hirsutism between COCP and combination treatment. Nevertheless, combination treatment was more effective in improving FAI, SHBG, and DHEAS compared with COCP alone. One systematic review (17) studied hirsutism and acne in PCOS, comparing monotherapy with metformin or COCP with combination treatment, with monotherapy being less effective for hirsutism compared with combination treatment. However, that review also included studies with shorter treatment durations (3 months) hindering interpretation of their findings.

As high insulin levels increase luteinizing hormone–mediated ovarian androgen synthesis (73), combination treatment presumably targets several distinct mechanisms leading to improved clinical and biochemical hyperandrogenism. Thus, combination treatment theoretically offers several benefits. Importantly, additional high-quality prospective studies are needed to better ascertain the efficacy of these treatment regimens. We also recognize the importance of self-assessment of severity and impact on quality of life, over and above clinical assessment.

Women with PCOS have an increased prevalence of the metabolic syndrome (74, 75). In our meta-analysis, there were no differences in weight, WHR, and BMI, but evidence was only available with very low certainty. Previously, metformin has been shown to improve BMI compared with placebo (76). A recent systematic review which included nonrandomized trials found that combined metformin, COCP, and anti-androgen treatment improved BMI and glucose tolerance (11). However, this study did not compare metformin monotherapy to a combination treatment with metformin and COCP. In our study the combination of COCP and metformin showed no benefit for anthropometric measures compared with COCP alone.

Hyperinsulinemia and insulin resistance play important roles in the pathophysiology of PCOS for both normal-weight women and women with obesity (2, 3, 73, 77). Metformin decreases insulin resistance and insulin levels. COCP does not have any major effects on carbohydrate metabolism in healthy women, whereas the effect on lipid metabolism depends on the level of estrogen and type of progestin, with potential negative effects (78, 79, 80, 81). Our systematic review confirms that metformin is superior in lowering fasting insulin levels, total cholesterol, and triglycerides compared with COCP, in line with findings reported previously (12).

Metformin can be used where COCPs are contradicted, in older women, women with obesity or medical conditions (such as migraine with aura, risk of venous thromboembolism or severe hypertension) or where pregnancy is desired. The major disadvantage of metformin treatment is the generally mild, usually self-limited, gastrointestinal side effects, limiting patient acceptance of appropriate doses. We also report that metformin lowered CRP compared with COCP treatment. CRP is often used as a metabolic risk marker and is associated with an increased risk of cardiovascular disease (82). Metformin has previously been shown to decrease CRP, both in obese and nonobese women, compared with placebo (83). COCP appears to increase CRP (79, 84, 85), yet the significance of these effects remains unknown.

Importantly, the addition of metformin to COCP improved insulin levels and insulin resistance, as compared with COCP alone. This is of special interest for women with PCOS with additional risk factors for type 2 diabetes, such as a high BMI (86, 87). Two systematic reviews found that adding metformin to COCP and anti-androgens improved BMI and glucose tolerance (11, 88).

This review presents the most up-to-date evidence on COCP and metformin treatment in women with PCOS. Strengths of the report include the rigorous processes, including PICO developed by clinicians, researchers, and patients; and the large number of RCTs included. Because critical appraisal is inherently subjective, both RoB and GRADE assessments were conducted by 2 authors independently. Several subgroup analyses were included to highlight benefits in population subgroups of interest. Limitations include those inherent in the included studies, with many studies having a high RoB, mainly due to often poorly described randomization processes or lack of blinding. Certainty of evidence was also affected by inconsistency of effect sizes and small sample sizes. Several studies did not provide data on BMI.

Identified research gaps include the lack of larger studies on adolescents with PCOS and on the comparison of metformin with combination treatment. Additional high-quality studies are needed to assess symptoms, especially related to clinical hyperandrogenism rather than biochemical markers of hyperandrogenism. Future directions include understanding mechanisms of action and factors that may impact on medication responses, such as genotype and phenotype of PCOS.

Conclusion

Results from this extensive systematic review and meta-analysis will advise the pending 2023 PCOS guideline update. The guideline will recommend COCP to be used over metformin for management of irregular cycles and hirsutism, and metformin over COCP for metabolic indications in PCOS. While our meta-analyses indicated that the combined treatment with metformin and COCP improved biochemical hyperandrogenism, insulin levels, and insulin resistance more than COCP alone, no difference was seen in clinical outcomes.

For women with PCOS, the choice of treatment should be based on clinical symptoms. Combination treatment appears to be beneficial in high metabolic risk groups, targeting the 2 major endocrine disturbances seen: hyperinsulinemia and hyperandrogenism. Our results, including several sub-analyses, add to the current evidence base and contribute toward reaching the ultimate goal of shared decision-making and effective, tailored, and individualized treatment for patients with PCOS.