Treatment Options for Hirsutism: A Systematic Review and Network Meta-Analysis

Abstract

Background

Several pharmacologic treatments for hirsutism are used in practice; however, their relative efficacy is unclear.

Methods

We searched MEDLINE, EMBASE, and CENTRAL through January 2017 for randomized controlled trials (RCTs) with follow-up of at least 6 months that evaluated antiandrogens, insulin sensitizers, and oral contraceptives in women with hirsutism. Independent pairs of reviewers selected and appraised trials. Random-effects network meta-analysis was used to compare individual drugs and classes.

Results

We included 43 trials. Estrogen-progestin oral contraceptives pills (OCPs), antiandrogens, and insulin sensitizers were superior to placebo, with standardized mean reductions (95% confidence intervals) of −0.94 (−1.49 to −0.38), −1.29 (−1.80 to −0.79), and −0.62 (−1.00 to −0.23), respectively. Antiandrogen monotherapy, the combination of OCP and antiandrogen, the combination of OCPs and insulin sensitizer, and the combination of antiandrogen and insulin sensitizer were superior to insulin sensitizer monotherapy. The combination of OCPs and antiandrogen was superior to OCPs. Antiandrogen monotherapy with flutamide, finasteride, and spironolactone were each superior to placebo but similar to each other in efficacy. OCPs containing levonorgestrel, cyproterone acetate, or drospirenone were similar in effectiveness to other OCPs or had trivial differences. The certainty in comparisons with placebo was moderate and for head-to-head comparisons was low.

Conclusions

Estrogen-progestin OCPs, antiandrogens, and insulin sensitizers are superior to placebo for the treatment of hirsutism.

Hirsutism is defined as the presence of excessive terminal (coarse) hair that appears in a male pattern in women and is often associated with an underlying endocrine disorder (most commonly polycystic ovary syndrome), psychosocial difficulties, and reduced quality of life (1, 2). The prevalence of hirsutism is ~5% to 10% (3, 4). The many approaches to the treatment of hirsutism include nonpharmacologic (lifestyle changes; cosmetic measures; and direct hair removal methods, such as photoepilation and electrolysis) and pharmacologic therapies (5–8). The current mainstay of pharmacologic therapies includes combination estrogen-progestin oral contraceptives pills (OCPs) and antiandrogens. Metformin was used in the past but has largely been abandoned due to perceived lack of efficacy. Glucocorticoids are occasionally used for women with adrenal disorders such as nonclassic 21-hydroxylase deficiency, but they are second-line therapy. Thus, the current focus is on understanding the relative effectiveness of OCPs and antiandrogens as a class and as individual agents (9). However, the evidence for the relative effectiveness of these therapies is constrained by multiple methodological limitations, including lack of blinded evaluation of effects of treatment, short duration of clinical trials, and small sample size (10, 11). Most important, few studies have addressed comparative effectiveness of these treatment options (10, 11).

In 2008, the Endocrine Society published its guidelines on the management of hirsutism that were based on two systematic reviews of OCPs, antiandrogens, and insulin sensitizers (10–12). To update these guidelines, the Endocrine Society commissioned this systematic review to summarize the totality of evidence regarding the efficacy of antiandrogens, insulin sensitizers, and oral contraceptives for the treatment of hirsutism and attempt a network meta-analysis approach to provide head-to-head efficacy estimates.

Methods

This protocol-driven systematic review is reported in adherence with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (13). The commissioning task force provided the clinical expert feedback.

Eligibility Criteria

We included randomized controlled trials (RCTs) that allocated women with hirsutism to antiandrogens, insulin sensitizers, OCPs (alone or a combination of the aforementioned classes), or placebo. No language or country restrictions were used. The minimum follow-up period was 6 months. Hirsutism was assessed by clinicians using the modified Ferriman-Gallwey (FG) score, the gold standard for evaluating hirsutism. A total score >8 is considered abnormal in black or white women, but the cutoff varies widely by ethnicity (12, 14). Hirsutism was also measured by patient self-assessment or laboratory assessment of hair (e.g., diameter, length, or rate of growth). RCTs were excluded if women received a gonadotropin-releasing hormone agonist, clomiphene, or glucocorticoids. RCTs were excluded if they enrolled women with hirsutism secondary to causes other than idiopathic hirsutism, polycystic ovary syndrome, or presumed nonclassic congenital adrenal hyperplasia. For the purposes of analysis, we considered a dose of cyproterone acetate (CPA) >2 mg as an antiandrogen. We also planned a priori to compare OCPs containing the antiandrogenic progestins CPA and drospirenone (DRSP) or the highly androgenic progestin levonorgestrel vs other OCPs.

Study identification

The database search strategy (Supplemental Table 1) was developed by a medical reference librarian with input from the study investigators. We performed a comprehensive search of MEDLINE, EMBASE, and CENTRAL through January 2017, updating two prior systematic reviews (10, 11) that used a search date through May 2006. We also reviewed the reference section of the included studies and of relevant systematic reviews and consulted with experts to identify additional candidate studies.

Study selection

Independent reviewers, working in duplicate, screened all abstracts and titles. Studies that met the inclusion criteria were obtained in full text. Reviewers working independently and in duplicate identified the eligibility of full text reports. Any disagreements were resolved by consensus.

Data extraction

Independent reviewers, working in pairs, extracted baseline characteristic data describing the patient population and treatments studied. The outcome of interest was severity of hirsutism, including self-assessed, when patients reported on their hirsutism; clinician-assessed (e.g., modified FG score); or assessed through laboratory measurements (e.g., hair diameter measurement).

We collected as outcomes the change of the scores from baseline or, when these were not available, the outcomes reported at the end of the study. We contacted the authors when data were not available.

Quality assessment (risk of bias)

We assessed the methodological quality of eligible RCTs using the Cochrane risk of bias tool (15). The adequacy of blinding, allocation concealment, extent of loss-to-follow up (i.e., the proportion of study patients for whom the outcomes could not be ascertained), and source of study funding were evaluated.

Statistical analysis

For head-to-head comparisons, we estimated the overall effect size by pooling standardized mean difference (SMD) using the DerSimonian and Laird random-effects methods, with the estimate of heterogeneity being taken from the Mantel-Haenszel model (16). SMD was chosen because hirsutism was evaluated by using different scales across trials. To evaluate effect sizes for all possible pairwise comparisons, whether they have been compared in previous trials, we calculated SMD for each included study and conducted network meta-analyses to combine head-to-head and indirect evidence, using multivariate random effects meta-regression (17). A node-splitting method was used to evaluate the consistency of network meta-analysis.

To assess heterogeneity across the studies, we used the I² statistic, where I² > 50% or P < 0.10 suggested a high level of heterogeneity. Although we planned to assess publication bias by using visual inspection of funnel plots and the Egger regression asymmetry test (18, 19), we were not able to conduct formal test because of the small number of studies available from head-to-head comparisons (20). Statistical analyses were conducted by using Stata software, version 14.1 (Stata Corp., College Station, TX).

Results

Search results

We included 43 RCTs in this systematic review. Eight of these trials compared specific OCPs containing antiandrogenic progestins (CPA and DRSP) or the highly androgenic progestin levonorgestrel vs other OCPs. Three trials provided dichotomized hirsutism results (no severity score). Hence, 32 trials provided data for network meta-analysis. Figure 1 describes the process of study selection.

Figure 1.

Study selection process.

Characteristics of the included studies

Supplemental Table 2 summarizes the characteristics of the included trials used in the network analysis. In general, the median age of participants was 25 years. Most studies used the modified FG score to measure severity of hirsutism. Baseline FG scores varied widely among the study patients, ranging from 5 to 23.

Six drug classes or a combination of classes were used, including OCPs, antiandrogen, insulin sensitizer, OCPs plus antiandrogen, OCPs plus insulin sensitizer, and antiandrogen plus insulin sensitizer. Specifically, a total of 14 agents or agent combinations were identified. Figures 2 and 3 depict the network of available trials and interventions per class and drug, respectively.

Figure 2.

Trial network by drug class. The diagram shows the available direct comparisons. Each circle represents a drug class. Each line represents a comparison. The number associated with the line and the thickness of the line reflect the number of cohorts included in each comparison (a single trial can include multiple cohorts).

Figure 3.

Trial network by specific drug or combination of drugs. The diagram shows the available direct comparisons. Each circle represents a specific drug or combination of drugs. Each line represents a comparison. The number associated with the line and the thickness of the line reflect the number of cohorts included in each comparison (a single trial can include multiple cohorts).

Methodological quality (risk of bias)

Risk of bias indicators are summarized in Supplemental Table 3. Most trials did not use or report allocation concealment (86%) or blind outcome assessors (69%). A third (34%) were solely supported by not-for-profit agencies. In summary, the risk of bias of the included trials was high because of inadequate allocation concealment and blinding and potential conflict of interest.

Network meta-analysis

Thirty-two RCTs were included in network meta-analysis. Table 1 summarizes pairwise effect size and 95% confidence intervals (CIs) of severity of hirsutism for class comparisons derived from network meta-analysis. Compared with placebo, OCPs, antiandrogen, insulin sensitizers, a combination of OCPs and antiandrogen, a combination of OCPs and insulin sensitizers, and a combination of antiandrogen and insulin sensitizers were associated with statistically significant reduction in hirsutism severity (P < 0.05). We also found that antiandrogen, a combination of antiandrogen and insulin sensitizers, a combination of OCPs and antiandrogen, and a combination of OCPs and insulin sensitizers were superior to insulin sensitizers. The combination of OCPs and antiandrogen was superior to OCPs alone. We did not observe a significant difference of one class over another in the other comparisons.

Table 1.

Class Comparison

Variable	Placebo	OCP	OCP + Antiandrogen	OCP + Insulin Sensitizers	Antiandrogen	Antiandrogen + Insulin Sensitizers
OCP	−0.94 (−1.49 to −0.38)a
OCP + antiandrogen	−1.40 (−1.95 to −0.84)a	−0.46 (−0.91 to −0.01)a
OCP + insulin sensitizers	−1.91 (−3.02 to −0.80)a	−0.97 (−2.03 to 0.09)	−0.51 (−1.63 to 0.60)
Antiandrogen	−1.29 (−1.80 to −0.79)a	−0.35 (−0.86 to 0.15)	0.10 (−0.33 to 0.54)	0.62 (−0.50 to 1.73)
Antiandrogen + insulin sensitizers	−1.53 (−2.29 to −0.77)a	−0.59 (−1.39 to 0.20)	−0.13 (−0.91 to 0.64)	0.38 (−0.88 to 1.64)	−0.24 (−0.92 to 0.44)
Insulin sensitizers	−0.62 (−1.00 to −0.23)a	0.32 (−0.15 to 0.80)	0.78 (0.25–1.31)a	1.29 (0.24–2.35)a	0.67 (0.19–1.16)a	0.91 (0.18–1.65)a

The comparison in this table is row heading compared with column heading. Negative effect size means improvement in hirsutism scores. Effect size is standardized.

^a

Statistically significant.

Considering individual drugs (Table 2), significant reduction in hirsutism compared with placebo was noted with OCPs, finasteride, flutamide, spironolactone, a combination of OCPs and CPA, a combination OCPs and finasteride, a combination of OCPs and metformin, a combination of OCPs and spironolactone, a combination of flutamide and metformin, and a combination of spironolactone and metformin. Metformin was associated with statistically significant reduction compared with placebo; however, that effect is supported by very-low-quality evidence due to network inconsistency and heterogeneity of the direct comparison. We did not find a significant difference between placebo and rosiglitazone, troglitazone, or a combination of OCPs and flutamide. A significant reduction was also noted with flutamide, a combination of OCPs and metformin, and a combination of OCPs and spironolactone, when compared with metformin.

Table 2.

Individual Medication Comparison

Variable	Placebo	OCP	OCP + CPA	OCP + Finasteride	OCP + Flutamide	OCP + Metformin	OCP + Spironolactone	Finasteride	Flutamide	Flutamide + Metformin	Metformin	Rosiglitazone	Spironolactone	Spironolactone + Metformin
OCP	−1.21 (−1.92 to −0.50)a
OCP + CPA	−1.70 (−2.55 to −0.85)a	−0.49 (−1.41 to 0.43)
OCP + finasteride	−1.64 (−2.72 to −0.55)a	−0.43 (−1.25 to 0.39)	0.06 (−1.17 to 1.29)
OCP + flutamide	−0.70 (−1.95 to 0.56)	0.51 (−0.93 to 1.95)	1.00 (−0.52 to 2.52)	0.94 (−0.72 to 2.60)
OCP + metformin	−2.09 (−3.33 to −0.85)a	−0.89 (−2.06 to 0.28)	−0.40 (−1.79 to 1.00)	−0.46 (−1.88 to 0.97)	−1.40 (−3.16 to 0.37)
OCP + spironolactone	−1.90 (−2.88 to −0.93)a	−0.70 (−1.49 to 0.10)	−0.21 (−1.31 to 0.90)	−0.27 (−1.41 to 0.87)	−1.21 (−2.80 to 0.38)	0.19 (−1.19 to 1.58)
Finasteride	−1.48 (−2.18 to −0.78)a	−0.28 (−1.05 to 0.50)	0.21 (−0.48 to 0.90)	0.15 (−0.97 to 1.28)	−0.79 (−2.22 to 0.65)	0.61 (−0.71 to 1.93)	0.42 (−0.58 to 1.42)
Flutamide	−1.85 (−2.55 to −1.14)a	−0.64 (−1.42 to 0.15)	−0.15 (−0.91 to 0.61)	−0.21 (−1.34 to 0.92)	−1.15 (−2.59 to 0.29)	0.25 (−1.06 to 1.56)	0.06 (−0.88 to 1.00)	−0.36 (−1.03 to 0.31)
Flutamide + metformin	−1.68 (−2.66 to −0.71)a	−0.48 (−1.55 to 0.59)	0.02 (−1.12 to 1.15)	−0.05 (−1.39 to 1.30)	−0.98 (−2.57 to 0.60)	0.41 (−1.07 to 1.89)	0.22 (−1.01 to 1.45)	−0.20 (−1.25 to 0.86)	0.16 (−0.79 to 1.11)
Metformin	−0.73 (−1.18 to −0.27)a†	0.48 (−0.13 to 1.09)	0.97 (0.14 to 1.80)a	0.91 (−0.11 to 1.93)	−0.03 (−1.37 to 1.31)	1.37 (0.20 to 2.53)a	1.17 (0.26 to 2.09)a	0.76 (0.05 to 1.46)a	1.12 (0.44 to 1.80)a	0.95 (0.01 to 1.89)a
Rosiglitazone	−1.06 (−2.64 to 0.51)	0.14 (−1.26 to 1.55)	0.64 (−1.04 to 2.31)	0.57 (−1.05 to 2.20)	−0.37 (−2.38 to 1.65)	1.03 (−0.79 to 2.86)	0.84 (−0.77 to 2.45)	0.42 (−1.18 to 2.02)	0.78 (−0.82 to 2.39)	0.62 (−1.14 to 2.38)	−0.33 (−1.86 to 1.19)
Spironolactone	−1.41 (−2.26 to −0.57)a	−0.21 (−1.14 to 0.73)	0.29 (−0.51 to 1.08)	0.22 (−1.02 to 1.47)	−0.72 (−2.23 to 0.80)	0.68 (−0.72 to 2.08)	0.49 (−0.64 to 1.62)	0.07 (−0.76 to 0.90)	0.43 (−0.41 to 1.28)	0.27 (−0.89 to 1.43)	−0.68 (−1.50 to 0.13)	−0.35 (−2.04 to 1.34)
Spironolactone + metformin	−1.92 (−3.42 to −0.41)a	−0.71 (−2.27 to 0.85)	−0.22 (−1.69 to 1.26)	−0.28 (−2.04 to 1.48)	−1.22 (−3.18 to 0.74)	0.18 (−1.69 to 2.05)	−0.01 (−1.69 to 1.67)	−0.43 (−1.93 to 1.06)	−0.07 (−1.57 to 1.43)	−0.23 (−1.93 to 1.46)	−1.19 (−2.67 to 0.30)	−0.85 (−2.95 to 1.24)	−0.50 (−1.75 to 0.74)
Troglitazone	−0.65 (−1.88 to 0.58)	0.56 (−0.86 to 1.98)	1.05 (−0.45 to 2.54)	0.99 (−0.65 to 2.62)	0.05 (−1.71 to 1.80)	1.44 (−0.30 to 3.19)	1.25 (−0.32 to 2.82)	0.83 (−0.58 to 2.25)	1.19 (−0.22 to 2.61)	1.03 (−0.54 to 2.60)	0.08 (−1.23 to 1.39)	0.41 (−1.58 to 2.41)	0.76 (−0.73 to 2.25)	1.27 (−0.68 to 3.21)

The comparison in this table is row heading compared with column heading. Negative effect size means improvement in hirsutism scores. Effect size is standardized.

^a

Statistically significant.

The network was overall consistent for both analyses (by class and by drug). One exception where direct and indirect estimates differed was for comparison of OCPs vs placebo (both direct and indirect estimates showed benefit with OCPs; however, larger benefit was seen with direct estimates). Some inconsistency was also noted for comparisons of OCPs vs metformin and metformin vs placebo. Supplemental Tables 4 and 5 summarize the results of direct comparisons by class and individual drugs; respectively.

OCP types

The trials that compared OCPs containing levonorgestrel, CPA, or DRSP are summarized in Supplemental Table 6. The quality of RCTs is summarized in Supplemental Table 7. Meta-analysis results are depicted in Supplemental Figs 1–3. OCPs containing levonorgestrel had a similar effect on hirsutism scores compared with all other OCPs (0.49; 95% CI, −0.22 to 1.20). OCPs containing antiandrogenic progestins (two trials using CPA and three trials using DRSP) were associated with similar effect (0.20; 95% CI, −0.42 to 0.81) for CPA and a small reduction in FG scores (−0.49; 95% CI, −0.96 to −0.03) for DRSP. This change may not be clinically relevant.

Supplemental Table 8 lists excluded trials.

Discussion

Main findings

We conducted a systematic review and network meta-analysis comparing commonly used treatments for hirsutism in adult women. Analysis has demonstrated that OCPs, antiandrogens, and insulin sensitizers were each effective means of treating hirsutism. Insulin sensitizer monotherapy appeared to be the least effective.

OCPs containing levonorgestrel, CPA, or DRSP were similar in effectiveness to other OCPs or had trivial differences. As individual agents, finasteride, flutamide, and spironolactone were more effective than placebo. Metformin was associated with a modest improvement (although supported by very-low-quality evidence due to risk of bias, heterogeneity, and network inconsistency), whereas rosiglitazone and troglitazone were not.

Practical implications

Women with mild hirsutism and no evidence of an endocrine disorder may not necessarily require treatment. However, when hirsutism is sufficiently important to women and they seek treatment, several pharmacological therapies are available. OCPs appear to be the first-line therapy, particularly when contraception is desired. Antiandrogens are effective, but because of their teratogenic potential, they may be more suited for women who are not sexually active, have undergone permanent sterilization, or use long-acting reversible contraception. Combination therapies can be used when women do not respond to a particular agent. Evidence is insufficient to recommend one OCP over another. Women at high risk for venous thromboembolism because of age, obesity, or smoking should consider a non-OCP therapy or use OCPs with the lowest effective dose of ethinyl estradiol and a low-risk progestin. When additional cosmetic benefit is desired, direct hair removal methods (e.g., laser therapy) are additional options.

Strengths and weaknesses

The strengths of this review relate to the comprehensive literature search and bias protection measures, such as selecting studies in duplicate and following an a priori–established protocol. The published systematic reviews and meta-analyses compared the effectiveness of hirsutism treatments for one specific class of drugs (10, 11, 21–24). The network meta-analysis approach allows combining direct and indirect effects to increase precision based on the assumption of consistency among the different trials. Nevertheless, the precision of the results remains limited because of the small number of small trials. Many of the RCTs included in this analysis had important methodological limitations. The primary outcome is measured with the modified FG scale, which does not have an established minimally important difference (25). Although we did not detect publication bias, it remains likely because available approaches to evaluate it are unreliable. The confidence in the pooled estimates and subsequent inferences are rated by using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach (26). Overall, at least moderate-quality evidence supports the superiority of OCPs, antiandrogens, and insulin sensitizers over no treatment. The quality of evidence supporting other comparisons (particularly head-to-head comparisons) is low.

Conclusions

Estrogen-progestin OCPs, antiandrogens, and insulin sensitizers are superior to placebo for the treatment of hirsutism.

Abbreviations:

 
• CPA

  cyproterone acetate

 
• DRSP

  drospirenone

 
• FG

  Ferriman-Gallwey

 
• OCP

  oral contraceptives pill

 
• RCT

  randomized controlled trial

 
• SMD

  standardized mean difference.

Acknowledgments

Financial Support: This study was partially funded by the Endocrine Society.

Disclosure Summary: The authors have nothing to disclose.

Author notes

These authors have contributed equally to the study.