Marshall and Dunaif Page 6

Second, ISDs are not a panacea, even for classic NIH PCOS (11). Meta-analyses of
metformin (12) indicate that it improves menstrual regularity and ovulation rates in PCOS.
Thiazolidinediones (TZDs) appear to have similar efficacy but there are fewer RCTs(12). A
NIH-PA Author Manuscript

large RCT (13) showed that clomiphene citrate was superior to metformin when the relevant
fertility endpoint, live births, was examined. Metformin is not recommended as a first-line
therapy for infertility in PCOS (14). Further, a recent RCT showed that metformin treatment
from the first trimester through pregnancy did not reduce pregnancy complications or alter
birthweight in PCOS (15). Metformin and other ISDs reduce hirsutism but appear to be
inferior to antiandrogens and contraceptive steroids (16). ISDs are also not recommended as
first-line therapy for hirsutism (17).

Meta-analyses (12, 18) of metabolic endpoints in women with PCOS indicate that
metformin has modest efficacy for reducing circulating glucose and insulin levels as well as
for reducing systolic blood pressure. However, the results of recent meta-analyses differ
with those limited to women with PCOS showing no benefit of metformin on lipid
parameters or body weight (12). The meta-analysis of Salpeter and colleagues (18) that
contained RCTs in women with PCOS as well as in other groups of subjects, found a
beneficial effect of metformin on body weight and lipid parameters. There are limited RCTs
of metabolic endpoints with TZDs. There is significant heterogeneity in many of the
endpoints examined in the meta-analyses, most likely because of differences in diagnostic
criteria and body weight as well as other potential differences, such as ethnic/racial
NIH-PA Author Manuscript

differences in the PCOS cohorts studied(12). Further, many of the RCTs limited to PCOS
examining metabolic endpoints have been small and/or metabolic parameters have not been
the primary endpoints of the trial(12, 18).

What can be extrapolated from large RCTs of ISDs for metabolic endpoints in other
cohorts? The efficacy of ISDs, both metformin and TZDs, for diabetes prevention has been
demonstrated in individuals with prediabetes(19), which, given the high risk of glucose
intolerance in women with PCOS(2), likely contained more than the 7% population
prevalence of affected women among the younger female participants. In the Diabetes
Prevention Program (DPP)(20), metformin both reduced incident metabolic syndrome and
reversed prevalent metabolic syndrome in individuals with prediabetes but this effect was
less pronounced in women. Metformin also resulted in modest weight loss that was greatest
at one year of therapy but was maintained at 10 years (21, 22). Hyperinsulinemia is an
independent risk factor for cardiovascular disease (CVD) (23). However, there is no
evidence to support treating insulin resistance per seto reduce CVD events. The PRO active
Study (24), a large RCT examining standard T2D therapy compared to standard therapy
with TZD in patients with T2D, who are at very high risk for CVD, found no significant
reduction in the primary endpoint of mortality from any cause, non-fatal myocardial
NIH-PA Author Manuscript

infarction, acute coronary syndrome, coronary revascularization, stroke, leg amputation or

Third, isolated insulin resistance cannot be reliably diagnosed with surrogate markers (25,
26). Fasting insulin levels reflect insulin secretion and clearance as well as insulin resistance
(26) and not sufficiently predictive of euglycemic clamp measures of insulin action to be
used for the diagnosis of insulin resistance in individuals patients (25). Other fasting
measures, such ashomeostatic model assessment (HOMA) (27), fasting glucose:insulin ratio
(G:I ratio) (28), and quantitative insulin sensitivity check index (QUICKI) (29), are all based
on fasting glucose and insulin levels and essentially provide identical information (26). As
discussed above, even when insulin resistance is assessed using the euglycemic glucose
clamp, it is clear that some women with PCOS have normal insulin sensitivity (3).
Therefore, in clinical practice, treatment of should be directed to the presumed sequelae of
insulin resistance, metabolic syndrome, dysglycemia (impaired fasting glucose [IFG],

Fertil Steril. Author manuscript; available in PMC 2013 January 1.