TABLE 2

Cardiovascular risk factors by insulin level.

Study group

Normoinsulinemic Hyperinsulinemic P value P value

Variable Controls PCOS patients PCOS patients (ANOVA) (ANCOVA with BMI)

CV risk count 0.047 ! 0.213 0.64 ! 0.929 1.58 ! 0.77 .0015* .0040*
Waist (cm) 74.7 ! 7.8 88.6 ! 16.7 110.0 ! 13.6 $.0001* .0007*
Waist-to-hip ratio 0.75 ! 0.04 0.81 ! 0.07 0.86 ! 0.06 .033 .15
BMI (g/m2) 23.6 ! 3.4 27.7 ! 7.6 38.9 ! 8.6 $.0001 —
Glucose level (mmol/L) 4.9 ! 0.3 5.1 ! 0.5 5.4 ! 0.8 .08 .14
Insulin level (pmol/L) 51.9 ! 18.0 53.6 ! 16.3 181.3 ! 73.9 — —
Cholesterol level (mmol/L) 4.08 ! 0.72 4.68 ! 1.02 5.11 ! 1.11 .002† .007†
HDL level (mmol/L) 1.39 ! 0.31 1.26 ! 0.34 1.01 ! 0.21 .021‡ .049‡
LDL level (mmol/L) 2.26 ! 0.68 2.84 ! 0.84 3.04 ! 0.70 .002† .006†
TG level (mmol/L) 0.94 ! 0.42 1.29 ! 0.82 2.32 ! 1.27 .002‡ .020‡
SBP (mm Hg) 109 ! 14 113 ! 11 129 ! 13 .21 .37
DBP (mm Hg) 67 ! 6 71 ! 11 79 ! 9 .50 .67
Note: Values are mean ! SD. Glucose, insulin, and lipid values are serum concentrations. Comparisons were made by one-way analysis of variance
(ANOVA) and subsequently with analysis for covariance (ANCOVA) with BMI in the metabolic and blood pressure analyses. For significant correlations
(P$.05), the Tukey-Kramer post hoc multiple comparison procedure was used. BMI " body mass index; DBP " diastolic blood pressure; HDL "
high-density lipoprotein cholesterol; LDL " low-density lipoprotein cholesterol; PCOS " polycystic ovary syndrome; SBP " systolic blood pressure; TG
" triglyceride.
* P$.001 (all three comparisons).
† P$.05 (both PCOS groups vs. control).
‡ P$.05 (normoinsulinemic PCOS vs. hyperinsulinemic PCOS).
Mather. Hyperinsulinemia in PCOS. Fertil Steril 2000.

Laboratory Measurements between the women with PCOS and the control women with
Measurements of lipids and glucose were obtained regard to activity level (2 hours per week overall in both
through our two local clinical chemistry laboratories, which groups) or alcohol consumption (two drinks per week aver-
used identical assays and methodologies. Insulin was mea- age in both groups). Androgen levels were available on a
sured using a double-antibody radioimmune assay (Pharma- subset of the PCOS subjects: serum testosterone was 2.26 !
cia & Upjohn, Inc., Mississauga, ON, Canada), with an 0.16 nmol/L (n " 26) and the free androgen index was
observed intraassay coefficient of variation of 2.4% at 51 13.2 ! 1.4 units (n " 15). Four of the PCOS patients were
pmol/L and 6.1% at 741 pmol/L. The interassay coefficient smokers, and none of the control patients smoked. No sub-
of variation is reported at 5.8% across the range of standards. ject had any documented cardiovascular disease.
Insulin measurements were performed in duplicate in a sin-
gle laboratory. CV Risk and Insulin Resistance
Table 2 documents the comparisons grouping PCOS
Statistical Analysis women into the hyperinsulinemic group if their fasting in-
StatView 5.0 (SAS Institute, Inc., Cary, NC) was used to sulin level exceeded 88 pmol/L (#mean ! 2 SD from the
perform statistical analyses. Student’s t-tests, one-way anal- control group data). Table 3 similarly uses the GIR to sub-
ysis of variance (ANOVA), and analysis of covariance classify women with PCOS, with insulin resistance defined
(ANCOVA) were used where indicated below. Because of as a GIR $ 0.032 mmol/pmol. In Tables 2 and 3, both a
differing numbers of subjects in each group, the Tukey- simple ANOVA for the variable and an ANCOVA to correct
Kramer multiple comparison procedure was used for post for the variation in BMI are presented.
hoc comparisons.
Our primary interest was in the distribution of the aggre-
RESULTS gate measure of CV risk represented by the calculated CV
risk score (14). Grouping subjects by degree of insulinemia
Study Population revealed significant differences in CV risk score between
The demographics for the control and PCOS subjects are groups (Table 2 and Fig. 1), and this difference persisted
presented in Table 1. The groups were of similar age but when BMI was included as a covariate (P".0040). This
differed significantly for most measures relating to CV risk, score also differed significantly across the groups defined by
including BMI. The survey data (not shown) did not differ the GIR (Table 3) (P".037). Seventy-five percent of our

152 Mather et al. Hyperinsulinemia and CV risk in PCOS Vol. 73, No. 1, January 2000