Among the PCOS participants, the mean BMI was 30.3 and the mean age was 27.2 years (). All three BMI groups contained women age under 25 years and over 35 years, and non-Caucasian women. The comparison cohort was somewhat older than the PCOS cohort, with a mean age in the 30–34 year bracket (P < 0.01). There were no differences between the PCOS and comparison cohorts by race (P > 0.29). Among the PCOS cohort, 49% self-reported acne and 73% self-reported hirsutism (data not displayed). Note that only limited demographics were included in the comparison cohort in order to maintain anonymity for IRB purposes.
Participant demographics and reproductive history.
Approximately half the PCOS and comparison cohorts (53% and 51%, respectively) scored as having symptoms of depression (P = 0.68, ). The majority of the PCOS sample with depression symptoms in each of the three BMI groups scored as having mild symptoms. The total depression symptom score did not vary by BMI strata within the PCOS cohort (P = 0.07), whereas the depression symptom severity was related to BMI strata within the comparison cohort (P = 0.0049). The depression symptom score was not related to age in the PCOS (P = 0.19) or comparison cohorts (P = 0.63).
QIDS-SR16 total depression symptom severity score (TDS) by cohort and BMI-stratum.
The body satisfaction distribution overall and within BMI strata was similar in the PCOS and comparison cohorts (P > 0.05, ). The body dissatisfaction within the PCOS cohort was generally greater in the obese group than the lean group (P < 0.001 for the weight concern and physical condition subscales, not significant for the sexual attractiveness subscale). Within the comparison cohort, only the weight concern subscale was related to BMI (P = 0.0025).
BES subscale scores by cohort and BMI-stratum.
Among the PCOS population, increased depression symptom severity was strongly correlated with increased body dissatisfaction (denoted by a negative coefficient because the lower BES scores represents greater dissatisfaction) measured through the weight concern and physical condition subscales (P < 0.01, ). Among the lean and overweight PCOS women, this same association was evident (all P < 0.04). In contrast, among the obese PCOS women, the depression symptom severity was only correlated with the BES-physical condition subscale (P < 0.04). Among the comparison cohort, there was strongly correlation between all three body dissatisfaction subscales and depression symptoms (P ≤ 0.01), however this pattern was not evident when the comparison women were stratified by BMI. Depression symptom severity was only correlated with body dissatisfaction for the physical condition subscale among the lean and obese non-PCOS women and only for the sexual attractiveness subscale among the obese non-PCOS women (all P < 0.03).
Spearman rank correlations (rs) for the univariate association between the body esteem score and the QIDS-SR total depression symptom score (TDS).
Within the PCOS cohort, total testosterone (ng/dL) ranged from 19.0 to 161.0 with a mean of 56.8 and 68.4 for non-obese and obese women, respectively. SHBG (nmol/L) ranged from 2.0 – 86.0 with a mean of 38.7 and 25.6 for non-obese and obese women. Free testosterone (pg/mL) ranged from 3.0 to 45.9 with a mean of 10.4 and 15.7 for non-obese and obese subjects. DHEAS (ug/dL) ranged from 27.1 to 479.3 with a mean of 197.4 for non-obese and 166.4 for obese women. There was no association between self-reported hirsutism and either total testosterone, free testosterone, or SHBG (P = 0.27, 0.14, 0.15, respectively). There was an association between self-reported acne and SHBG (P = 0.02), but not between self-reported acne and either total or free testosterone (p=0.38 and 0.84, respectively).
Correlations of these androgens (plus SHBG) with the depression symptom severity indicated no relationship with either SHBG or DHEAS (P ≥ 0.21, data not shown). Graphs of the individual data points (not shown) supported the observation that testosterone and free testosterone were the two most likely androgens to consider in a multivariate model, and that the relationship appeared to be inverse (i.e., greater depression symptom severity at lower androgen levels).
The linear regression model for the non-obese women with PCOS () shows that the depression symptom score had the strongest association with the BES-physical condition subscale (inverse association, P = 0.02), followed by a non-significant association with the BES-weight concern subscale (inverse association, P < 0.10). These relationships held true after the inclusion of androgens, which were not statistically significant (P > 0.50). In contrast, the linear regression model for the obese women with PCOS indicated that the depression symptom severity was not related to either the BES-weight concern subscale or the BES-physical condition subscale scores (P > 0.35), but that depression symptoms may be related to lower levels of free testosterone in a larger cohort, as the p-value was marginal (P = 0.08). If we had included a covariate for BMI in the model and used the entire PCOS cohort instead of stratifying by BMI, these relationships above would not have been apparent. Using the entire PCOS cohort, all covariates in the full model (BES weight concern, BES physical condition, testosterone, free testosterone, age, BMI) except for BMI were not significant (p>0.14, model not displayed). Post-hoc, the restriction of the PCOS cohort to those who did not report current use of depression/anxiety medications (and for whom this data was known) did not alter the multivariate relationships within either the non-obese or obese PCOS cohorts (data not shown).
Multivariate negative-binomial regression model for the prediction of the loge (Total Depression Symptom Score) in PCOS women by BMI stratum.