Of the 432 participants screened, 26 women did not meet eligibility criteria and a further 17 participants were excluded due to a lack of ovarian ultrasound data (Fig. 1). Of the remaining 389 eligible participants, 134 (34.4%) were classified as PCOS and the remaining 255 (65.6%), as the comparator healthy non-PCOS population. Overall 45.2% of women had high BMI (≥ 23). Compared to non-PCOS subjects, women with PCOS patients were younger (mean (SD): 29.84 ± 4.00 vs 32.24 ± 5.25 years, p < 0.001) and had higher BMI (25.14 ± 6.46 vs 23.08 ± 4.36, p < 0.001).
Table 1 displays the characteristics of women under 4 categories: non-PCOS (normal BMI), non-PCOS (high BMI), PCOS (normal BMI), and PCOS (high BMI). There were no significant differences in age comparing high and normal BMI in PCOS subjects. However in non-PCOS subjects, those with high BMI were a mean 2.21 years older. Malays and Indians had significantly higher BMI compared to those of Chinese ethnicity in those, with or without, PCOS. There were no differences in marital and employment status, monthly income, smoking habits, alcohol and coffee intake between the non-PCOS and PCOS subjects, whether of high or low BMI.
We utilized three models to dissect the relative effects of BMI on four categories of women (Table 2). In Model A, variables were analysed with reference to healthy-normal BMI women. In Model B, variables in normal and high BMI PCOS subjects were compared to healthy-high BMI women. In Model C, the effects of BMI on PCOS subjects were analysed with reference to PCOS women with normal weight.
Effect of BMI on clinical and ovarian parameters (Table 2)
As expected, about 70% of women with PCOS had oligomenorrhea. High BMI had no effect on menstrual cycle length in healthy women (Model A) while having PCOS increased the risk of oligomenorrhea 6.44-fold (Model A, OR 24.07, 95% CI 9.24–62.69). Compared to healthy women with high BMI (Model B), PCOS women whether of normal or high BMI had similar increased risks of oligomenorrhea (6.16- and 5.97-fold respectively). High BMI did not further affect rates of oligomenorrhea in women with PCOS (Model C). We were therefore unable to observe any independent effect of BMI on rates of oligomenorrhoea in healthy, or PCOS women.
In contrast, we observed a step-wise effect of BMI and PCOS on hair growth as measured by mFG scoring. About 7.6% of our cohort were hirsute as defined by mFG score ≥ 5. Compared to healthy women of normal weight (Model A), mFG scores increased by 1.27-fold in healthy women with high BMI, and by 1.61-fold in PCOS women of normal weight, although these differences did not reach statistical significance. Interestingly, women with PCOS and high BMI had a mFG score that was 2.96-fold higher compared to healthy women with normal BMI (Model A, adjusted mean difference; 1.85, 95% CI 0.80–2.90). In comparison to healthy women with high BMI (Model B), PCOS women with high, but not normal, BMI had a significantly higher mean difference in mFG (1.79, 95% CI 0.64–2.93). In women with PCOS (Model C), having high BMI also significantly increased mFG score by 1.85-fold (adjusted mean difference 1.82, 95% CI 0.52–3.12). In total, these data suggested an additive effect of high BMI and PCOS to increase mFG scores, with maximum scores observed in PCOS women with high BMI.
Although there was a trend towards higher acne scores in PCOS women, these differences did not reach statistical significance, and there was no further effect of BMI on acne scores in both healthy and PCOS women. In normal weight women, having PCOS was associated with higher mean adjusted differences in AFC (12.11, 95% CI 8.44–15.78), ovarian volume (2.77, 95% CI 1.60–2.94) mls, and AMH (49.27, 95% CI 35.67–62.86) pmol/L compared to healthy women (Model A). However, there was no additional effect of BMI on mean AFC, mean ovarian volume and AMH in both healthy and PCOS women (Models A, B, C).
Effect of BMI on reproductive hormones (Table 3)
Compared to their healthy normal BMI counterparts (Model A), fasting serum testosterone levels were 1.86- to 1.87-fold higher in PCOS women of normal and high BMI respectively. When the reference was healthy-high BMI women (Model B), PCOS women whether of normal or high BMI exhibited similarly increased testosterone levels of 1.83- and 1.84-fold respectively. High BMI did not change testosterone levels amongst PCOS women (Model C). In healthy women (Model A), an increase in BMI did not change levels of ADT, DHEAS and DHT. PCOS women with high BMI had higher adjusted mean differences in DHEAS (1.24, 95% CI 0.19–2.29) μMol/L and DHT (0.73, 95% CI 0.19–1.26) nMol/L compared to healthy women with normal BMI (Model A). Like testosterone, these differences were no longer observed when the references were healthy-high BMI (Model B) or PCOS-normal BMI women (Model C).
As expected, high BMI significantly decreased SHBG levels by 35% and 52% in healthy and PCOS women respectively (Model A). PCOS status did not have an independent effect on SHBG levels (Model B). Since decreases in SHBG directly increases FAI, healthy women with high BMI had higher FAI levels (Model A). Compared to healthy women with high BMI (Model B), PCOS in the absence of high BMI did not affect FAI levels. The combined effect of PCOS and high BMI, however, resulted in FAI increasing by 2.72-fold (adjusted mean difference 5.27, 95% CI 3.42–7.12). In PCOS women (Model C), high BMI also increased FAI by 2.26-fold (adjusted mean difference 4.93, 95% CI 2.84–7.02). In total, BMI and PCOS had an additive effect to increase FAI levels, with the highest FAI observed in PCOS women of high BMI.
Compared to healthy-normal BMI women (Model A), LH levels were 1.93- and 1.71-fold higher in PCOS women of normal weight and high BMI respectively. Likewise, PCOS women of normal weight and high BMI had a 2.26- and 2.01-fold increase in LH levels respectively when compared to healthy women with high BMI (Model B). While LH levels were higher in PCOS women when compared to their healthy non-PCOS counterparts, we observed that a high BMI significantly lowered LH levels amongst high BMI PCOS women when compared to PCOS women of normal weight (Model C, adjusted mean difference − 2.58, 95% CI − 24.83, − 0.34).
Similarly, an increase in BMI did not change FSH levels in healthy women (Model A). PCOS women with high BMI had lower adjusted mean difference in FSH (− 1.62, 95% CI − 2.75, − 0.49) when compared to healthy women of normal weight (Model A). However, no effect of BMI on FSH levels was observed when compared to normal weight PCOS women (Model C). Estradiol was not affected by weight in both healthy (Model A) and PCOS women (Model C).
Effect of BMI on metabolic biomarkers (Table 4)
Unsurprisingly in healthy women (Model A), high BMI worsened HOMA-IR, cholesterol, triglycerides, LDL and HDL levels. High BMI increased HOMA-IR by 1.71-fold (adjusted mean difference 0.68, 95% CI 0.01–1.35) in healthy women (Model A), and by 2.87-fold (adjusted mean difference 1.43, 95% CI 0.48–2.38) in PCOS women (Model C). Although high BMI increased insulin resistance by 1.60-fold in PCOS women when the reference group was healthy women with high BMI (Model B), this increase of 0.80 did not reach statistical significance. Increased BMI was associated with higher HOMA-IR, triglycerides and lower HDL in the PCOS group (Model C).