Demographic and anthropometric data
A total of 465 women with PCOS completed the IPAQ and were included in these analyzes. The characteristics of the study population are given in Table 1. Within the study population, 62% (n = 287) identified as white, 15% (n = 71) as Hispanic, 11% (n = 52) as / Southeast (SE) Asians, 7% (n = 32) as South Asia (SA) and 5% (n = 23) as Black / African American (AA).
As shown in Table 1, the patient population had differences in baseline age, body mass index (BMI), and waist circumference. In addition, the patient population also exhibited differences in socioeconomic factors including income, education level, parenthood, and percentage of people born in the United States. A pairwise comparison of baseline characteristics between racial / ethnic groups showed that the Black / AA cohort was significantly older than the White and South Asian cohorts. The comparison between the Black / AA and Hispanic cohorts did not show significant differences in BMI and waist circumference; however, these two cohorts both had significantly higher BMI and waist circumference than the White, East / Southeast Asian, and South Asian cohorts (Supplementary File 1: Table AT).
Differences in exercise behaviors among racial / ethnic groups
We first quantified the exercise behaviors of each group by Metabolic Equivalents (METs) and compared METs of moderate-intensity exercise, vigorous-intensity exercise, and total exercise (moderate-intensity and vigorous) in different ethnic groups (Table 2). We showed that the METs of vigorous-intensity exercise and total exercise were significantly different across racial / ethnic groups (p<0.01 and <0.01, respectively). However, we did not observe any significant differences in METs compared to moderate intensity exercise between racial / ethnic groups. It should be noted that South Asian patients consistently exhibited the lowest metabolic expenditure among ethnic groups from moderate-intensity exercise, vigorous-intensity exercise, and total exercise. Unadjusted pairwise comparisons for MET outcome suggested that the statistically significant differences for vigorous intensity and total exercise were due to differences between the South Asian and White cohorts (Supplementary File 1: Table B).
Predictors of adequate physical activity
We then assessed exercise behavior using a threshold approach (adequate or inadequate) applying DHHS guidelines. . Using a logistic regression model, controlled for age and BMI, we assessed whether the following socio-demographic factors predicted APA: ethnicity (white, Hispanic, East / South Asian East, South Asian or Black / AA), income level (dichotomized into two earning ≥ $ 75,000 / year or not), education level (graduation from college or not), parity (nulliparous or no) and place of birth (born or not in the United States).
We found that there was a statistically significant difference in achievement of ABS between different racial / ethnic groups when controlled for age (p= 0.01, Table 3) and a trend towards statistical significance when age and BMI are controlled (p= 0.05). It should be noted that South Asian patients had the lowest frequency of APA. White ethnicity as a whole is associated with a higher likelihood of achieving ABS compared to non-white ethnicity (p<0.01, Table 4). We also found statistically significant differences in obtaining the APA according to parity status and level of education (Tables 3, 4). Nulliparous patients were more likely to get APA than parous patients. Patients who had a college degree were more likely to get APA than those who did not. On the contrary, income level and being born in the United States are not predictors of ABS.
We then sought to understand the factors that contribute to differences in exercise behaviors in patients with PCOS of different racial / ethnic groups. We performed multivariate logistic regression analyzes to assess the relationship between race / ethnicity and APA, controlling for factors in addition to age and BMI, such as education level and the status of parous. While the association between ethnicity and APA was significant when checked only for age (p= 0.01), this significance was slightly attenuated in multivariate logistic models that controlled for age and BMI (p= 0.05), as well as for age, BMI, education level and parity status (p= 0.16; Additional file 1: Table C).