Background
Worldwide, there is an alarming rise in weight gain and obesity [
1], fueled by adverse lifestyle risk factors [
2]. This increasing trend is especially concerning among women of reproductive age in developed countries like Australia, where up to 44% of women entering pregnancy are overweight or obese [
3]. Obesity increases insulin resistance (IR), worsens pregnancy complications such as gestational diabetes (GDM) [
4], and exacerbates other common insulin-resistant conditions including polycystic ovary syndrome (PCOS) [
5]. Obesity, PCOS, and GDM all impose additional risks in pregnancy and present long-term health risks for mothers and their progeny. However, little is known about the relationships between these common insulin-resistant conditions, with prior meta-analyses exploring relationships between obesity and PCOS or GDM and PCOS showing significant heterogeneity [
6].
PCOS is a common endocrine disorder underpinned by hyperandrogenism and IR that affects up to 18% of women of reproductive age [
7]. The syndrome is associated with a range of metabolic and pregnancy complications, including obesity, IR, type 2 diabetes (T2DM), GDM, and preeclampsia [
7]. PCOS is further aggravated by extrinsic or obesity-related IR, increasing the prevalence and severity of the condition [
8]. GDM is defined as carbohydrate intolerance first recognized during pregnancy and is a common pregnancy-related complication, affecting approximately 7% of women [
9,
10]. GDM with gestational weight gain has significant neonatal and maternal complications, including large for gestational age, macrosomia [
11], premature deliveries, and stillbirth, as well as long-term risks such as the development of T2DM [
12]. While existing studies recognize GDM as a complication of PCOS, with an approximately three-fold risk, most have not adjusted for important confounders, including obesity or body mass index (BMI) [
13,
14].
Studies in the general population that have longitudinally explored BMI or weight change have mostly used conventional growth modeling, assuming the population is homogenous with one weight trajectory [
15]. Studies using group-based trajectory modeling (GBTM) to examine changes in BMI in association with lifestyle-related disorders, such as T2DM and cardiovascular disease (CVD), typically showed heterogeneity of the disorders in terms of differing pathophysiological disease pathways among different segments of the study population categorized according to change in BMI over time. Dhana et al. [
16] identified three distinct trajectories of BMI among participants who developed CVD. The majority of participants who developed the disease were characterized with a stable BMI over time, suggesting BMI alone to be a poor indicator for identifying populations at risk of CVD. Conversely, Vistisen et al. [
17] explored patterns of BMI change over time among 6705 white British participants before the development of T2DM and found three distinct BMI trajectories associated with differing rates of change in IR and cardiometabolic risk factors, with the majority of patients showing modest weight gain prior to diagnosis.
There is evidence to suggest that a higher BMI is more commonly seen in women with PCOS, with BMI over time increasing more in women with PCOS compared to women without [
18,
19]. Currently, there is no literature examining empirically derived categorizations of BMI taking into account change over time both among women of reproductive age in general or specifically among higher-risk women, including those with PCOS. GBTM captures BMI change over time by identifying and empirically grouping individuals into latent classes, thus providing insight into BMI trajectory determinants and taking into account population heterogeneity, allowing group comparisons [
20].
To our knowledge, no studies have examined latent BMI trajectory groups in community-based cohorts of women of reproductive age or among women with PCOS. Furthermore, there is no prior research exploring the relationship between BMI trajectories, their sociodemographic predictors, PCOS, and GDM development. Therefore, we aimed to address these knowledge gaps in a large national longitudinal cohort study of community-based women of reproductive age.
Discussion
Insulin-resistant conditions like obesity, PCOS, and GDM are increasing at an alarming rate in women of reproductive age in developed countries like Australia. Despite sharing IR as an underlying mechanism for the development of diabetes, the relationship between these conditions remains unclear. Here, we have generated novel longitudinal weight gain trajectories from a large community-based Australian cohort of women of reproductive age followed over 13 years. We show significantly distinct trajectories of weight gain predicted by early adult life (beginning of adult reproductive age) BMI. We also show a higher tendency to be characterized as LSG among women of Asian descent. We demonstrate that BMI trajectory is a stronger correlate of GDM than PCOS or other traditional confounders such as age, socioeconomic status, and parity.
Studies exploring developmental trajectories of BMI among women of reproductive age are scarce. In this novel longitudinal study, we discuss developmental trajectories of BMI among Australian women of reproductive age. We identified three rising BMI trajectories, namely LSG, MRG, and HRG. Our findings are relatively consistent with WHO cut-off points for classifying BMI, except for MRG, which began at a slightly lower BMI than the standard recommended value of 25 kg/m
2, the existing WHO cut-off point for the overweight category [
22]. Characterizing BMI trajectories in this cohort revealed that a higher early adult life BMI predicted membership in higher BMI trajectories. This highlights the differential impact of weight in predicting membership in high BMI trajectories. In contrast to prior research reporting significant correlations of socioeconomic factors with BMI, we did not find any significant association of income, education, or health behaviors, such as exercise and smoking, with BMI trajectory group membership [
23,
24]. This could be because socioeconomic status and related factors, such as education and income, tend to remain relatively stable past a certain point in adulthood, possibly displaying less variability among this age cohort. Women of Asian descent were more likely to belong to the LSG compared to the MRG trajectory, consistent with prior literature reporting that women of Asian descent generally have a lower BMI [
25] than Caucasian women, with central obesity rather than BMI being responsible for the metabolic impacts seen among South Asian women [
26]. We were unable to explore similar associations among the HRG trajectory due to the small subgroup sample size and this should be studied further. Overall, this study highlights the need to target weight gain prevention in adolescents or young women of reproductive age.
Trajectories among women with and without PCOS, as well as within the entire sample of women of reproductive age, were similar, in terms of the number of trajectories as well as the observed growth patterns describing change in BMI over time. However, a greater proportion of women with PCOS were characterized as belonging to the overweight and obese (MRG and HRG) trajectories compared to the proportion of women without PCOS or those within the entire sample of women of reproductive age. Our findings are consistent with prior research reporting women with PCOS to have a higher baseline weight as well as an increased rate of weight gain over time compared to women without [
18,
19].
BMI trajectories were independently associated with GDM development and were a stronger correlate than factors such as PCOS, maternal age, socioeconomic status, and parity. Prior evidence and meta-analyses also report an increased risk of GDM in overweight and obese women [
6,
13,
27]. Our 2.50-fold increased risk among those in the HRG trajectory is greater than assessments of GDM adjusting for cross-sectional BMI, possibly reflecting the greater accuracy of using BMI trajectories in predicting the development of GDM. Given that our study findings suggest a predictive role of early adulthood BMI in predicting future BMI, interventions aimed at reducing adolescent and early reproductive life weight could be beneficial for IR-mediated diseases such as GDM. IR is one of the key pathophysiological features of obesity, PCOS, and GDM [
8], independently of BMI. However, IR is further exacerbated by an increased BMI with obesity increasing the prevalence of GDM independent of PCOS [
28]. Previous research has reported that the state of normal pregnancy induces a state of hyperinsulinemic IR [
27], which, when compounded with baseline IR observed among obese women, as well as women with PCOS, may amplify the absolute risk of pregnancy complications [
29] and contribute to a higher risk of GDM among this subgroup.
The main strengths of the study include the large, unselected community cohort with a good retention of participants over time, limited information bias, and the ability to adjust for a number of important confounders. ALSWH staff regularly compare the most recent census data and data from national health surveys with corresponding data from ALSWH surveys to enable them to document the extent to which representativeness is maintained and to quantify biases that might affect the generalizability of findings [
30]. A recent comparison of women who participated in the baseline ALSWH survey with women from the Australian 1996 census within the same age range showed participants to be representative of the general population, underscoring the generalizability of our findings for Australia and similar contexts [
31].
Limitations include the use of self-reported measures of BMI, PCOS, and GDM. However, self-reported measures of PCOS have been validated with menstrual irregularity among this cohort of women. Self-reported BMI, as used in this study, has been validated with anthropometric measurement among the mid-age cohort from ALSWH. Further, the self-reported measure of GDM used in this study has also been validated by Gresham et al. [
32] against objective medical records from the New South Wales Perinatal Data Collection, showing very high validity (≥92%) and reliability between the two datasets. Another limitation is that we do not know the specific timing of GDM and T2DM within the surveys. In Survey 6, it is possible for women to have had GDM and then post-partum T2DM within the 3-year period since the last survey. Thus, we have not excluded these women from our analysis, but acknowledge that we do not know the specific timing of these two events. Finally, it is unclear if the young cohort is representative of Asian women given the very low numbers of Asian women participating in the study. This was most likely due to lower immigration rates of women from non-English speaking countries before 1996, as the cohort did not include women who arrived in Australia after 1996 [
33].