GROUP BASED TRAJECTORY MODELING (GBTM) OF ANTI-MULLERIAN HORMONE (AMH) IN A COHORT OF REPRODUCTIVE AGE AFRICAN AMERICAN WOMEN (AAW)

Abstract

AMH has been used as a marker of ovarian reserve. Limited data are available to observe longitudinal changes in AMH over time. Our objective was to evaluate group based AMH trajectory models in a group of reproductive age AAW over the course of 5 years. This longitudinal study used an established cohort of AAW from Detroit, MI aged 23-35 years, who were recruited as a part of the Study of the Environment, Lifestyle and Fibroids. Anthropometric measurements, self reported medical history, and serum AMH levels (picoAMH assay, Ansh Labs, Webster, TX) were collected from participants at 4 visits over 5 years. GBTM was used to establish clusters of women with similar AMH trajectories using age as the time variable. The number of groups and form of temporal trend (linear, quadratic, etc.) for each group trajectory was determined using Bayes Information Criteria. Time varying BMI and hormonal contraception (HC) use were tested to determine their effect on group trajectories. AMH is not normally distributed so analyses were carried out on the logarithmic scale and back transformed to report predicted AMH. 1551 women with AMH for at least 2 visits were included with mean baseline age of 29.2 ± 3.4 years and median AMH of 4.08ng/mL (IQR 2.28 - 6.70). At visit 4, mean age was 34.2 ± 3.4 years and median AMH was 3.00ng/mL (1.52 - 5.15). GBTM model results indicate 5 different groups of AMH trajectories in this population. Group 1 (0.9% of participants) has a linear decline in log AMH with predicted AMH of 0.03ng/mL at age 23 and 0.001ng/mL at age 35 for a normal weight woman not using HC. Group 2 (6.7% of participants) has a cubic trajectory with an AMH value of 1.80ng/mL at age 23 and 0.31ng/mL at age 35. Groups 3, 4, and 5 (comprising 27.8%, 45.4%, and 19.3% of participants respectively) all had a quadratic trajectory for log AMH, although with different slopes. In group 3, at age 23 the predicted AMH is 2.87ng/mL while at age 35 is 1.37ng/mL; for group 4 the predicted AMH at age 23 is 6.13ng/mL while at age 35 is 3.61ng/mL, and for group 5 the predicted AMH at age 23 is 13.8ng/mL and at age 35 is 8.7ng/mL. Time varying HC use and overweight or obese status is associated with AMH trajectories in some groups. In groups 3, 4, and 5, HC use is associated with 15.4% (p<0.001), 15.5% (p<0.001), and 23.9% (p<0.001) lower AMH compared to not using HC, respectively. Being obese is associated with 10.2% (p=0.016) and 12.1% (p=0.029) lower AMH compared to normal weight in groups 4 and 5, respectively. Group 1, with very low AMH at all 4 visits, is associated with 846.6% (p<0.001) higher AMH in HC users, 258.8% (p<0.001) higher AMH in overweight women, and 611.4% (p<0.001) higher AMH in obese women. We identified five groups of AMH trajectories over time. HC and overweight or obese status were found to influence AMH trajectories in some, but not all, groups. Further study is needed to understand other factors that may impact AMH trajectories or risk factors for group membership.