Characterization of intra- and interpatient variability and trajectory of ovarian reserve measures in 121,737 women

Abstract

Measures of ovarian reserve including basal antral follicle count (BAFC), follicle-stimulating hormone (FSH), and anti-Müllerian hormone (AMH) are routinely utilized to counsel women undergoing fertility treatment or preservation. These measures also help in personalizing clinical protocols to minimize risks and optimize success rates, however, no large-scale study has been performed interrogating intra- or inter-individual variability in these metrics. To characterize variability and trajectory of ovarian reserve measures among a spectrum of women. We included 152,858 FSH, 330,266 BAFC, and 79,565 AMH measurements from 121,737 women at 13 fertility clinics (2000-2017). The mean number of FSH, BAFC, and AMH measurements per woman was 2.6, 3.1, and 1.3, respectively. Women were classified as UnknownFecundity if they used a sperm donor and had a female partner or no partner. Others were defined as FertilityTreatment or subgrouped based on diagnosis. We evaluated intra-individual variability of measurements within 3 months using the coefficient of variation (CV; standard deviation/mean) and used Kruskal-Wallis and Mann-Whitney Wilcoxon tests for difference of means between groups. We evaluated the relationship with age using a Generalized Estimating Equations (GEE) model with Poisson family (BAFC) or with Gaussian family (FSH, AMH), controlling for clinic and diagnosis. For trajectory models, we used GEE models controlling baseline age, clinic, and diagnosis. We also tested for interactions between time, age, and diagnosis. The intra-individual CV increased significantly with age for all three metrics, regardless of patient type. We saw greater variation in BAFC and AMH for FertilityTreatment, while FSH variation was greater for UnknownFecundity. For each metric within the FertilityTreatment diagnoses, the greatest variation was seen with diminished ovarian reserve (DOR), while the least was with polycystic ovary syndrome (PCOS). We next examined inter-individual variability among different patient groups. When we controlled for age and clinic, BAFC and AMH in UnknownFecundity, MaleFactor, TubalFactor, and UterineFactor both had similar values. Compared with UnknownFecundity, statistically significant differences in BAFC and AMH were found in DOR (BAFC 45% lower, AMH 1.3 ng/mL lower), Endometriosis (16% lower, 0.8 ng/mL lower), PCOS (50% higher, 3.3 ng/mL higher), and OvulatoryDysfunction (31% higher, 1.95 ng/mL higher). In contrast, for FSH, we only found differences in DOR (1.6 IU higher) and PCOS (1.4 IU lower), while UnknownFecundity was similar to other diagnoses in the FertilityTreatment group. Finally, we studied intra-individual rate of change of these metrics over time using trajectory modeling. We found that the rate of change increased with age, starting at age 30.0 for BAFC, 29.1 for AMH, and 25.9 for FSH. For subtypes, we found that the rate of change for FSH increased at a slower rate for PCOS and UnknownFecundity, compared with other diagnoses. Interestingly, we found that DOR patients had a slower rate of decrease in AMH compared with other patients. We found significant differences in variability and trajectory of ovarian reserve measures among women and at different ages. These findings underscore the importance of taking multiple factors into account when leveraging these markers in counseling and decision-making.