Personalizing elective oocyte cryopreservation: a novel predictive model for oocyte yield across multiple stimulation cycles.

Study Question: Can patient age and ovarian reserve tests predict the number of cryopreserved oocytes in patients undergoing one or more ovarian stimulation cycles for elective oocyte cryopreservation (EOC)?

Summary Answer: A predictive model incorporating patient age, antral follicle count (AFC), anti-Müllerian hormone (AMH), and FSH levels achieved the greatest predictive accuracy.

What Is Known Already: As a consequence of societal evolution, women are increasingly delaying starting a family. However, the natural decline in ovarian reserve and oocyte quality as age advances can increase the risk of age-related fertility decline (ARFD) and involuntary childlessness. EOC is a fertility preservation procedure designed to mitigate against the risk of ARFD. Multiple studies have evaluated the optimum number of cryopreserved oocytes to achieve one or more live births, with many women requiring more than one cycle. Previous studies have modelled oocyte yield in response to ovarian stimulation in single-cycle sub-fertile populations, which limits translatability to a population who are presumed fertile and electively cryopreserving their oocytes. Predictive models incorporating data from multiple cycles in an elective population could aid clinician-patient counselling in women undergoing EOC.

Study Design, Size, Duration: This retrospective cohort study was conducted using data from patients (N = 579) who underwent one or more ovarian stimulation cycles for EOC at the Centre for Reproductive and Genetic Health (CRGH) between 2016 and 2023 inclusive. Baseline characteristics at each cycle, including age, BMI, AFC, AMH, and FSH levels, were recorded.

Participants/materials, Settings, Methods: Cryopreservation of ≥10 oocytes following an ovarian stimulation cycle was classified as a good response, while ≥5 oocytes indicated an adequate response. The following parameter combinations were subsequently evaluated in negative binomial regression models with generalized estimation equation: (i) age, AFC, AMH, and FSH; (ii) age, AFC, and AMH; (iii) age, AMH, and FSH; (iv) age and AMH; and (v) age and AFC. Receiver operating characteristic curves, with corresponding AUC, sensitivity, and specificity values, were generated for all models. R version 4.4 was used for all statistical analyses.

Main Results And The Role Of Chance: Model 1 achieved the highest AUC for predicting a good response, AUC: 0.7922 (95% CI: 0.7628-0.8217), with a corresponding sensitivity of 0.7631 (95% CI: 0.7190-0.8095), and a specificity of 0.694 (95% CI: 0.6580-0.7297). Model 2 achieved the second highest AUC of 0.7919 (95% CI: 0.7625-0.8213), followed by Model 3, AUC 0.7770 (95% CI: 0.7463-0.8078). Model 5 achieved an AUC of 0.7749 (95% CI: 0.7441-0.8056), and Model 4 achieved the lowest AUC of 0.7727 (95% CI: 0.7417-0.8038). Similarly, Model 1 achieved the highest AUC for predicting an adequate response, AUC: 0.7917 (95% CI: 0.7586-0.8249), with a corresponding sensitivity of 0.7255 (95% CI: 0.6940-0.7571), and a specificity of 0.7481 (95% CI: 0.6890-0.8036). Model 2 achieved the second highest AUC of 0.7837 (95% CI: 0.7504-0.8169), followed by Model 5, AUC 0.7729 (95% CI: 0.7391-0.8068). Model 3 achieved an AUC of 0.7723 (95% CI: 0.7376-0.8069), and Model 4 similarly achieved the lowest AUC of 0.7607 (95% CI: 0.7257-0.7958).

Limitations, Reasons For Caution: This analysis, based on data from a single fertility centre, does not incorporate patient ethnicity or previous oocyte yield as model variables. Consequently, while we evaluate the impact of age and baseline ovarian reserve on predictive accuracy, model performance may vary across different patient cohorts.

Wider Implications Of The Findings: Predictive models incorporating patient age and baseline ovarian reserve tests across multiple cycles may aid clinician-patient discussions for women undergoing EOC. Model accuracy could be enhanced by the incorporation of ethnicity and prior EOC outcomes as model variables in large multicentre studies.

Study Funding/competing Interest(s): No external funding was used for this study. None of the authors have any competing interests, nor have they received or are due to receive any payment for writing this article.

Trial Registration Number: This cohort study did not require registration. Following consultation with the Medical Advisory Committee at CRGH, ethical approval was not deemed necessary.