Analysis of relative factors and prediction model for optimal ovarian response with gonadotropin-releasing hormone antagonist protocol

ObjectiveTo explore the relative factors for best ovarian response in patients undergoing assisted reproductive technology with the gonadotropin-releasing hormone antagonist protocol and to establish a nomogram prediction model of ovarian response. MethodsA retrospective cohort analysis of the clinical data of 1,944 patients who received assisted reproductive treatment in the Center for Reproductive Medicine of Fujian Maternity and Child Health Hospital from April 1, 2018, to June 30, 2020. According to the number of oocytes obtained, there were 659 cases in the low ovarian response group (no more than five oocytes were retrieved), 920 cases in the normal ovarian response group (the number of retrieved oocytes was >5 but <= 18), and 365 cases in the high ovarian response group (>18 oocytes retrieved). Independent factors affecting ovarian responsiveness were screened by logistic regression, which were the model entry variables, and a nomogram prediction model was established based on the regression coefficients. ResultsThere were statistically significant differences in age, anti-Mullerian hormone, antral follicle count, the diagnosis of endometriosis, decreased ovarian reserve, polycystic ovary syndrome, basal follicle-stimulating hormone and basal luteinizing hormone among the three groups (P < 0.001). Multifactorial stepwise regression analysis showed that female age (0.95 [0.92-0.97], P = 0.000), decreased ovarian reserve (0.27 [0.19-0.38]), P = 0.000), endometriosis (0.81 [0.56-0.86], P = 0.000), antral follicle count (1.09 [1.06-1.12], P = 0.000), basal follicle-stimulating hormone (0.90 [0.85-0.96], P = 0.001), Anti-Mullerian hormone (1.19 [1.13-1.26], P= 0.000) and luteinizing hormone on trigger day (0.73 [0.66-0.80], P= 0.000), were independent factors for the occurrence of different ovarian responses during ovarian hyperstimulation. The predictive model of ovarian responsiveness was constructed based on the above factors, and the model was verified with 589 patients' data from July 1, 2020, to December 31, 2020, at this center. The predicted ovarian response (number of eggs obtained) of a total of 450 patients was consistent with the actual results, with a coincidence degree of 76.4%, and the consistency index of the model is 0.77. ConclusionThe nomogram model was successfully developed to effectively, intuitively, and visually predict the ovary reactivity in the gonadotropin-releasing hormone antagonist protocol and provide guidance for clinical practice.

Automated summary

A retrospective study of nearly 2000 patients undergoing assisted reproductive treatment found that factors such as age, anti-Mullerian hormone, antral follicle count, endometriosis, decreased ovarian reserve, and polycystic ovary syndrome can affect ovarian response. A predictive model based on these factors can help guide clinical practice.