PREDICTABILITY OF OVARIAN SENSITIVITY INDEX FOR ANTAGONIST-PROTOCOL ICSI CYCLE OUTCOMES

Abstract

This retrospective cohort study aims to revisit the generalizability and determine the predictability of the ovarian sensitivity index (OSI) for ICSI cycle outcomes. Women for ICSI in tertiary ART unit from January 2016 through 2020 were screened. Women were eligible if they had oocyte retrieval in fresh antagonist FSH/hMG stimulated cycles. OSI is calculated as oocytes retrieved × 1000/total gonadotropin dose. OSI standards determined by Huber et al. (1) were utilized to categorize women with OSI <1.69, 1.69-10.07, and >10.07 to poor-, normal-, and high-responders, respectively. For discriminative analysis, predictive and validation analytic models were constructed from total sample (n=622) by randomization in a 1:2 ratio, respectively. The OSI cutoffs, based on ROC analysis with Youden index that predicted clinical pregnancy (CP) and live birth (LB) in the predictive model, were further validated in the validation cohort. Multivariate bootstrapped logistic regression was done for the predictability of OSI for cycle outcomes, adjusted for other covariates. 622 cycles were eligible for analysis. Women had a median (IQR) age of 30 (6) years, BMI of 28 (5.5) kg/m2, and AMH of 2.1(2.3) ng/ml. CP and LB rates per retrieval were 36.2% and 27.7%, respectively. Based on Huber’s OSI, the cohort included poor (18.8%), normal (71.2 %) and high (10%) responders with a mean OSI of 1.04±0.41, 4.89±2.25, and 13.05±2.90, P< .001, respectively. High-responders had a higher CP (58.1% vs 36.4% vs 23.2%, P< .001), LB (50% vs 28% vs 14.5%, P<.001) and top-quality embryos (TQE) formation (87.1% vs 53.5% vs 2.6%, P<.001) rates compared to normal and poor responders, respectively. In the discriminative analysis, the predictive model (n=208) showed that OSI ≥ 3.6 was a predictor for CP and LB (OR 95% CI: 2.26, 1.25-4.08, P=.007 ; 3.00, 1.49-6.05,P=.002, respectively). Also, the selected cutoff has AUC, sensitivity and specificity of 0.60, 70.9 and 49.1 for CP, and 0.64, 75.4, and 48.9 for LB. In the validation cohort (n=414), women with OSI ≥ 3.6 had higher CP, LB, and TQE formation rates compared to their counterparts. The highest predictability of OSI ≥ 3.6 was for TQE production (OR 15.8, 95%CI 9.6-26.0) (r coefficient= .61) (AUC=0.86) compared to CP (OR 2.32, 95% CI 1.52-3.56) (r coefficient=.19) (AUC=0.65) and LB (OR 3.22, 95% CI 1.96-5.27) (r coefficient=.24) (AUC=0.68) with P< .001 for all. Similar predictability trend for TQE formation was also yielded by Huber’s OSIs. In polynomial regression curves, pregnant women yielded more oocytes when compared to non-pregnant receiving the same gonadotropins dose with reduced oocyte yield at doses of ≥2250 and 3000 for the pregnant and non-pregnant cohorts, respectively. In adjusted multivariate regression, AMH and OSI ≥ 3.6 were the only predictors for CP and LB. The gonadotropins requirement per oocyte retrieved is superior to other cycle parameters for predicting outcomes in antagonist cycles, specifically the potentiality to have TQE.