Predictors of pregnancy in assisted reproductive technology (ART) based on baseline characteristics: a retrospective analysis of 9067 cycles

Abstract

OBJECTIVE: A new proposal of a statistical model to predict chances of pregnancy after IVF/ICSI, based on large data base from baseline characteristics of patients. DESIGN: Retrospective study. MATERIALS AND METHODS: An observational data base was constituted from 2000 to 2009 on 9067 cycles individually documented by conditions at baseline, during stimulation and until pregnancy. We conducted a Linear Stepwise regression to estimate the probability of ongoing pregnancy based on the following potential predictors at baseline: Age, Indication, type of infertility, baseline FSH, uterine abnormality, smoking habits, body mass index (BMI), number of previous attempts and previous number of miscarriage. First-order Interactions and main effects were tested at 0.99 confidence level. RESULTS: Our model was constituted by five predictors: primary infertility (Odd Ratio OR=0.592 [0.533, 0.657]), Age 25-35 category (OR=1.61, 1.42-1.82), BMI 18-24 (OR= 1.68, 1.51-1.88), cervical factor infertility (OR=1.826, 1.139-2.92) and male infertility (OR=1.27, 1.13-1.42). These baseline characteristics had a reasonable predictive value (Roc-Curve, C=0.66) and a dispersion of probability predictions between 0.04 and 0.48. CONCLUSION: Our results are based on one of the largest existing data base in ART. Compared with the other proposed models, we confirm age, primary infertility and male-factor infertility as key predictors. In addition, we highlight the important effect of BMI. Our next objective will be to improve the predictive value in integrating stimulation and treatment variables in our prediction model.