A simple online test may give women considering in vitro fertilization (IVF) a much more accurate idea of their chances of a live birth on the first try, compared with a model based on age alone, the authors of a new study say.
The test, called PreIVF-Diversity (PreIVF-D), uses the patient’s age, body mass index, day 3 follicle-stimulating hormone level, and medical and fertility history, as well as a semen analysis, to estimate her chances of a live birth resulting from her first IVF treatment.
In the study, published online March 21 and in the June issue of Fertility and Sterility, researchers used data from 3 clinics in North America and Europe to develop and validate the model.
PreIVF-D was developed by Univfy Inc. The company’s founders, Wing Wong, PhD, and Mylene Yao, MD, are coauthors of the study. Several other authors are employees of, or hold stock in, Univfy.
Lead author Bokyung Choi, PhD, and colleagues retrospectively analyzed 13,076 first IVF treatment cycles performed at clinics in Spain, Canada, and the United States. They initially developed clinic-specific PreIVF models using 10,957 of the available cases, taking into account baseline clinical variables and data available from each clinic before the patients started IVF, as well as variables and outcomes data from each individual clinic.
The team then used those models to develop a clinic-independent model. “We built PreIVF-D by blending and weighting the individual components from all 3 clinic-specific models to form a resulting model that was adjusted for the different numbers of cases available from each clinic,” they write.
The team simultaneously developed a control age-based model from the same 10,957 cases. They classified the women into age categories identified by the Society for Assisted Reproductive Technologies and Centers for Disease Control and Prevention (Age-D).
The team further used 1061 independent cases from the 3 clinics as a training set for the PreIVF-D model and then validated it on an independent set of 1058 cases from clinics.
The validation was performed by comparing PreIVF-D against the Age-D model based on objective measures: predictive power, discrimination (the ability to distinguish patients based on their prognoses), accuracy, and the percentage of patients that would have been reclassified as having a different predicted probability of having a live birth. “We determined the posterior probability of having a live birth in the first IVF cycle based on the collective phenotype profile of the patient and her male partner, or the patient”s phenotype profile alone if donor sperm is used,” the authors explain.
Compared with Age-D, PreIVF-D showed a 35.7% improvement in the ability to predict live birth, representing more than a 1000-fold increase in predictive power on a linear likelihood scale. PreIVF-D resulted in significantly different live birth probabilities in 86% of cases (P < .05), compared with the Age-D model, with 57% showing higher probability of live birth and 28% a lower probability.
The authors suggest that many patients may avoid IVF because they underestimate their chances of success. If more accurate testing encourages even a small percentage of those women to pursue a pregnancy, “that represents an increase in IVF utilization by good-to-excellent prognosis patients, which in turn would improve the overall success rates and utilization of IVF,” they write.
However, at least 1 independent expert sounds a cautionary note. “IVF is an effective, but expensive, technology for assisting with reproduction,” said Sanjay Agarwal, MD, clinical professor, director of fertility services, and director of the Center for Endometriosis Research and Treatment, University of California, San Diego. “Based on the woman’s age, many couples wrongly assume their chances of success are either substantially better or worse than they really are,” he told Medscape Medical News in an email.
With that issue in mind, Univfy has offered to perform a free analysis for any clinic interested in learning how their patient-specific success rates compare to the PreIVF model.
Published: March 25, 2013