Pregnancy: Prognostic models for the probability of achieving an ongoing pregnancy after in-vitro fertilization and the importance of testing their predictive value

The aim of this study was to create reliable models to predictthe probability of achieving an ongoing pregnancy during in-vitrofertilization (IVF) treatment: model A, at the start of thefirst treatment, model B, at the tune of embryo transfer, andmodel C, during the second treatment at the end of the firstIVF treatment Prognostic models were created using data fromthe University Hospital Nijmegen (n = 757) and applied to thedata from the Catharina Hospital Eindhoven (n = 432), The Netherlands,to test their predictive performance. The predictions of modelB (made at time of embryo transfer) were fairly good (c = 0.672in the test population). For instance, 93% of the patients whohad a predicted probability of achieving an ongoing pregnancyof <10% did not achieve an ongoing pregnancy. However, thepredictions of the other two models (A and C) for Eindhovenwere less reliable. The predictive value of model C was fairlyhigh in Nijmegen (c = 0.673). Its poor performance in the testpopulation may be explained partly by differences in effectivenessof the ovulation stimulation protocols and the decision aboutwhen to discontinue the cycle. Thus, before using prognosticmodels at an IVF centre, their reliability at that specificcentre should be tested.