External validation of prognostic models for ongoing pregnancy after in-vitro fertilization

This study aimed to validate prognostic models for predicting ongoing pregnancy after the first and second in-vitro fertilization cycles. Models were developed using data from the University Hospital, Nijmegen, 1991-1994 and tested using more recent data from the same centre and data from two other centres. Although the variables included in the models seemed plausible, the predictions of the models were unsatisfactory. The models did not discriminate between women who had achieved pregnancy and women who did not achieve pregnancy; neither could they indicate which women had a (very) low probability of ongoing pregnancy. Taking into account the success rate of a specific clinic or the success rate during a specific period did not show any advantage. The predictions were even inaccurate in the same hospital during another period. It is obvious that these prognostic models should not be used. This study shows the importance of validating prognostic models before their implementation in clinical practice.