Templeton prediction model underestimates IVF success in an external validation

Prediction models for IVF can be used to identify couples that will benefit from IVF treatment. Currently there is only one prediction model with a good predictive performance that can be used for predicting pregnancy chances after IVF. That model was developed almost 15 years ago and since IVF has progressed substantially during the last two decades it is questionable whether the model is still valid in current clinical practice. The objective of this study was to validate the prediction model of Templeton for calculating pregnancy chances after IVF. The performance of the prediction model was assessed in terms of discrimination, i.e. the area under the receiver operation characteristic (ROC) curve and calibration. Likely causes for miscalibration were evaluated by refitting the Templeton model to the study data. The area under the ROC curve for the Templeton model was 0.61. Calibration showed a significant and systematic underestimation of success in IVF. Although the Templeton model can distinguish somewhat between women with a high and low success rate in IVF, it systematically underestimates pregnancy chances and has therefore no real value for current IVF practice.