Accuracy of pharmaceutical company licensing predictions: projected versus actual licensing dates

Objectives To determine the accuracy of pharmaceutical companies' predictions of drug licensing timeframes for their products in late stage clinical development. Methods We compared predicted licensing dates provided to the National Institute for Health Research Horizon Scanning Research and Intelligence Centre by pharmaceutical companies against actual marketing authorisation application ( MAA ) and marketing authorisation ( MA ) dates published by the European Medicines Agency for drugs granted authorisation between 2009 and 2013. Key findings One hundred and twenty‐three drugs met our inclusion criteria. About 78% were new drugs and 16% had orphan designation. Less than half (44%) and less than a quarter (24%) of MAA and MA predictions respectively were considered accurate (same month or 1 month either side of the actual date). Pharmaceutical companies were significantly more accurate in predicting MAA dates than MA dates ( P < 0.001). For accurate predictions, the mean duration between the prediction being made and the actual MAA and MA dates were 17.5 and 18.7 months respectively. Out of the total 108 MA predictions, almost two‐thirds (65.4%, 16/26) of short‐term predictions (made in the 2 years prior to the actual MA ) were accurate. For predicted dates that were earlier than the actual MA date, there was a positive relationship between accuracy and the time between the prediction and authorisation. Conclusions Even in predicting near events from well‐informed sources, accuracy is imperfect. There appears to be an optimum time for the provision of accurate information on predicted MAA and MA dates for drugs. This information is crucial for effective early awareness and alert activities.