Patent landscape reporting quality and predicting drug approval

Thesis

Attrition is a major contributor to the cost of new drug development. Therefore, the ability to predict drug approval could improve R&D; efficiency. The success of drug development programs is governed both by scientific factors and by “external” factors, such as intellectual property, company characteristics, and economics. Incorporating these external factors into predictions of drug approval could help to improve decision-making.

We postulated that patents might represent a particularly under-used information source in this context. However, in reviewing the academic patent literature, it was apparent that improvements in the quality of available evidence were needed before patent landscapes, a common form of patent analysis, could be useful. To examine the extent and severity of this problem, we conducted a systematic review of the reporting quality of patent landscape articles. Finding evidence that reporting was insufficient, we developed the Reporting Items for Patent Landscapes (RIPL) statement, a reporting guideline which aims to improve it.

Subsequently, we focussed on predicting approval and failure of drug candidates. A systematic review identified only three papers developing multivariable models to predict approval on the basis of analysis of approved and failed drugs. These models are of limited utility due to methodological and reporting quality issues. Therefore, we developed and internally validated two models to predict approval: one incorporating candidate predictor variables to represent the external factors previously mentioned, as well as physicochemical parameters, and the second including only physicochemical parameters. Performance was modest but positive and further work is needed to externally validate the models.

Authors: Smith, J

• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English