RE1 Factors Influencing Drug Reimbursement Decision in Scotland

Abstract

A reimbursement prediction model was previously developed based on a dataset of submissions to the Scottish Medicines Consortium (SMC) between 2008 and 2010. The aim was to update and re-analyze the dataset, and to test internal and external validity of the prediction model on submissions from 2011 and 2012. A database of submissions between January 2005 and March 2012 was created. Data of 405 applications were collected, including information on the reimbursement decision (yes/no), clinical data and indicators of the health economic model quality supporting the submission. The impact of these variables was estimated with univariate and multivariate logistic regression models. The multivariate model was identified by a backward selection procedure. Internal validity was assessed by the area under the receiver operating characteristic (ROC) curve. External validity was conducted and judged by a classification test to predict the SMC decision based on 2011-2012 data. Out of 405 applications 226 received positive recommendation (56%) and 131 (58%) of them were reimbursed with restriction, e.g. limited patient population or restricted time period. Based on univariate analyses, three factors had the largest significant effect on the reimbursement: poor pharmacoeconomic analysis design (OR=0.03), high ICER (OR=0.16) and unclear/ inferior efficacy outcomes (OR=0.25). The final multivariate model included the following further factors: antineoplastic-immunomodulating agent (OR=0.47), combination therapy (OR=2.00), biological drug (OR=0.16), placebo-uncontrolled trial (OR=0.50), extended indication (OR=4.24), innovative drug (OR=2.07). The area under the ROC curve was high; 87.7%. Based on the external validation, using a model estimated on data until December 2010 and a cut-off point of 50%, 79.8% of the predicted reimbursement decisions in 2011-2012 were correctly classified. The new prediction model demonstrates internal and external validity for 2011-2012. Therefore, the model could be used as input when further optimizing the market access strategy for products in clinical development.