Objectives: To develop a quantitative process to estimate the probability of a recommendation for reimbursement for a new drug with a cost per QALY of between 20,000 and 30,000 pounds in the UK.
Methods: A multi-criteria process was used that included: selection of 7 UK decision makers; pre-workshop questionnaires to identify the most important attributes for the decision and their relative importance; a workshop to develop levels for the most important attributes, map each attribute level to a value function and identify marginal drug profiles (drugs just acceptable for a recommendation for reimbursement); post-workshop estimation of weighted scores for each marginal product based on the estimated values and relative importance of their attribute levels and a logistic regression model to estimate the probability of a favorable reimbursement recommendation as a function of the weighted scores; post-workshop questionnaire to validate the logistic model using participant ratings for a series of hypothetical products.
Results: The most important attributes identified for a recomendation for reimbursement and their relative importance weights (%) were: robustness of clinical evidence (31%); robustness of CE estimates (25%); availability of alternative treatments (8%); incremental efficacy (8%); relative safety (7%); ease of adoption (7%); incremental impact on QOL (5%); budget impact (4%); unmet need (3%); size of population (1%). The attribute levels and relative value for a positive reimbursement recommendation (0-1) for the most important attribute, robustness of clinical evidence, were: ‘endpoints and/or comparators not relevant to payers' (0); ‘weak intermediate clinical endpoints, indirect comparisons needed' (0.25); ‘all clinical endpoints and comparators relevant for NHS' (1). The estimates of the probability of a favorable reimbursement recommendation for the hypothetical products included in the post-workshop questionnaire using the logistic regression model had 71% positive predictive value and 91% negative predictive value when compared to participant decisions for these hypothetical products provided in responses to the post-workshop questionnaire.
Conclusions: An MCDA process can provide both a qualitative understanding and quantitative estimates of the relative importance, attribute levels, and value scales of different product attributes that influence a positive reimbursement decision in the UK.