PRM94 - Modeling Treatment Sequences In Health Technology Assessments

Abstract

As interventions available in a therapeutic area increase, the relevant decision question in health technology assessment (HTA) expands to include identifying the optimal treatment sequences or position for a treatment in a sequence. This study reviewed economic models capturing treatment sequences published by National Institute for Health and Care Excellence (NICE). Economic models including a treatment sequence assessed by NICE were reviewed as these HTAs generally provide comprehensive detail on modeling. The rationale for modeling a sequence, modeling technique used, and approach to characterizing clinical, cost and utility impacts were evaluated. Forty models were identified that considered treatment sequences in the following therapeutic areas: 32.5% oncology, 17.5% auto-immune, 15% cardiovascular, 10% neurology/mental health, 5% infectious, 5% diabetes, and 15% other. Modeling techniques included discrete event simulation (12%), individual state-transition (15%) and most commonly state transition with tracking states. In most cases treatment sequencing was modeled to reflect clinical practice or clinical trial design. Other reasons included assessing where in a treatment sequence the new treatment belongs or evaluating the addition of more efficacious treatment options to the current treatment sequence. Efficacy inputs were generally based on trials that considered a single intervention and not a treatment sequence. Efficacy was commonly assumed to be the same regardless of line of therapy. Disease-related costs and utilities were mostly determined by disease status or its related events, and only seldomly by line of therapy. Capturing treatment sequences in economic models is important for informing reimbursement, policy and clinical decisions. Key considerations for determining how to best model sequences include: the number of treatment options, patient heterogeneity, key outcome events and event risk (time-varying or constant). A key challenge to model treatment sequences is the scarcity of the clinical data as clinical trials do not commonly study sequences.