Analytical approach:
A previous systematic review and meta-analysis and model were updated and extended for a Finnish context (see Other Publications of Related Interest). The model was conducted from a societal perspective and evaluated treatments over a 24-hour period. Two analyses were presented based on a clinical outcome and a health-related quality-of-life outcome.
Effectiveness data:
The updated systematic review included double blind randomised controlled trials that compared any interventions and dosages in adult patients with moderate or severe intensity migraines.
The network meta-analysis simultaneously evaluated four outcomes using a fixed-effect model: reduction from 2 or 3 on the four-point International Headache Severity (IHS) scale to 0 or 1 (defined as response) (R2); pain free at two hours (PF2); recurrence at 24 hours (Rec24); and adverse events (AE). Treatment effects were evaluated using log-odds ratio and then transformed into probabilities.
The primary effectiveness outcome in the cost-effectiveness analysis was sustained pain-free with no adverse events (SNAE); a composite outcome that combined PF2, Rec24, and adverse events. The formula for SNAE was as follows: SNAE = PF2 * (1-Rec24) * (1-AE).
A cost-utility analysis was conducted using health-related quality-of-life as the outcome.
Monetary benefit and utility valuations:
Two sets of utility scores were used. The primary analysis used utility scores from the quality of well-being self-administered instrument (QWB-SA). A sensitivity analysis was conducted using EQ-5D from another study.
Measure of benefit:
Two measures of benefit were used: the composite SNAE derived from the three clinical outcomes for the cost-effectiveness analysis; and quality-adjusted life-years (QALYs) for the cost-utility analysis.
Cost data:
Costs included acquisition costs of drugs and the cost of productivity losses derived from a published Finnish study. Costs were presented in euros (€). The price year was 2010. Data were derived from a post-hoc analysis of a published report.
Analysis of uncertainty:
Data from the mixed treatment comparison were presented with 95% credible intervals. Sensitivity analyses were conducted that simulated drug purchasing by the cheapest packet instead of the cheapest price per tablet, quality of life using EQ-5D utilities instead of QWB-SA utilities and using the effectiveness results from the previously published meta-analysis rather than the update. The model was fully probabilistic.