OBJECTIVES: Placebo can affect subjective outcomes like pain. Cost-effectiveness analyses (CEAs) that compare interventions versus no intervention (rather than “sham intervention/placebo”) should account for the placebo effect. We demonstrate a method to adjust network meta-analysis (NMA) for placebo effects (specifically, by route of treatment administration [ROA]).

METHODS: A Cochrane review of pain treatments estimated a standardized mean placebo effect, which was assumed to reflect the difference between no treatment and mean placebo effect. Bannuru 2015 estimated differences in placebo effect between three ROAs. It was assumed that three levels of placebo in Bannuru 2015 center around the mean placebo effect as estimated by Cochrane and that distribution of these ROAs correspond to the differences between subcutaneous (SC), intramuscular (IM), and intravenous (IV) ROAs. We applied these to an NMA of change in monthly migraine days (MMDs) informing a CEA of monoclonal antibody calcitonin gene-related peptide inhibitors (both IV and SC) or onabotulinumtoxinA (IM) over weeks 1-4 in patients with chronic migraine. We evaluated the impact of placebo adjustment on observed outcomes modeled as the additive effect of placebo and treatment.

RESULTS: Compared to no adjustment, estimates for SC treatments with adjustment had decreased treatment effect (increase in MMDs) over weeks 1-4, ranging from 0.59 to 1.51 days depending on the specific treatment. Results over weeks 1-4 for IM treatment showed an improvement in treatment effect of -0.74 days, while IV treatment showed an improved treatment effect of -1.92 days. When added to the treatment effect, inclusion of the adjustment resulted in IV treatment having higher absolute improvements over all other treatments. Results were consistent with an analysis of weeks 9-12.

CONCLUSIONS: This case study illustrates the importance of addressing varying placebo responses in indirect treatment comparisons informing CEAs.