Challenges of Network Meta-Analysis for Safety Outcomes: The Analysis of Tirbanibulin Against Common Treatments in Europe for Actinic Keratosis

OBJECTIVES: We conducted a systematic literature review and network meta-analysis (NMA) to assess the safety of a new drug, tirbanibulin 1% ointment, compared with common treatments for actinic keratosis (AK) in Europe.

METHODS: We conducted a feasibility assessment (FA) to identify the available networks.

RESULTS: Of 40 randomized controlled trials (RCTs) relevant to the European network, 22 RCTs reported data on discontinuations due to adverse events (AEs) or local skin reactions (LSRs). The 22 RCTs reported discontinuations across disparate populations, including intent-to-treat, per-protocol, safety, and completers, with disparate reasons for discontinuation, including any AE, local AEs, serious AEs, treatment-related AEs (TRAEs), treatment-emergent AEs (TEAEs), LSRs, serious TRAEs, and local TEAEs.

When available data were restricted to the comparable sub-set deemed most clinically relevant (discontinuations due to TEAEs, TRAEs, local AEs or LSRs), discontinuations in both arms of these data comparing tirbanibulin against placebo were zero. Therefore, it was impossible to estimate relative treatment effects under a binomial model: NMA was not the best method for the comparison.

Data on discontinuations were instead compared qualitatively. While eligible interventions generally reported low rates of ≤2%, including tirbanibulin with 0% and placebo between 0 and 0.6%, two of the interventions, fluorouracil 0.5%+salicylic acid and diclofenac sodium 3%, reported rates between 6 and 9%.

Rates of any single severe LSRs were analysed qualitatively because of a lack of data to create a network. Tirbanibulin, fluorouracil 0.5%+salicylic acid and diclofenac sodium showed low severe LSR rates (≤11%); imoquimod showed higher rates (severe redness: 31%; erythema: 25%).

CONCLUSIONS: Although severe LSRs and resulting discontinuations were key outcomes of interest, a lack of reporting and methodological challenges hindered any in-depth comparison of these safety data.

FA is a useful tool to identify issues with data quality, which informs whether to compare data quantitively or only summarize qualitatively.