Model based meta-analysis (MBMA) to support the development of decision criteria in dental pain studies
Rujia Xie, Bernie Surujbally, Alex Phipps, Zahid Ali, James Hadlow, Maria Sudworth, and Scott Marshall
Objectives: Pre-specified decision criteria for go/no go decision-making of a compound are critical in drug development, especially for novel mechanisms. Model based meta-analysis (MBMA) is a key component in this process estimating a distribution of effects over placebo for the endpoints of interest from observed trials. This enables determination of which endpoint is most sensitive and also helps quantify what a meaningful effect may be. MBMA was used to integrate prior information of drugs studied in the post-surgical dental pain model to quantify an effect over placebo for the endpoint (last observation carried forward) LOCF-based TOTPAR (area under the Pain Relief (PR) curve).
Methods: Placebo controlled, post-surgical dental pain trials including NSAIDs, COX2 inhibitor or opiates reporting PR in subjects, following surgical extraction of at least 1 third molar post-operatively, were included in the meta-database. This consisted of 33 trials encompassing 102 treatment arms and 6188 subjects. A meta-analysis was conducted for both internal and literature data to analyse the efficacy endpoint of TOTPAR6 (0-6 hr post-treatment) and to determine its prediction interval for a given drug class. This information aided determination of a minimum meaningful target value (TV) of TOTPAR6 that would need to be observed in a dental pain study, with a certain level of probability, in order to further invest in the compound. Two decision rules governing the decision were evaluated for probability of being: i)superior to placebo; ii)greater than TV.
Results: The mean [95%CI] TOTPAR6 change from placebo for Ibuprofen 400mg was estimated to be 8.37 [7.45, 9.29] from the meta-analyses. Between study variability was described and an approximate predictive interval for the effect in an unspecified study could be 8.37 [4.83, 11.91]. Using this information, and considering the main objective of the study, it was determined that a TV of 6 would be used as part of the decision criteria to progress a compound to the next stage. The (posterior) probability of the true effect being 6 would need to be at least 25 %, as well as at least 95% (posterior) probability of the true effect being greater than placebo.
Conclusions: This meta-analysis provided a broad overview and understanding of effect size of different classes of analgesic drugs in order to develop quantitatively target values for endpoints in early phase efficacy studies that do not have pre-defined meaningful values.