Statistics Bedeviled But Did Not Derail Cymbalta Fibromyalgia Approval
This article was originally published in Pharmaceutical Approvals Monthly
Lilly ran into one review issue with its fibromyalgia NDA for Cymbalta (duloxetine) because it did not adhere to a "well established policy" of the FDA division responsible for analgesic drugs regarding imputation of missing data, a recurring statistical issue in pain drug reviews, review documents show
You may also be interested in...
FDA officials respond to the recommendations of a National Research Council Panel on handling missing data in clinical trials, giving early indications of how FDA could be viewing a forthcoming guidance on the issue.
InterMune was left with a mixed bag when it released its major Phase III program of pirfenidone in idiopathic pulmonary fibrosis, which yielded conflicting primary endpoint results. But the company thinks there are enough tricks in that bag to win FDA approval
Lilly acknowledged the importance of using FDA's preferred statistical analysis methods when the firm withdrew its sNDA for Cymbalta (duloxetine) for chronic pain management. One of three new studies submitted did not meet its primary endpoint when analyzed by the FDA analgesic division's preferred statistical method for handling data for patients who don't complete the study: baseline observation carried forward. The study met its primary endpoint using an analytical method known as Mixed Model Repeated Measures; BOCF was a secondary endpoint. The sNDA withdrawal was prefigured by debate over the agency's preference for the more conservative imputation method during the review of Cymbalta's fibromyalgia indication (1Pharmaceutical Approvals Monthly November 2008, p. 42). Lilly expects to submit a revised application in the first half of 2009 that will include data from a recently completed study that the company said achieved statistical significance using both BOCF and MMRM