Cel-Sci's 11-year cancer drug odyssey ends in failure, sinking stock, but plots FDA filing anyway

A phase 3 clinical trial of Cel-Sci’s immunotherapy candidate has missed its primary endpoint. The drug failed to improve on the overall survival achieved by standard of care in the overall population of head and neck cancer patients, but Cel-Sci plans to seek FDA approval on the basis of a subgroup analysis.

Investigators randomized 928 patients to receive standard of care or one of two regimens featuring Cel-Sci’s cytokine cocktail Multikine. Cel-Sci was looking to show adding Multikine to standard of care causes at least a 10-percentage point improvement in overall survival, but the study fell short of that goal. The failure sent shares in Cel-Sci down 45%.

However, Cel-Sci said it plans to seek FDA approval anyway. The plan is to focus the filing on a group of participants who didn’t receive chemotherapy as part of their treatment. In that subgroup, overall survival at five years was 14 percentage points higher in the Multikine group, 62.7%, than in the control cohort, 48.6%. While such subgroup analyses are typically less robust than assessments based on the full data set, Cel-Sci said it expects the assessment to support a filing for approval.

“The analysis of this separate group is expected to meet regulatory requirements for FDA submission based on the protocol and Statistical Analysis Plan, which were prospectively concluded before database lock and unblinding,” the company wrote in a statement.

If Cel-Sci seeks approval, it may face questions about the cause of the differences between overall survival in the chemotherapy and non-chemotherapy subgroups. Cel-Sci said “the chemotherapy, cisplatin, was given intravenously and may have negated the survival benefit imparted by Multikine immunotherapy in these patients.” 

The biotech has outlined the next steps without committing to a timeline. The history of Multikine development suggests the process may take time. Cel-Sci began the phase 3 clinical trial in 2010 and disclosed the study had recorded the number of deaths needed for the primary endpoint analysis almost 14 months ago.