FDA panel votes that phase III PFS data for sotorasib in KRAS G12C+ advanced NSCLC cannot be reliably interpreted

“I voted no. No one expects a perfect randomized clinical trial, but what we hope for is a small number of issues in trial conduct, and [we hope the trial is] large enough to withstand the uncertainties caused by those sorts of issues,” said Mark R. Conaway, PhD, a professor in the Division of Translational Research and Applied Statistics Department of Public Health Sciences at the University of Virginia School of Medicine. “For [CodeBreaK 200], we seemed to have the opposite [outcome]: a large number of issues that cloud the interpretation of a small observed [PFS] effect.”

During the meeting, the committee outlined several points of potential bias during the study and statistical analysis, including asymmetric early dropouts between the treatment arms, investigator imaging assessments favoring sotorasib, and early crossover per investigator assessment affecting BICR assessments.

In the control arm, 13% of patients (n = 23) withdrew consent and were censored at day 1 for not having a post-baseline assessment compared with 1% of patients (n = 2) in the sotorasib arm. The censorship of patients who dropped out could lead to an overestimation of PFS if withdrawn patients would have had better outcomes.

Study protocol was also amended to include a crossover option for patients in the docetaxel arm after efficacy data for sotorasib were reported from the phase II CodeBreaK 100 trial (NCT03600883). However, the crossover amendment was implemented after 99% of patients had been enrolled on CodeBreaK 200.

“The question before the committee was not one of the efficacy of sotorasib in lung cancer. Rather, specifically, [the question surrounded] the ability to interpret data from a relatively small clinical trial conducted with a highly anticipated agent in a hyper-information age where both patients and providers had high expectations,” Ravi A. Madan, MD, a senior clinician at the National Cancer Institute, said following the vote. “Given that we had powers of statistical permutations discussed that could change interpretations, I had to vote no on the reliability of the PFS benefit for this study. The factors that contributed to the lack of certainty come from the small size [of the trial], investigator conduct, and the small 5-week PFS benefit. I do think if the PFS benefit were greater, this would have been a much shorter conversation.”

Topline results from the CodeBreak 200 showed that patients in the sotorasib arm (n = 171) experienced a median PFS of 5.6 months compared with 4.5 months for docetaxel (HR, 0.66; 95% CI, 0.51-0.86). The committee noted that although there was an approximate 5-week difference in PFS between the arms, imaging to evaluate for progression occurred once every 6 weeks, leaving a potential margin of error for when progression occurred. Furthermore, the median follow-up was 15.2 months for the sotorasib arm vs 6.9 months for the docetaxel arm.

The phase III CodeBreaK 200 was intended as a confirmatory study following the accelerated approval of sotorasib. It enrolled patients with KRAS G12C–mutated advanced or metastatic NSCLC who received at least 1 prior line of therapy including chemotherapy and an immune-oncology agent. Patients were randomly assigned 1:1 to receive 960 mg of oral sotorasib per day or 75 mg/m2 of docetaxel once every 3 weeks. Patients in the control arm were permitted to cross over to receive 960 mg of sotorasib once per day upon disease progression.