NEW YORK – Advisors to the US Food and Drug Administration on Thursday said they could not reliably interpret progression-free survival data from the Phase III confirmatory study of Amgen's Lumakras (sotorasib) in KRAS-mutant advanced non-small cell lung cancer due to potential bias in the trial.
The FDA's Oncologic Drugs Advisory Committee (ODAC) voted 10-2 that Amgen's progression-free survival estimates from the CodeBreaK 200 trial were unreliable due to that potential bias. In the study, Amgen compared the activity of Lumakras against docetaxel chemotherapy in patients with locally advanced, unresectable or metastatic NSCLC with KRAS G12C mutations who had at least one prior systemic treatment. The FDA granted Lumakras accelerated approval in this setting in 2021 based on data from a single-arm Phase II study, and Amgen is hoping to confirm the benefit seen in the earlier study and garner full approval for the drug based on data from CodeBreaK 200.
The FDA's reviewers had flagged the possibility of systemic bias in the CodeBreaK 200 trial after noting patterns in patient dropouts, data censoring, and how investigators were determining which patients experienced cancer progression. These potential sources of bias led the FDA to question whether Amgen's conclusions about the improvement in progression-free survival were reliable.
The FDA is not required to take the recommendation of its advisory committees, but it does more often than not. While the ODAC's vote does not directly affect Lumakras' accelerated approval status, the FDA may still factor in the expert panel's determination that the Phase III trial data is unreliable as it decides whether to grant Lumakras full approval. The regulator is expected to make that decision by Dec. 24.
At the meeting, Harpreet Singh, director of FDA's division of oncology II, noted that it is not the agency's intent to immediately withdraw approval for Lumakras due to the problems in CodeBreaK 200, adding that the agency has other regulatory pathways available. "We do hear the conflict in [the expert panel's] thought process around the vote about totality [of benefit] and the desire to keep sotorasib on the market as an option for patients," Singh said.
In fact, other companies with cancer drugs granted accelerated approval have been able to submit data from a second trial after their first confirmatory trial failed. Merck did this for Keytruda (pembrolizumab) in second-line liver cancer and Jazz Pharmaceuticals did so for Zepzelca (lurbinectedin) in small cell lung cancer. However, Singh did not say whether the FDA would afford Amgen the opportunity to submit further confirmatory data.
In its presentation, the FDA noted that it would consider all the results from the confirmatory trial in deciding whether to potentially withdraw Lumakras' accelerated approval, such as whether the drug had a survival or safety benefit, along with the current treatment landscape in the disease setting.
Potential bias
The FDA flagged several areas where there may be systemic bias in the CodeBreaK 200 trial. The open-label trial was randomized 1:1 but the regulator argued that because a greater proportion of patients dropped out in the control arm, this may have affected the randomization and, thus, the results.
In the docetaxel arm, 23 patients (13 percent) withdrew consent and never received the treatment, compared to just two patients (1 percent) who never received treatment in the Lumakras arm. These patient dropouts were censored at day 1 in both arms.
The agency also expressed worries about data censoring in the trial after looking into how patients were allowed to crossover from the docetaxel to the Lumakras arm. The FDA noted that 19 patients on docetaxel were determined to be progressing by trial investigators and allowed to cross over and begin Lumakras treatment prior to assessment by blinded independent central review. These patients were then censored to their last BICR assessment prior to crossover because the BICR assessment was potentially confounded by the new therapy start.
The unknown data from censored patients in these two groups may have resulted in an overestimation of Lumakras' progression-free survival benefit because the arms are no longer directly comparable, and the dropouts in the control arm and early crossover patients may have had better outcomes or prognosis, Singh said.
The FDA also noted that potential patient and investigator bias may have led investigators to favor Lumakras. They found discordance in investigator and BICR determinations about when NSCLC patients were progressing in both study arms. In the docetaxel arm, investigators appeared to deem patients as progressing earlier than BICR, but investigators were more likely to say Lumakras-treated patients were progressing later, suggesting that they were willing to either keep patients on Lumakras longer or take patients off docetaxel earlier.
Jeevan Puthiamadathil, a clinical reviewer at FDA's division of oncology II, noted that this discordance may have been due to a perceived loss of equipoise or the belief that one intervention is better than another. Puthiamadathil suggested that the public awareness of Lumakras, including a positive data readout from the CodeBreaK 100 trial and its breakthrough therapy designation while CodeBreaK 200 was ongoing, may have influenced patients' and investigators' beliefs about the efficacy of the drug. In an open-label randomized trial like CodeBreaK 200, this loss of equipoise can affect behavior in the study, he said.
Amgen argued that it took adequate measures to minimize bias in the trial. Gregory Friberg, Amgen's VP of medical affairs, presented models to try to show that the benefit of Lumakras persisted even if the dropouts and censored patients had better than average outcomes on docetaxel.
Trial results
Amgen argued that Lumakras should remain available to patients based on the benefit and tolerability seen in the trial. In CodeBreaK 200, median progression-free survival on Lumakras was 5.6 months versus 4.5 months on docetaxel. Overall survival, however, was similar between the two arms, 10.6 months on Lumakras and 11.3 months on docetaxel, although the trial was not powered to measure this endpoint.
The FDA, meanwhile, questioned whether the five-week progression-free survival benefit seen with Lumakras was enough, especially in light of its concerns about bias. NSCLC patients are typically screened for progression via imaging every six weeks, but their progression could have happened any time between the two scans, Puthiamadathil pointed out. That means the five-week benefit could be a small as five days for some patients, he said.
Richard Pazdur, director of the FDA's Oncology Center of Excellence, also suggested that progression-free survival may not have been the optimal endpoint for this trial. "All of these discussions [about bias] could have been mitigated if we chose the right endpoint [for CodeBreaK 200], and that is overall survival in this setting," he said. "I would hope that the field would have moved forward, that we would be able to show superiority over [docetaxel], a drug that was approved 23 years ago."
Melissa Johnson, director of lung cancer research at Sarah Cannon Research Institute who spoke on behalf of Amgen at the meeting, said she was "not surprised" to see the high rate of dropout in the docetaxel arm, due to its known toxicities. "Sotorasib has been criticized for not beating docetaxel on overall survival, … but if [patients] can live the same amount of time and live better, … they will pick sotorasib every time," she said. "Patients need options beyond docetaxel."
Patient advocates who spoke during the public comment portion of the meeting echoed Johnson's statements that the safety profile for Lumakras, along with the outcomes reported in this study, make it a valid choice for lung cancer patients whose only other option may be chemotherapy.
The two committee members who voted yes, supporting the reliability of the progression-free survival data, said they felt the data as presented was robust enough to show Lumakras' benefit. Pamela Shaw, a senior biostatistics investigator at the Kaiser Permanente Washington Health Research Institute, said she felt the scenarios presented by FDA and Amgen to model different outcomes for the censored patients showed "remarkably consistent efficacy" for Lumakras.
"In the context of an open-label trial, I'm able to make the kind of interpretation I want, which is seeing a statistically different value in progression-free survival [between Lumakras and docetaxel] that I can interpret as modest at best," Shaw explained.
The voting committee members also felt that Lumakras shouldn't be taken off the market. While 10 members voted that the progression-free survival data was unreliable, nearly all of them emphasized that their vote was only reflective of the potential bias in the trial results, not about the potential benefit of Lumakras.
Ravi Madan, chair of the ODAC and a senior clinician at the National Cancer Institute, said there was "relative unanimity" among the voting committee on the potential benefit of Lumakras for KRAS G12C-mutant NSCLC patients. "Despite voting no [on data reliability], most people expressed optimism that this treatment can be effective and perhaps we just need more data from a different trial to give a reliable readout on that," Madan said.