CodeBreaK 200 & Sotorasib

A flawed trial

The CodeBreaK 200 trial, investigating the role of sotorasib (Lumakras) in patients with lung cancer, has recently been released in The Lancet. Previously, we published a peer-reviewed analysis of the data presented in the 2022 ESMO congress. Most of our points were confirmed, however the Lancet’s publication raised additional questions. 

But first a primer

KRAS is one of the most commonly mutated oncogenes in cancer and has been considered to be un-druggable. Targeting the oncoprotein coded by the mutated KRAS gene had been a pharmacological challenge remained unsolved until recently.

Sotorasib was the first KRAS inhibitor to be approved in the history of drug development and is surely a chemical success. The drug demonstrated its ability to shrink tumors (= its activity) targeting one specific KRAS mutated protein (G12C). KRAS p.G12C is found in 14% in non-small cell lung cancer (NSCLC) tumors and varying frequencies in other cancers. 

However “activity” (shrinking tumors) doesn’t automatically translate into “efficacy” (living longer or better). That’s why sotorasib was approved under the accelerated pathway, a conditional approval. This means that a confirmatory trial is necessary to ensure patients derive clinical benefits. This is a condition to maintain the drug on the market. This trial was the CodeBreaK 200 trial.

The CodeBreaK 200 trial has several limitations.

The CodeBreaK 200 trial was a randomized controlled trial that was conducted on patients with advanced lung cancer, in the second line setting. The trial has several limitations. 

First, the control used in the trial was docetaxel, which is not the best treatment. At least three other options previously showed superiority over docetaxel alone, in terms of progression-free survival (PFS) and/or overall survival (OS) (docetaxel plus ramucirumab, paclitaxel plus bevacizumab, docetaxel plus nintedanib). If you test a new drug against a treatment which is not the best one, you automatically help your trial to be positive, but it is poorly informative for patients and health-care providers. 

Second, the primary endpoint of the study was not whether people lived longer thanks to sotorasib (OS), but instead whether sotorasib could increase PFS, a composite endpoint which is largely based on arbitrary thresholds of tumor shrinkage. In CodeBreaK 200, against a comparator which is not the best one, the median PFS increase was 33 days.

Even this limited benefit is questionable, because an important number of patients left the study. When this happen with imbalance between groups, like in CodeBreaK 200, the PFS benefit can be artificial because of a phenomenon called « informative censoring ». For more on this: read this paper from Vinay Prasad and Usama Bilal. 

Thirdly, a crossover was allowed, and this was problematic. This is a technical issue, but critical to understand oncology trials. To delve deeper into this topic, I can only recommend my favorite paper on this from Alyson Haslam and Vinay Prasad.

Fourthly, the quality-of-life data were measured over a very limited time-frame, not capturing the overall patient's journey. This is a common limitation in trials previously described in our research group by Alyson Haslam et al.

Our main concern is about the possibility of survival decrement. 

A late amendment occurred during the trial which reduced the sample size from 650 to 330 patients. This amendment occurred “per regulatory guidance”, without detailed explanation. Notably, the trial lost power to evaluate OS, a secondary endpoint.

The trial doesn’t show a survival benefit. What does that mean?

Some basic understanding of statistics is necessary. When a trial aims to demonstrate a survival benefit and fails to do so, a common misinterpretation is that the trial proved equivalence. This is erroneous. The only conclusion that can be derived is that the trial failed to prove superiority. This is simply because, to be able to prove equivalence, you need a much larger number of patients.

In CodeBreaK 200, the absence of survival benefit cannot rule out a survival decrement. More concerning, there are several red flags indicating this could be possible. 

In Figure 3 of the Lancet publication, we can see dramatic early tumor progression in the sotorasib group and none in the docetaxel arm. Numerically, more patients died in the sotorasib group (63.7%) as compared to the docetaxel arm (54.0%). Last, we now know there is a 10-point percentage more fatal treatment-emergent adverse event in patients taking sotorasib (22%) compared to patients in the docetaxel group (12%). 

Last words, important words. 

In medicine, you will (almost) always find patients benefiting from a therapy. If you are a patient, you may benefit from sotorasib and this analysis is a not intended to be a substitute for medical advice.

However, in medicine, we often evaluate benefits and harms on aggregate. That’s why intensive high-dose chemotherapy followed by autologous hematopoietic stem cell transplantation has been abandoned for patients with breast cancer. Did some patients benefited? Probably, but on aggregate, this was a very toxic regimen with no survival benefit and it was halted. In the case of sotorasib, the story is ongoing… it is in the hands of regulatory bodies (FDA, EMA, Swissmedic, and others) to appraise if such data are satisfying. I think we owe better to patients.