Amgen's KRAS Inhibitor Lumakras Faces FDA Scrutiny Over Statistical Bias in Confirmatory Trial

Oct 17, 2023 12:16 CST Updated 12:16

Amgen

Developer of Treatment Drugs for Serious Diseases

1
Amgen Non-Small Cell Lung Cancer

Lumakras

Amgen's marketed drug sotorasib (Lumakras) is an oral targeted therapy for treating non-small cell lung cancer patients with the KRAS G12C mutation. The KRAS G12C mutation arises from cancer-related genetic mutations and is very common in non-small cell lung cancer patients. This mutation alters the structure of the KRAS protein, leading to abnormal growth and division of cancer cells. Sotorasib slows or halts the growth of cancer cells by inhibiting the activity of this mutated KRAS protein.Lumakras is the first therapy approved by the FDA targeting a mutation once considered "undruggable."

Figure 1.Sotorasib Knowledge Graph, Source: PharmaDATA

On May 28, 2021, sotorasib received accelerated FDA approval for the treatment of patients with non-small cell lung cancer harboring KRAS G12C mutations who had previously received at least one systemic therapy. Prior to the accelerated approval of sotorasib, the preferred standard-of-care regimen was single-agent docetaxel.

After Lumakras Launch

Study questioned by FDA

As part of the post-marketing study (PMR) to validate the clinical benefit of sotorasib, Amgen conducted the open-label clinical trial CodeBreaK 200. The trial randomly assigned patients in a 1:1 ratio to receive either oral sotorasib monotherapy or intravenous docetaxel monotherapy, with the primary endpoint being progression-free survival (PFS).

On October 5, the FDA Oncology Drug Advisory Committee discussed the possibility of converting Lumakras' accelerated approval into full approval. But before the official discussion,FDA Releases Document Expressing Concerns Over "Systematic Bias" in Statistics from CodeBreaK 200 Clinical Trial

Specifically, the error of statistical systematic bias generated by Amgen in clinical trials,Mainly due to the contradiction between their experimental design and data analysis.

According to the FDA document, in the CodeBreaK 200 trial, the median Progression-Free Survival (PFS) for the sotorasib group was 5.6 months, while the median PFS for the control group receiving docetaxel was 4.5 months. Based on the data, the p-value derived from the median PFS of both groups and the corresponding 95% Confidence Interval (CI) range was 0.002, leading Amgen to conclude that the primary endpoint had been achieved.

FDA noted that the median PFS for subjects using sotorasib was nearly 5 weeks higher compared to the control group. However, the patient testing intervals in this trial were once every 6 weeks during the first 49 weeks, followed by once every 9 weeks. Comparing the testing intervals with the PFS gain, the FDA concluded that there was a systematic bias in the analysis of the trial results.

PFS events (i.e., the first detection of tumor growth) can occur at any time point between two measurements; therefore, the determination of PFS may be overestimated (e.g., if the first PFS event occurs on the first day of the sixth week but is only detected during the measurement at the 12th week, PFS would be recorded as 12 weeks, whereas it should actually be 6 weeks and 1 day).

Despite such biases, this method of measurement and analysis remains acceptable because the anticipated error in the measurement should be similar between random groups. However, in the case of the CodeBreaK 200 trial, the situation is not so simple. BecauseThe median difference in PFS between the two groups was only 5 weeks, while the measurement interval reached 6 weeks. Therefore, the error caused by measurement was too large relative to the PFS gain. Consequently, the FDA concluded that the PFS gain could not be fully attributed to the drug itself, as it was impossible to rule out that the difference between the two groups was not due to measurement error.

Experimental Design and Data Analysis

The Importance of Clinical Conclusions

In response to statistical flaws in Amgen's data analysis and conclusions, the FDA re-evaluated the PFS data from CodeBreaK 200 using the interval censoring method. Interval censoring is an alternative to the primary analysis. It assumes that for any patient assessed as having progressive disease at the end of the evaluation interval, their PFS event could have occurred at any time during the analysis interval. For individuals who have already experienced a PFS event, the time of the event is considered the left endpoint of the interval. That is, in the case of testing every six weeks, if a PFS event on the 6th week and one day is detected (which can only be detected at the 12th week), then this PFS will be considered as occurring at week 6 in interval censoring, rather than at week 12 in primary censoring.

FDAThe median PFS obtained by the interval censoring method was 4.2 months in the sotorasib group and 4.3 months in the docetaxel group.Contrary to the conclusion reached by Amgen. This further intensified the FDA's doubts about the validity of sotorasib's PFS gain conclusion.Given the issues with trial design and analysis, the FDA is concerned that CodeBreaK 200 may not be considered an adequate and well-controlled trial, and thus may not provide substantial evidence to support the claim of sotorasib's efficacy.

AmgThe statistical error committed in CodeBreaK 200 can actually be analogous to the problem of blindly starting experiments without clarifying the signal-to-noise ratio before implementing DoE (Design of Experiments).

InIn DoE, understanding the signal-to-noise ratio in advance helps researchers better plan and execute experiments, ensuring the credibility and reproducibility of the results.A high signal-to-noise ratio usually indicates that the noise is relatively small compared to the signal, and the impact of noise on the conclusion is also minimal.Understanding this can improve the reliability and stability of experimental results.If the signal-to-noise ratio is low, the experimental results may be affected by random errors, leading to unstable and irreproducible outcomes.Understanding the signal-to-noise ratio also helps determine the required sample size and whether multiple experimental replicates are needed to achieve statistically significant results.

In Amgen's trial, the inherent bias of the measurement method (excessive sampling interval) itself may pose challenges to the reliability and validity of the conclusions.If this issue could be predicted in advance,Perhaps Amgen should shorten the interval of measuring once every six weeks to once every two weeks, instead of relying solely on the p-value to draw conclusions.After all, the cost increase brought by improving the measurement frequency is much smaller than the possible consequences of being questioned by the FDA regarding the clinical trial conclusions.

References:

FDA Brief Document. Oncologic Drugs Advisory Committee Meeting. NDA 214665 s005. Sotorisab.

Scan the WeChat QR code, add the editor of Antibody Circle, and those who meet the requirements can join the Antibody Circle WeChat group!

Please indicate: Name + Research Direction!

All articles reprinted by this official account are intended to convey more information, with the source and author clearly indicated. Media or individuals who do not wish to be reprinted can contact us at (cbplib@163.com), and we will immediately delete the content. All articles represent the views of the author and do not represent the position of this site.