Jul 31 2024 10 mins
In this JCO Precision Oncology Article Insights episode, Miki Horiguchi summarizes two articles: “Biomarker-Driven Oncology Trial Design and Subgroup Characterization: Challenges and Potential Solutions” by Wang, et al. published on June 7, 2024, and “Biomarkers in Oncology: Complexities in Biomarker-Driven Studies and Statistical Analysis” by Uno, et al. published on July 22, 2024.
TRANSCRIPT
Miki Horiguchi: Hello and welcome to JCO Precision Oncology Article Insights. I'm your host, Miki Horiguchi, an ASCO Journal Editorial Fellow. Today, I'll be providing summaries for two articles.
The first article is a review article titled, “Biomarker-Driven Oncology Trial Design and Subgroup Characterization: Challenges and Potential Solutions,” by Dr. Jian Wang and colleagues. Biomarker driven clinical trials represent a key component of precision medicine, focusing on tailoring treatments to patients based on specific biomarkers. By identifying and targeting therapies to patients who are most likely to benefit, these trials aim to improve treatment outcomes and reduce adverse events.
The article highlights several important points to optimize biomarker driven clinical trials. The authors first reviewed US FDA approvals in biomarker defined subgroups and conducted an in-depth analysis of key regulatory considerations. They developed an innovative decision tree to guide designing biomarker based clinical trials. In addition, they clarified the statistical challenges, including ones found in the all-comers study design. The authors found that most of the US FDA approvals are being restricted to the biomarker positive subgroup, indicating that observed treatment benefits in the overall population are heavily influenced by the biomarker positive patients. This raises concerns as the treatment effect in the biomarker negative subgroup may be smaller but still clinically meaningful.
Additionally, achieving adequate statistical power for the biomarker negative subgroup is often not feasible. These factors could limit access to the treatment for biomarker negative patients who might benefit from it. To address these challenges, the authors introduced various statistical methods and conducted numerical studies to compare the performance of several of these methods. They found that a promising approach is a Bayesian Dynamic Borrowing Method that leverages evidence from the biomarker positive subgroup to evaluate the treatment effect in the biomarker negative subgroup. The authors emphasize that any statistical method used for subgroup analysis must be prespecified. Proactive engagement with regulatory authorities and alignment with the guidelines before finalizing study designs and analysis plans are also essential.
The second article is an editorial which accompanies the first article, "Biomarkers in Oncology: Complexities in Biomarker-Driven Studies and Statistical Analysis” by Dr. Hajime Uno and Dr. Miki Horiguchi. In this editorial, the authors introduced additional statistical considerations that can further enhance informed decision making based on the results of biomarker driven oncology clinical trials. Specifically, the authors raised three key points to consider.
Number one is controlling the type 1 error rate. The qualitative assessment of a new treatment involves a statistical test, while regulatory decisions consider the totality of evidence rather than evidence based solely on P values. Statistical tests play a crucial role in determining treatment benefits in each of the three analysis populations, that is, the biomarker positive, the biomarker negative, and the all-comers population. The type 1 error rate of a statistical test is the probability of rejecting the null hypothesis when it is actually true. The threshold for the type 1 error rate is conventionally at 0.05. The threshold value can vary depending on the situation, but maintaining the type 1 error rate at the nominal level is essential to ensure the reliability of the conclusions drawn from a statistical test. Any inflation or deflation of the type 1 error rate from the nominal level can lead to significant issues in regulatory decision making.
Number two is choosing robust and interpretable quantitative summaries of treatment effect. Statistical tests provide a binary outcome aiding regulatory decisions like drug approval. However, quantifying the magnitude of the treatment effect is more informative for clinicians and patients when assessing the risk benefit balance of the treatment. Therefore, the choice of a summary measure to quantify the between group difference is also important. Dr. Wang and colleagues use the Cox Hazard ratio in their study, which is the most common summary measure in oncology trials. Yet this measure relies on several assumptions. Specifically, when it is applied to biomarker driven trials, the proportional hazards assumption must hold in both biomarker positive and biomarker negative subgroups.
In addition, when a stratified Cox analysis is used to integrate the hazard ratio of the two subgroups to derive the hazard ratio for the all-comers population, there is an underlying assumption that the hazard ratios from the biomarker positive and biomarker negative subgroups are the same. These assumptions do not usually hold in practice, and violations of these assumptions can compromise the interpretability of the estimated between group difference and its generalizability to future patient populations. It has also been discussed widely in both statistical and clinical journals that the hazard ratio is difficult to interpret because of the lack of absolute hazards from the treatment and control groups. To address these limitations, Doctors Uno and Horiguchi suggested using alternative summary measures, including restricted mean survival time and average hazard with survival rate, which do not share these limitations and offer more robust and interpretable results than the conventional hazards ratio approach.
Number three is using coherent statistical analysis models for the three analysis populations. In the first article, Dr. Wang and colleagues introduced a Bayesian Dynamic Borrowing approach. The primary analysis of their approach borrowed information from the biomarker positive subgroup only when analyzing the biomarker negative subgroup. They did not perform the borrowing when they analyzed the biomarker positive subgroup. The accompanying editorial highlights the potential bias introduced by this asymmetric approach. Specifically, suppose the treatment effect in the biomaker positive subgroup is pronounced, but that in the biomarker negative subgroup is weaker. In this case, their asymmetric approach produces a more favorable result for the biomarker positive subgroup compared to the symmetric approach, where each subgroup follows the information from the other subgroup. Providing a convincing rationale for using an asymmetric approach or conducting a sensitivity analysis with a coherent approach for both subgroups would be required.
To conclude, biomarker driven oncology trials are diverse and complex, requiring a tailored approach to statistical analysis that considers the unique characteristics of each trial. The Bayesian approach represents one useful analytic approach, but might not be a universal solution for all biomarker driven studies. Further discussions among stakeholders, such as those from regulatory authorities, clinicians, and biostatisticians will stimulate further research on the optimal design and analysis methods for biomarker driven clinical trials in precision oncology.
Thank you for listening to JCO Precision Oncology Article Insights and please tune in for the next topic. Don't forget to give us a rating or review and be sure to subscribe so you never miss an episode. You can find all ASCO shows at asco.org/podcast.
The purpose of this podcast is to educate and to inform. This is not a substitute for professional medical care and is not intended for use in the diagnosis or treatment of individual conditions.
Guests on this podcast express their own opinions, experience, and conclusions. Guest statements on the podcast do not express the opinions of ASCO. The mention of any product, service, organization, activity, or therapy should not be construed as an ASCO endorsement.