Mar 10, 2021
This podcast considers the impact of exclusion criteria on clinical trials, generalizability, and the complexity of modernizing eligibility while maintaining trial integrity.
This JCO Podcast provides observations and commentary on the JCO article “Impact of Organ Function-Based Clinical Trial Eligibility Criteria in Diffuse Large B-cell Lymphoma (DLBCL) Patients. Who Gets Left Behind?” by Khurana et al. My name is Richard Little, and I am at the National Cancer Institute. My oncologic specialty is lymphoid and myeloid malignancies. I am a federal employee with no conflicts of interest to disclose.
The authors have contributed a timely and provocative analysis examining the lack of generalizability and disappointing results of repeatedly negative randomized phase 3 trials conducted over the past 15 years failing to improve on R-CHOP. The authors have proposed that these failures may be in part explained by enrollment onto clinical trials patients who are not really representative of those with the disease, because eligibility criteria too often unnecessarily eliminate patients with laboratory values that reflect organ impairment not pertinent to the agents under study. To test this inference, the authors leveraged a unique resource: The Molecular Epidemiology Resource of the University of Iowa/Mayo Clinic Lymphoma Specialized Program of Research Excellence, or SPORE. This SPORE is an NCI-funded research project initiated in 2002 and collects data in a uniform manner among consenting participants with newly diagnosed DLBCL who undergo treatment managed by their physician. This rich database enabled the evaluation of data in patients treated with R-CHOP or R-CHOP-like immunochemotherapy. The investigators were able to categorize the SPORE participants in reference to the eligibility criteria of the important phase 3 studies recently conducted as either meeting eligibility of the phase 3 studies or not meeting eligibility due to out-of-range laboratory values. Additionally, the SPORE data in some ways mimics clinical trial data in that events such as disease progression and death are captured, and these outcomes could be evaluated comparatively between patients categorized as clinical trial eligible and not eligible. They indexed 7 trials, and for example, found that 12.9% of the SPORE participants did not meet eligibility for the PHOENIX study—that is the phase 3 trial of ibrutinib-R-CHOP vs placebo-R-CHOP showing no benefit of the addition of ibrutinib. Not surprisingly, those scored as ineligible had worse outcomes compared to those scored as eligible. For example, indexing eligibility for the PHOENIX trial, the overall survival hazard ratio for those scored not eligible versus eligible was 1.49 with a p-value of .002. The median survival of the SPORE cohort scored as ineligible for PHOENIX had a median overall survival of around 80 months, and the median survival for those sored as eligible had not yet been reached. Interestingly, treatment-related deaths were not found to be increased in the SPORE patients scored as ineligible compared to those scored as eligible, but death due to progressive lymphoma was higher among those scored as ineligible.
So how does this relate to the inability to improve upon standard DLBCL therapy?
Khurana and colleagues' data highlight several features of clinical trials conduct related to trial outcomes. The unnecessary exclusion of patients based on criteria not specific to the treatment can translate into eligible patients having a more favorable prognosis compared to individuals with the disease who are excluded from studies, but who nevertheless are treated with standard therapy used as a control in DLBCL clinical trials. Populating randomized trials with the best prognosis patients can lead to reduced power to detect an outcome difference, even if one exists.
But what about the unnecessary exclusion of patients from clinical trials based on laboratory values that reflect organ dysfunction and poor outcomes as shown in the SPORE patients? The authors document unequivocally that those patients deemed ineligible are at higher risk of death due to treatment failure: that is they die due to progressive lymphoma more than the SPORE patients scored as meeting clinical trial eligibility criteria. What we don’t know from the study as presented, is whether there were more dose delays and dose reductions among those scored as ineligible. This is fundamental toward an improved understanding of how to repair our clinical trials enterprise in DLBCL. And what I mean by repair, is to broaden and expand clinical trials access to as many patients as possible and to have rigorous design and conduct of trials to detect true signals. There are two essential elements to address, and this data points the way but does not fully answer the questions. To make trials generalizable, we must include as broadly as possible those patients with DLBCL. However, if the SPORE outcomes are explained by an inability to administer therapy as planned in those scored as trial ineligible, then just broadening eligibility criteria could undermine the ability to detect a difference with new effective treatment by including too many patients unable to tolerate the therapy. For example, some have suggested the negative results of the PHOENIX trial may be due to treatment intolerance among those aged 60 years and over. So it appears that trials increasing generalizability and reaching a positive trial endpoint for a novel treatment have a complex interaction, to say the least. It will be essential as diffuse large B-cell lymphoma therapeutic trials are developed to be mindful of differential prognosis and perhaps interrelated treatment tolerance of patients due to multiple factors, including degree of organ dysfunction. The exclusion of patients with the disease understudy from clinical trials should be minimized. Efforts to modernize clinical trial eligibility are being embraced by most stakeholders, and in my opinion is an essential social and medical responsibility for clinical trialists to meet. The challenge is how to accomplish this important objective and appropriately design studies to enhance the ability to detect the desired study endpoint. To answer the question posed by the authors, we must endeavor to leave no one behind. We can accomplish this through statistical designs such as stratification of the randomization, powered for the groups of interest -- as one such solution. Separate and specific trials for older or more frail patients is another solution. And I dare say, we should focus on efforts to eliminate the CHOP backbone and develop better-tolerated therapy that those with organ dysfunction can benefit from. The VIPOR regimen presented by Melani and colleagues at the 2020 ASH meeting provides an example of the type of DLBCL research that could be of interest. The challenge of generalizability and its relevance to negative clinical trial results is becoming clearer if only to recognize the increasing complexity in meeting the needs of our patients. But isn’t that what motivates us?
This concludes this JCO Podcast. Thank you for listening.