2 Clarke Drive
Suite 100
Cranbury, NJ 08512
© 2024 MJH Life Sciences™ and OncLive - Clinical Oncology News, Cancer Expert Insights. All rights reserved.
Fewer than 4% of patients with cancer in the United States participate in clinical trials, the elderly and individuals with clinically relevant comorbidities are strikingly underrepresented in the research portfolio.
Maurie Markman, MD
Asserting that the clinical trials process in oncology is in a state of disarray is only to declare the obvious. Fewer than 4% of patients with cancer in the United States participate in clinical trials, the elderly and individuals with clinically relevant comorbidities (the “real world” of oncology) are strikingly underrepresented in the research portfolio, and there is no comprehensive and rational clinical trials strategy to move the essential precision medicine paradigm forward. These distressing observations make clear that major reform at multiple levels is needed.
To some, the phase III randomized trial, which requires a minimum of hundreds of patients and takes years to complete, remains the gold standard to move any new concept— or even a minor modification in an existing paradigm—into the clinic for routine use. The requirements for conducting such studies, particularly the need to ensure as much homogeneity of the study population as possible in an attempt to isolate the single study question by excluding real-world patients with cancer and complex clinically relevant comorbidities, does not seem to concern academic and regulatory purists. All that appears to matter is that a statistically significant level of “P <.05” or perhaps “P <.005” is reached so that investigators can declare that one approach is superior to another in a phase III randomized trial that is preferably double-blind and placebo-controlled.In the realm of oncology, noninferiority trials offer another approach for examining the clinical utility of a novel product or combination regimen. In these studies, the fundamental goal is to improve a nonefficacy endpoint such as a reduction in treatment-related adverse effects or shorter duration of therapy while not negatively affecting the efficacy, such as overall survival (OS), that has been established for the current standard of care.
A recently published commentary in the New England Journal of Medicine discusses the concept of noninferiority studies in light of the growing gap between real-world oncology patient care and the need to rapidly investigate novel approaches to improve outcomes versus the requirements of mandating high degrees of statistical certainty before the ponderous bureaucratic machinery moves from a fixed position.1
Although the prospective design of measurable endpoints deemed acceptable to declare a trial finding “positive” (eg, statistically significant percentage reduction in the risk of grade 3 mucositis or stomatitis) can itself be a major challenge in noninferiority trials, the issue highlighted here is the required sample size for such studies in order to reassure regulators and some academics that there is no meaningful loss of efficacy associated with utilizing a specific investigative strategy in routine care.
The authors do not cite specific examples of sample size requirements in oncology studies but they make their point with an example from the realm of arthritis management. A randomized noninferiority trial comparing celecoxib versus ibuprofen or naproxen required a population of more than 24,000 participants to satisfy the statistical and regulatory mandates. Although such a large population could never be contemplated in an oncology trial, the potential for substantially larger studies than currently employed to document noninferiority in cancer care strategies is a very realistic possibility.Consider, for example, the impressive activity observed in trials of combination immunotherapy in numerous clinical settings. The clinically relevant toxicity observed in studies of these therapies is highly concerning. No longer is the discussion regarding toxicity focused on grade 4 neutropenia or anemia, which are generally quite controllable, but rather on complex, expensive, often very difficult-to-manage, and potentially fatal immunological adverse effects that may require emergency department visits and hospitalizations.2
Against this background, it is understandable that a number of investigative teams would devote considerable effort to maintaining the demonstrated efficacy—such as OS, time to disease progression, or objective response rate (ORR)—of a particular strategy in a given setting while attempting to improve the regimen’s profile of serious toxicities. This might include modifying the treatment schedule, reducing the dose of 1 or more agents in combinations, or adding a unique supportive care medication.
So the question is: What will the requirements be for a particular modified protocol, such as a novel combination immunotherapy regimen, to receive regulatory approval or to be employed in routine clinical practice following the publication of promising phase II trial results? And, will these rules apply to each specific proposed modification in each individual clinical setting?
Consider, for example, the recent report of a modified regimen for the use of pembrolizumab plus ipilimumab in the management of advanced melanoma.3 The 2-drug combination of nivolumab plus ipilimumab, employing standard-dose ipilimumab and a reduced dose of nivolumab, had previously been shown to improve OS compared with the drugs administered individually, but with the substantial risk of severe toxic effects.4
In the report of a nonrandomized experience using standard-dose pembrolizumab plus reduced-dose ipilimumab, the ORR associated with the dose-modified regimen was comparable to that reported with the regulatory-approved dosing but with a more favorable toxicity profile.3
What level of evidence will be required to permit the routine use of this or other possible variations of individual dosing or scheduling of the therapeutic regimen in the treatment of patients with advanced melanoma? Furthermore, if the regimen is shown to be safe and effective in a particular tumor type, will individual trials be required in all settings where the strategy has been shown to be clinically useful?
If one believes that randomized trials must be mandated, the next question should be: What impact would such a requirement have on the development of innovative, less-toxic, more broadly beneficial, and far more cost-effective anticancer therapy?