Tribulations of trials in aggressive lymphoma

In this issue of Blood, Qualls et al elegantly demonstrate the stark discrepancies in relapsed or refractory diffuse large B-cell lymphoma (RR-DLBCL) outcomes between registration studies and routine clinical populations.¹ The authors analyzed 178 patients from 11 institutions receiving Food and Drug Administration (FDA)-approved tafasitamab plus lenalidomide. The original, registrational, single-arm Lenalidomide with MOR00208 in patients with RR DLBCL (L-MIND) study screened 156 patients to recruit 81 with relapsed, favorable-biology, DLBCL (primary refractory and double-hit were excluded), limited to 1 to 3 prior treatments with creatine clearance >60 mL/min. In this study, the overall and complete response rates (ORR and CR) were 60% and 43%, respectively, with median progression-free survival (PFS) of 12.1 months and overall survival (OS) of 33.5 months; in contrast, the real-world population experienced ORR of only 31% (CR 19%), median PFS of 1.9 months, and median OS of 6.5 months. Notably, 89% failed to meet the L-MIND trial eligibility, with no difference in eligible versus ineligible group outcomes, nor largely in therapeutic toxicity.

RR-DLBCL is almost universally fatal and thus remains a significant area of unmet need. Excitingly, 8 pivotal trials have led to accelerated FDA approvals for RR-DLBCL since 2017, all based on phase II studies using surrogate end points, with only 1 being randomized.

The true value of clinical trials lies in applicability of results to clinical decision-making in relevant patient populations. Major problems arise when recruited populations do not sufficiently represent patients in routine care or end points are not aligned with patient needs. Increased sophistication in capabilities of real-world data (RWD) has facilitated high-quality research in identifying such gaps. In lymphoma trials, the gap between trial populations and the real world is growing with direct correlations between increasing complexity of eligibility criteria and the proportion of patients eligible.² The major impact patient selection has on trial results is evidenced by several RWD analyses that show meeting eligibility is itself independently associated with improved outcomes regardless of treatment strategy.^3-5 

Other RWD analyzing approved medicines including polatuzumab and chimeric antigen receptor (CAR)-T cell therapy mirror the findings of Qualls et al. Patients receiving routine care are generally less fit with more aggressive biology and have poorer outcomes than those treated in small registrational phase II trials.^4,6 RWD show as few as 6.5% to 26% of patients with RR-DLBCL receiving routine care are eligible for registration studies, rising only to 47% in treatment-naive cohorts.^2,7,8 Even among recent published RR-DLBCL trials, outcomes were influenced by recruited population heterogeneity with respect to proportions of refractory disease, molecular subtypes, prior therapy, and patient characteristics such as age, performance status, and prognostic index scores.

The ethical dilemma of offering available therapies to patients who were not strictly represented in these registrational studies is hugely challenging. Even though uncertainty exists over genuine benefit in these broader cohorts, such therapies have mostly unrestricted approvals within the disease setting and are rapidly incorporated into national guidelines.

The drive to recruit fitter populations with favorable disease biology to smaller trials for accelerated registration is multifactorial; however, outcomes from new agents evaluated in highly selected cohorts are likely to be enhanced independent of treatment effect and, consequently, achieve the commercial goal of regulatory approval based on the favourable outcomes seen. Although FDA cancer-related accelerated approvals stipulate postmarketing requirements for randomized validation data, a review of this program identified only ∼20% of confirmatory studies demonstrated OS benefit, replicated preapproval study benefits, or demonstrated benefit but in a different surrogate measure altogether.⁹ Many postmarketing phase III trials fail to complete or modify original study populations and take years to report. Moreover, FDA validation requirements for recently approved RR-DLBCL agents vary considerably in specified end points (PFS versus OS), comparators (eg, investigator choice for some, individually listed regimens for others, or no comparator but observational evidence), and end dates.

Toxicity influences patient survival and quality of life—the 2 most valued outcomes by hematology patients. In addition to the documented inferior outcomes from real-world studies, some learnings are to be gained from the recently withdrawn FDA approval for phosphatidylinositol 3-kinase inhibitors. Four agents in 6 randomized trials demonstrated durable response rates and PFS benefits in indolent lymphomas but subsequent detrimental OS resulting from substantial toxicity.¹⁰ 

Some trial design and eligibility decisions inadvertently exclude entire subgroups—for example, the incorporation of crude risk surrogates such as strict diagnosis-to-treatment intervals, in which short screening periods are coupled with arduous screening assessments leading to patient exclusion due to the inability to complete these tests in brief time frames. Restricting eligibility to high International Prognostic Index (IPI) cohorts to capture high-risk patients, but concurrently excluding older patients or those with Eastern Cooperative Oncology Group (ECOG) performance status > 1, leads to results only applicable to the fittest, youngest minority of patients, where toxicity-to-benefit ratios differ considerably from broader populations. Likewise, strict pregnancy prevention strategies, impossible to meet for many women of childbearing potential, leads to this physiologically unique population being underrepresented in analyses.

Early-phase trials do need to test new treatments in relatively fit patients, and exceptionally frail patients with significantly poor disease biology will always struggle to benefit from therapy; however, the widening gap between registrational trials and real-world populations needs to close. The major risk to patients from such poor applicability of registrational studies is not only ineffective treatment and financial and time toxicity but also direct harm via exposure to side effects from potentially futile treatments. The median PFS of 1.9 months described by Qualls et al is not meaningful by any measure.

To successfully improve patient outcomes, not only does trial eligibility need to be less restrictive, but surrogate end points must be appropriately validated, and accurate predictive biomarkers must be identified. In many global jurisdictions, approvals are restricted to agents with randomized evidence of survival benefit, and outcomes from cancer in such jurisdictions vary minimally. RWD ensure accountability of trial results, as exemplified by the accompanying study, but can also successfully augment drug development pathways by robustly and prospectively identifying representative populations from national registries for future research efforts.

To ensure clinicians can communicate meaningful, accurate outcomes and obtain true informed consent, interpretation and application of data in the context of individualized clinical assessment are essential, but regulatory bodies and those designing trials have a responsibility to ensure trials are applicable not only to the very fittest minority but aim to improve the outcomes of the majority of patients in our care.

Conflict-of-interest disclosure: A.B. declares honoraria from Roche, Gilead, Novartis, and Beigene. E.A.H. declares the following conflicts: consultant or advisory role: Roche, Merck Sharpe & Dohme, Astra Zeneca, Gilead, Antigene, Novartis, Regeneron, Janssen, Specialised Therapeutics; travel expenses: Astra Zeneca; research funding (paid to institution): Roche, Bristol Myers Squibb, Merck KgA, Astra Zeneca, and Merck.