“Everybody knows that pestilences have a way of recurring in the world; yet somehow, we find it hard to believe in ones that crash down on our heads from a blue sky. There have been as many plagues as wars in history; yet always plagues and wars take people equally by surprise.”
—Albert Camus, The Plague
On a drug’s path to regulatory approval, phase I clinical trials evaluate its toxicity and suggest a recommended dose for the phase II trial to follow. Phase II trials provide an estimate of the activity of a drug or regimen and a better assessment of toxicity. However, it is the randomized clinical trial alone that can confirm the true safety and efficacy of a treatment.
Bruce D. Cheson, MD, FACP, FAAAS, FASCO
Randomized trials are, in theory, supposed to balance for the prognostic factors that we know of, as well as those unknown at the time of randomization. This point has been the mantra for decades, as it harmonizes the treatment populations, providing the best estimate of “truth” regarding the clinical value of any treatment. Theoretically, it removes any excuses for differences in safety or toxicity not directly attributed to the differences in the drug or treatment regimen. What randomization at baseline cannot account for is what happens when variables that might differentially affect the treatment arms are introduced after randomization. Such might be the case if, for example, a devastating plague were to “crash down on our heads from a blue sky,” especially should that plague afflict one arm of a study to a greater extent than the other.
Case in Point: PI3K Inhibitors
On April 21, 2022, the U.S. Food and Drug Administration (FDA) convened an unusual Oncologic Drug Advisory Committee (ODAC) meeting, which followed the publication of its analysis of the status of the four PI3K inhibitors that had been approved.1 Instead of focusing on a specific new drug or regimen, the topic was a diverse class of drugs, namely the phosphoinositide-3-kinase (PI3K) inhibitors, studied most extensively in lymphoid malignancies. As of the date of that ODAC meeting, four such PI3K inhibitors had been approved by the FDA across a panoply of indications. These four drugs include idelalisib, duvelisib, copanlisib, and, most recently, umbralisib.
Idelalisib is approved in combination with rituximab for relapsed or refractory chronic lymphocytic leukemia (CLL) based on a comparison against rituximab plus placebo (even though rituximab did not have an indication as a single agent in relapsed or refractory CLL), with a progression-free survival hazard ratio (HR) of 0.18 (95% confidence interval [CI] = 0.10–0.31). At the time of the approval, overall survival was not sufficiently mature. It also received accelerated approval for follicular lymphoma, based on a single-arm trial, with an overall response rate of 54%, and for small lymphocytic leukemia (SLL), with an overall response rate of 58%. The median duration of response was not reached for the patients with follicular lymphoma, but it was 11.9 months for those with SLL.
“Of the four PI3K inhibitors, the umbralisib story particularly presents several features that should prove instructive as we contemplate how to move forward.”— Bruce D. Cheson, MD, FACP, FAAS, FASCO
Tweet this quote
Duvelisib, which targets both PI3Kδ and PI3Kγ also received regular approval for relapsed or refractory CLL, and SLL based on a comparison of the single agent vs the rarely used, anti-CD20 ofatumumab, with a progression-free survival hazard ratio of 0.52 (95% CI = 0.39–0.69). Similarly, overall survival was immature at the time of the approval. It received accelerated approval for relapsed or refractory follicular lymphoma, with a 42% overall response rate. A total of 17% of responses were ongoing at 1 year.
Copanlisib, a pan-PI3K inhibitor that primarily targets PI3Kα and PI3Kδ, received accelerated approval for relapsed follicular lymphoma, with an overall response rate of 59%, lasting a median of 12.2 months.
Umbralisib, a highly selective inhibitor of PI3Kδ and casein kinase-1 epsilon (CK1ε), was the most recent PI3K inhibitor to receive accelerated approval for relapsed or refractory follicular lymphoma after three prior regimens. This approval was based on an overall response rate of 43%, lasting a median of 11.1 months, in follicular lymphoma; and in relapsed or refractory marginal zone lymphoma, the overall response rate was 49%, with a median duration of response not yet reached.
However, although the primary endpoint of the subsequent, requisite, confirmatory randomized trials was met for all four drugs—progression-free survival—for each study, the secondary endpoint of overall survival trended away from the experimental arm and favored the outdated control. Three randomized trials with idelalisib in CLL were terminated for toxicity and increased deaths, notably in previously untreated patients, with a hazard ratio for these studies of 2.29 (95% CI = 1.26–4.18).1 Whereas the drug remains on the market for this indication, for the present time, its label carries a number of black box and other warnings. The drug has been voluntarily withdrawn from the market in follicular lymphoma and SLL because of an inability to complete the confirmatory studies.
In the CHRONOS 3 confirmatory trial for copanlisib in relapsed or refractory FL, copanlisib plus rituximab was more effective than placebo plus rituximab, with a progression-free survival hazard ratio of 0.52 (95% CI = 0.39–0.69); however, the preliminary overall survival hazard ratio was 0.87; and the upper limit to the confidence interval was 1.82.1 As a result, the sponsor has pulled its filings for combination studies in the United States, the European Union, and other markets, pending submission of additional analyses.
Although these trends in overall survival in favor of the control arm were not statistically significant, were associated with large confidence limits, and had no alpha assigned in the statistical design, the observation was considered by the FDA to suggest the experimental arm was “causing harm” to patients, which indicated the dose and schedule selected for the study were inappropriate and thus resulted in an unfavorable safety profile.1 The consequence was that idelalisib and duvelisib were voluntarily withdrawn by the sponsors for certain indications.
Learning From the Umbralisib Story
Yet, of the four PI3K inhibitors, the umbralisib story particularly presents several features that should prove instructive as we contemplate how to move forward. In the confirmatory UNITY-CLL trial, patients with either previously untreated or relapsed/refractory CLL were randomly assigned to receive either chlorambucil plus the anti-CD20 obinutuzumab or the investigational regimen of umbralisib plus the investigational anti-CD20 ublituximab. UNITY-CLL met its primary endpoint of progression-free survival, with a hazard ratio of 0.55 (CI = 0.41–0.72) and had an improved overall response rate (83.3% vs 68.7%) as a secondary endpoint. Nonetheless, the immature overall survival analysis was initially 1.23, with broad confidence intervals.
Several important points emerged from the postmortem on these drugs over the past few months. To begin with, the PI3K inhibitors are not the same drug with different names but have marked variability in structure and isoform specificity. Yet they have been lumped together as a class under a black cloud.
For example, umbralisib differs from its three cousins not only in its structure and specificity, but it also targets CK1ε, which conveys several important immunologic attributes. Each of the drugs has at least one study with a highly favorable progression-free survival, yet with an overall survival hazard ratio, albeit immature in most cases, suggesting potential harm to patients. Umbralisib has been better tolerated than most of its relatives, with a reasonably low incidence of grade 3 or worse class-specific toxicities of special interest (eg, transaminitis, rash, pneumonitis, infection).2 In the UNITY-CLL trial, umbralisib and ublituximab exhibited a remarkably similar pattern of toxicity as the control arm within the same exposure period.
And then cometh the plague. Despite the randomization to ensure balance of prognostic factors between the arms at the time of randomization, COVID-19 provided an unanticipated imbalance postrandomization that impacted umbralisib/ublituximab disproportionately to chlorambucil/obinutuzumab. At the first ad hoc, FDA-mandated survival analysis, there were more deaths in the umbralisib/ublituximab arm resulting in the unfavorable hazard ratio. It is noteworthy, however, that the chlorambucil/obinutuzumab therapy lasted for 6 months, and all patients were off therapy at least 2 years prior to the start of the pandemic. In contrast, both drugs in umbralisib/ublituximab were administered until disease progression or intolerance, with about 40% of patients on active treatment with umbralisib/ublituximab at the start of the pandemic. The median follow-up in the umbralisib/ublituximab arm was 21.1 months compared with 5.1 months with chlorambucil/obinutuzumab, which resulted in marked differences in the collection of adverse events, as all data collection across both arms, by design, was terminated 30 days after completion of therapy.
“B-cell depletion and age have been established as two of the most important prognostic factors associated with severe COVID-19 hospitalization or death.”— Bruce D. Cheson, MD, FACP, FAAS, FASCO
Tweet this quote
Detailed analyses of the cause of death revealed a major discordance between the arms, with COVID-19–related deaths dominating the umbralisib/ublituximab arm disproportionately. Remarkably, the umbralisib/ublituximab deaths from COVID-19 occurred predominately during the surges in infection. When the hazard ratio for the COVID-related deaths was adjusted to account for the impact of COVID-19, the overall survival hazard ratio actually favored umbralisib/ublituximab at 0.975. Unfortunately, in a subsequent survival analysis, there was an increasing disparity in deaths, essentially all of which were attributed to infection with COVID. Clearly responsible was the duration of B-cell–depleting treatment. The unfortunate outcome was not only voluntary withdrawal of the UNITY-CLL study from the April 22 ODAC meeting, but also the biologics license application for umbralisib and also termination of the sale of umbralisib and its accelerated approval for follicular and marginal zone lymphomas.
What is glaringly obvious is that UNITY-CLL is not the only clinical trial that is presently being (or intended to be) conducted during the time of COVID, and certainly not the only one to deploy at least one B-cell–depleting agent. B-cell depletion and age have been established as two of the most important prognostic factors associated with severe COVID-19 hospitalization or death. B cells play a fundamental initiating role in the generation of an immune response following either natural infection or active immunization, which is followed by a T-cell response. In many patients, the additive consequence of such drugs confounds the already underlying immunocompromised status of the patients. Numerous recent studies have demonstrated that patients with lymphomas (especially CLL) have a reduced response to COVID vaccination, which is further compromised by the prior or current use of drugs including anti-CD20 antibodies, Bruton tyrosine kinase (BTK) inhibitors, and PI3K inhibitors. And when hospitalized, they have a higher incidence of death.3 This dismal picture has been mitigated by improvements in the management and treatment of these patients.4
Even as the pandemic slowly morphs into being “merely” endemic, COVID-19 and variants will be the sword of Damocles hanging over us for many years to come. It is imperative that we be more proactive in approaching this dilemma in our clinical trials, not only those being planned, but amending those currently ongoing.
As such, several recommendations might be considered:
Greater Care in the Selection of Drug Dose: It is not clear, especially with targeted agents, that phase I trials have routinely arrived at the appropriate phase II dose, and that at the recommended phase II dose, the dose/toxicity ratio has not always been optimized. Indeed, lower doses may have similar efficacy to the higher doses used in late-phase trials, with fewer adverse events. This point is particularly relevant in creating combinations in which the original single-agent phase II doses are often combined, failing to take into account additive toxicities and drug-drug interactions.
Utilize Discontinuous Therapy: Similarly, rather than daily administration of many targeted agents, an intermittent dosing schedule of a brief, fixed dosing period followed by a drug holiday might provide similar efficacy, while limiting toxicity. A newer investigational PI3K in clinical trials, zandelisib, is administered in such a fashion.
Minimize the Duration of Treatment: One of the critical design flaws of umbralisib/ublituximab was its indefinite administration of therapy. Certainly, the continuous anti-CD20 agent puts patients at risk from chronic B-cell depletion, as does the prolonged umbralisib dosing. Nevertheless, numerous other agents used to treat lymphoma and CLL are administered indefinitely, including other PI3K and BTK inhibitors and monoclonal antibodies directed at CD20 and other targets, such as CD19. The development of time-limited approaches with retreatment upon disease progression or risk-adapted (ie, undetectable measurable residual disease [MRD]) strategies with these agents would likely reduce the likelihood of COVID. Importantly, maintenance therapies should be avoided whenever possible, not only in clinical trials, but in clinical practice as well.
“The stigma from the past PI3K inhibitors will likely impede accrual to any future PI3K studies—guilt by association. Can the class ever be resurrected, or is it forever defunct?”— Bruce D. Cheson, MD, FACP, FAAS, FASCO
Tweet this quote
COVID Prevention and Management: Although many patients will be unable to mount an immune response, based on their underlying disease and previous anti-CD20–based therapies status of prior COVID vaccination, many do, and vaccine response should be assessed during participation in clinical trials with B-cell–depleting agents. Standardized vaccination guidelines need to be developed to help permit a clearer assessment of the safety and efficacy of a regimen before it is released to the general population.
Although not uniformly effective, COVID-preventing therapies such as tixagevimab with cilgavimab should be incorporated as a part of the treatment regimen, administered at regular intervals during and after therapy.
Regular polymerase chain reaction testing for COVID should be conducted, with close monitoring of patients during and following B-cell–depleting therapy, whether targeted agents, cellular therapies, or immunotherapies, for the development of COVID or other potentially severe pathogens and rapid intervention instituted:
Assess Patient Immune Status: Determine quantitative immunoglobulin levels, and administer intravenous immunoglobulins as indicated.
Improve Data Capture: Continue to capture adverse events and identify causes of death for all patients, regardless of any difference in the duration of study therapy. Carefully distinguish morbidity and mortality with and from COVID.
Statistical Considerations: The potential for the risk of COVID infection and death over time should be considered in the statistical design, especially with regimens of differing durations.
Patient safety is a primary concern in the design and conduct of clinical trials. However, without appropriate modifications to current approaches to study design and conduct, a drug that might favorably impact the lives of patients may undergo pharmacoptosis, or programmed drug death.5 We have a finite number of effective therapies for patients with CLL and lymphomas. It would be sad to flush an entire class of drugs, each characterized by respectable activity, because they were not developed in the proper fashion.
And what are the implications for the other PI3K inhibitors currently in development? Even if there were a clearer and more careful development strategy for the newer agents, the stigma from the past PI3K inhibitors will likely impede accrual to any future PI3K studies—guilt by association. Can the class ever be resurrected, or is it forever defunct?
Whether we like it or not, the challenges that clinical research has faced in the past are now magnified by COVID-19. Close interaction among clinical researchers, pharmaceutical sponsors, regulatory agencies, and patient representatives will be essential to develop strategies that will be safe and effective in the coming years of the plague.
DISCLOSURE: Dr. Cheson has served as a consultant or advisor to AbbVie, MorphoSys, Karyopharm, Kite, Epizyme, Roche-Genentech, Janssen, Pharmacyclics, Astra Zeneca, BeiGene, and Bristol Myers Squibb; and has received research funding from AbbVie, Roche-Genentech, Trillium, TG Therapeutics, Pharmacyclics, AstraZeneca, and Epizyme.
1. Richardson NC, Kasamon Y, Pazdur R, et al: The saga of PI3K inhibitors in haematological malignancies: Survival is the ultimate safety endpoint. Lancet Oncol 23:563-566, 2022.
2. Davids MS, O’Connor OA, Jurczak W, et al: Integrated safety analysis of umbralisib, a dual PI3Kδ/CK1ε inhibitor, in relapsed/refractory lymphoid malignancies. Blood Adv 5:5332-5343, 2021.
3. Mato AR, Roeker LE, Lamanna N, et al: Outcomes of COVID-19 in patients with CLL. Blood 136:1134-1143, 2020.
4. Roeker LE, Eyre TA, Thompson MC, et al: COVID-19 in patients with CLL: Improved survival outcomes and update on management strategies. Blood 138:1768-1773, 2021.
5. Cheson BD: Letter from the editor. Clin Adv Hematol Oncol 9:515, 2011.
Dr. Cheson is a medical oncologist at The Center for Cancer and Blood Disorders, Bethesda, Maryland, and Scientific Advisor for the Lymphoma Research Foundation.
Disclaimer: This commentary represents the views of the author and may not necessarily reflect the views of ASCO or The ASCO Post.