Phase III Trials and Tribulations

Get Permission

Louis M. Staudt, MD, PhD

Louis M. Staudt, MD, PhD

Wyndham H. Wilson, MD, PhD

Wyndham H. Wilson, MD, PhD

Imagine this. You are a large pharmaceutical company that launches an international randomized phase III trial to assess whether one of your drugs improves the outcome of patients with a common type of cancer. The trial was solidly backed by preclinical evidence that the drug target was essential in this cancer type and equally compelling phase II evidence that the drug had the expected clinical activity. The phase III trial involves the participation of hundreds of patients, scores of physicians, and costs hundreds of millions of dollars.

Six years later, when you analyze the outcomes, the results land with a thud on the floor: the trial failed to meet its primary endpoint. Although a planned subset analysis shows a clinically meaningful and statistically convincing benefit for certain patients, you abandon your efforts to seek regulatory approval and move on to the next challenge.

What is wrong with this picture?

Most importantly, patients with this type of cancer will be deprived of a potentially life-saving treatment because insurance providers will not cover the drug costs and doctors will follow suit in the absence of regulatory approval. Not to mention the fact that the willing participation of hundreds of patients in this clinical trial will have been for naught.

Real-Life Example: PHOENIX Trial of Ibrutinib

This is not a hypothetical scenario but rather precisely what transpired in the phase III randomized PHOENIX trial. This study evaluated the addition of the Bruton’s tyrosine kinase (BTK) inhibitor ibrutinib to standard R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone) chemotherapy in newly diagnosed diffuse large B-cell lymphoma (DLBCL).1

Early preclinical studies clearly indicated that DLBCL tumors of the activated B-cell–like (ABC) subtype were addicted to chronic active signaling by the B-cell receptor (BCR) and were vulnerable to inhibition of BTK, a kinase that links BCR signaling to the prosurvival NF-κB pathway.2 A phase I/II clinical trial involving 80 patients demonstrated that responses to ibrutinib were frequent in ABC DLBCL (37%) but rare in the other major gene-expression subtype, germinal center B-cell–like (GCB) DLBCL.3 Because of this molecular association, the PHOENIX trial enrolled only patients with non-GCB DLBCL. The primary endpoint of the trial was event-free survival, and a prespecified analysis was age at diagnosis, since age had previously been shown to influence treatment efficacy and toxicity in DLBCL.4,5

Although ibrutinib was not associated with an event-free survival advantage when considering all patients, it was associated with a significant 3-year event-free survival benefit of 11% in younger patients (age < 60) and associated 3-year overall survival benefit of 12%.1 Unexpectedly, many patients older than 60 who received ibrutinib had significantly more toxicity and received fewer cycles of R-CHOP compared those on the control arm, leading to no benefit and actually a trend toward worse overall survival. Despite the observed benefit of ibrutinib in younger patients, the trial was declared negative because the effect of age was not a predetermined formal endpoint. No regulatory approval was sought.

Focus on Molecular Subtypes

This decision appears especially problematic in light of subsequent molecular analyses of tumors from patients on the PHOENIX trial.6 By way of background, seven genetic subtypes of DLBCL have been recently defined by the co-occurrence of particular genetic abnormalities in the same tumor.7-9 The genetic abnormalities that occur in these subtypes drive the dependency of each genetic subtype on distinct oncogenic pathways, suggesting the possibility they might respond differentially to targeted therapies.

For example, the MCD genetic subtype frequently acquires mutations targeting the BCR subunit CD79B and the signaling adapter MYD88, which cooperate to promote BCR-dependent NF‑κB signaling.7,10 Notably, ibrutinib monotherapy induces objective tumor regression in most relapsed or refractory DLBCLs with CD79B and MYD88 mutations (80%3) and most primary central nervous system lymphomas,11,12 which typically bear MCD genetic abnormalities.9

Via DNA and RNA sequencing analysis of tumor biopsies from patients on the PHOENIX trial, cases were assigned to three genetic subtypes that are prevalent in non-GCB DLBCL—MCD, BN2, and N1—with others remaining unassigned. With the addition of ibrutinib to R-CHOP, younger patients with MCD and N1 tumors had a 100% event-free survival rate at 3 years but had no more than a 50% survival rate if they received R-CHOP alone.6 The exceptional benefit of ibrutinib in MCD DLBCL fits well with the preclinical and clinical studies previously cited, whereas the benefit in N1 DLBCL was unanticipated. However, analysis of recurrently mutated genes in N1 tumors was consistent with dependence on the BCR-dependent NF-κB pathway and sensitivity to BTK inhibition.6 Thus, a clear biologic basis exists for the therapeutic benefit of ibrutinib in younger patients enrolled on the PHOENIX trial.

How does this genetic analysis alter our view of the PHOENIX clinical trial results?

The implicit reason that subset analyses are downplayed when interpreting clinical trial results is the null hypothesis that the effect was observed by chance, either because of multiple hypothesis testing or sheer dumb luck. The genetic analysis of the PHOENIX trial tested exactly three prespecified hypotheses (ibrutinib benefit in MCD, N1, or BN2), and it was conducted in a fashion that was blind to the clinical outcome results. Therefore, the striking association in younger patients between ibrutinib benefit and genetically sensitive subtypes refutes the null hypothesis, supporting the view that the PHOENIX trial should be considered positive in this age group.

Based on this evidence, we believe a reasonable case could have been made to the U.S. Food and Drug Administration (FDA) for accelerated approval of ibrutinib with R-CHOP therapy in this patient population, particularly given the clinically meaningful, 12% increase in 3-year overall survival observed with this combination. When the FDA grants accelerated approval, it requires confirmatory clinical trial(s) and will proceed to full approval if they are positive but withdraw approval if they are negative.13 Confirmatory trials need not be as costly as the original phase III trial if a clinical or laboratory parameter can be used to enrich for responding patients, based on phase III trial results.

Filling the ‘Interpretive Void’

At present, an oncologist must weigh the risk and benefit of adding ibrutinib to R-CHOP in a newly diagnosed patient with non-GCB DLBCL who is younger than age 60. Such a “grassroots” approach to clinical decision-making is not optimal for patients, since oncologists may be unaware of the evidence and/or insufficiently able to weigh the arguments, not to mention convincing medical insurers of the potential survival value of ibrutinib.

We believe this “interpretive void” can be filled by deliberative bodies constituted by professional societies such as ASCO. The need for such authoritative opinions is likely to grow over time as more and more trials incorporate molecular profiling of tumors and responses to targeted agents are identified in narrow subsets of molecularly defined cancers. The substantial genetic and phenotypic heterogeneity of tumors within a given diagnostic category of cancer is well documented: the malignant growth of tumors in many patients is due to rare genetic “driver” events, a phenomenon known as the “long tail distribution” of cancer drivers.14-16 For example, in prostate cancer, most significantly mutated cancer drivers occur in less than 3% of cases.16

Molecular Profiling in Clinical Trials

The inherently rare nature of most oncogenic alterations in human cancer poses a serious challenge to clinical trialists. For example, consider a prospective trial to validate the exceptional benefit of ibrutinib in the N1 subtype of DLBCL.6 Since the N1 genetic subtype constitutes a small fraction (~1.7%) of all DLBCL tumors, a trial to confirm the benefit of ibrutinib in this subtype would require the screening of hundreds of patients for the activating mutations in NOTCH1 that are pathognomonic for this subtype. Since this is evidently costly and impractical for both patients and trialists, an alternate solution must be designed to accelerate progress in precision cancer medicine.

This nut can be cracked using two complementary approaches. The first relies on carefully conducted clinical trials in which tumors are profiled molecularly and a limited number of hypotheses regarding response in particular molecular subtypes are prespecified. After appropriate correction for multiple hypothesis testing, a significant validation of one or more of these hypotheses could be used to seek regulatory approval for the use of the drug in a molecular subtype(s) of cancer.

A variation would be to design the trial using a training set/validation set approach. In such a setup, multiple hypotheses regarding drug activity are freely evaluated in the training set, and a limited number of promising hypotheses are taken forward for testing in the validation set. This might at first glance seem to be a nonstarter concept that would potentially double the size of the clinical trial with resultant expense and opportunity costs. However, it could be argued that such a concept might be a better bet than the current all-or-nothing paradigm in which a trial is declared positive or negative based on a single primary endpoint.

A second general approach relies on the molecular profiling of all patients with cancer at diagnosis or disease progression. Many major cancer centers are prospectively profiling tumor biopsies from all patients at their institutions, with rapid turnaround of results. In this way, cohorts of patients with relative rare cancer driver alterations can be identified and offered appropriate clinical trials.

Recent noteworthy advances using this approach include the approval of the TRK inhibitor entrectinib in cancers with chromosomal rearrangements involving the genes encoding TRK-family kinases (NTRK1, NTRK2, NTRK3), which occur in ~1% of solid tumors.17 The FDA granted accelerated approval for this drug based on impressive clinical activity in three small, single-arm clinical trials, which included 54 patients with advanced or metastatic solid tumors that had NTRK gene rearrangements.18

Prospective profiling of tumor biopsies also mitigates an overlooked problem of conventional clinical trials involving aggressive cancers, namely a bias toward more favorable outcomes stemming from the time it takes to enroll a patient on a trial and the resulting delay in treatment. This has been convincingly demonstrated in DLBCL based on a meta-analysis of three phase III trials, demonstrating that the survival of patients on the trial was positively correlated with the time between diagnosis and trial enrollment.19 This trend is exacerbated in trials that incorporate a laboratory test to select patients for enrollment, which inevitably lengthens the diagnosis to treatment interval.

The presumptive cause for this bias is the reality that patients with aggressive tumors requiring expeditious treatment cannot wait the several weeks it may take to complete the tests and documentation necessary for trial enrollment. This is particularly a problem in DLBCL, where R-CHOP has a high cure rate in good-risk patients, potentially obviating the benefit of an added agent. This bias was evident in the PHOENIX trial in which patients with ABC DLBCL had a 5-year overall survival of 77% when treated with R-CHOP plus placebo compared with a “real-world” estimate of survival for nonclinical trial patients of ~52%.20 This bias not only skews the biologic attributes of tumors treated in clinical trials, but also makes it more difficult to observe a statistically significant benefit of an investigational agent. By enabling molecular profiling locally at each institution participating in a trial, the time-to-treatment bias of clinical trials would be diminished.

Closing Thoughts

In summary, we are advocating for randomized clinical trials that are conducted in a fashion that does not “throw the baby out with the bath water.” By allowing a limited number of prespecified clinical and molecular endpoints as well as designing the trial to test them in a statistically sound manner, we would accelerate precision cancer medicine and identify patient populations with exceptional responses to targeted therapies. Ideally, the existing regulatory agencies would accept these trial designs and consider the trials for approval. If for some reason the sponsor of a trial does not seek regulatory approval, review boards constituted with domain experts could step into the interpretive void and help interpret the trial results to maximize patient benefit. 

DISCLOSURE: Dr. Staudt has received royalties for gene-expression profiling in lymphoma that has been licensed by NanoString Technologies. Dr. Wilson reported no conflicts of interest.


1. Younes A, Sehn LH, Johnson P, et al: Randomized phase III trial of ibrutinib and rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone in non-germinal center B-cell diffuse large B-cell lymphoma. J Clin Oncol 37:1285-1295, 2019.

2. Davis RE, Ngo VN, Lenz G, et al: Chronic active B-cell-receptor signalling in diffuse large B-cell lymphoma. Nature 463:88-92, 2010.

3. Wilson WH, Young RM, Schmitz R, et al: Targeting B cell receptor signaling with ibrutinib in diffuse large B cell lymphoma. Nat Med 21:922-926, 2015.

4. Pfreundschuh M, Trümper L, Kloess M, et al: Two-weekly or 3-weekly CHOP chemotherapy with or without etoposide for the treatment of young patients with good-prognosis (normal LDH) aggressive lymphomas: Results of the NHL-B1 trial of the DSHNHL. Blood 104:626-633, 2004.

5. Pfreundschuh M, Trümper L, Kloess M, et al: Two-weekly or 3-weekly CHOP chemotherapy with or without etoposide for the treatment of elderly patients with aggressive lymphomas: Results of the NHL-B2 trial of the DSHNHL. Blood 104:634-641, 2004.

6. Wilson WH, Wright GW, Huang DW, et al: Effect of ibrutinib with R-CHOP chemotherapy in genetic subtypes of DLBCL. Cancer Cell 39:1643-1653, 2021.

7. Schmitz R, Wright GW, Huang DW, et al: Genetics and pathogenesis of diffuse large B-cell lymphoma. N Engl J Med 378:1396-1407, 2018.

8. Chapuy B, Stewart C, Dunford AJ, et al: Molecular subtypes of diffuse large B cell lymphoma are associated with distinct pathogenic mechanisms and outcomes. Nat Med 24:679-690, 2018.

9. Wright GW, Huang DW, Phelan JD, et al: A probabilistic classification tool for genetic subtypes of diffuse large B cell lymphoma with therapeutic implications. Cancer Cell 37:551-568, 2020.

10. Phelan JD, Young RM, Webster DE, et al: A multiprotein supercomplex controlling oncogenic signalling in lymphoma. Nature 560:387-391, 2018.

11. Lionakis MS, Dunleavy K, Roschewski M, et al: Inhibition of B cell receptor signaling by ibrutinib in primary CNS lymphoma. Cancer Cell 31:833-843, 2017.

12. Grommes C, Pastore A, Palaskas N, et al: Ibrutinib unmasks critical role of Bruton tyrosine kinase in primary CNS lymphoma. Cancer Discov 7:1018-1029, 2017.

13. Beaver JA, Howie LJ, Pelosof L, et al: A 25-year experience of US Food and Drug Administration accelerated approval of malignant hematology and oncology drugs and biologics: A review. JAMA Oncol 4:849-856, 2018.

14. Bailey MH, Tokheim C, Porta-Pardo E, et al: Comprehensive characterization of cancer driver genes and mutations. Cell 173:371-385, 2018.

15. Loganathan SK, Schleicher K, Malik A, et al: Rare driver mutations in head and neck squamous cell carcinomas converge on NOTCH signaling. Science 367:1264-1269, 2020.

16. Armenia J, Wankowicz SAM, Liu D, et al: The long tail of oncogenic drivers in prostate cancer. Nat Genet 50:645-651, 2018.

17. Drilon A, Siena S, Ou SHI, et al: Safety and antitumor activity of the multitargeted pan-TRK, ROS1, and ALK inhibitor entrectinib: Combined results from two phase I trials (ALKA-372-001 and STARTRK-1). Cancer Discov 7:400-409, 2017.

18. National Cancer Institute: FDA Approves Entrectinib Based on Tumor Genetics Rather Than Cancer Type. September 17, 2019. Available at Accessed February 28, 2022.

19. Maurer MJ, Ghesquières H, Link BK, et al: Diagnosis-to-treatment interval is an important clinical factor in newly diagnosed diffuse large B-cell lymphoma and has implication for bias in clinical trials. J Clin Oncol 36:1603-1610, 2018.

20. Lenz G, Wright G, Dave SS, et al: Stromal gene signatures in large-B-cell lymphomas. N Engl J Med 359:2313-2323, 2008.

Dr. Staudt is Director of the National Cancer Institute’s Center for Cancer Genomics. Dr. Wilson is Head, Lymphoma Therapeutics Section and Senior Investigator of the Lymphoid Malignancy Branch, Center for Cancer Research at the National Cancer Institute.

Disclaimer: This commentary represents the views of the author and
may not necessarily reflect the views of ASCO or The ASCO Post.