Maurie Markman, MD
Much has been written regarding the serious limitations of phase III randomized trials in the oncology arena. Although it is appropriately acknowledged that such efforts were truly foundational in the establishment of modern oncology and specifically in the area of antineoplastic drug therapy, there is growing evidence that continued reliance on the mantra of “randomized phase III trials” is highly problematic.
For example, an essential feature of a randomized trial is the goal to isolate and carefully define the contribution of the investigative strategy. Therefore, the study arms need to be carefully balanced concerning known cancer-related variables such as stage, tumor grade, prior therapy, and any nonmalignancy-related risk factors such as age, sex, and the presence or absence of specific comorbidities.
As a result, major efforts are made by trial sponsors and investigators to reduce the inherent heterogeneity of study populations, particularly for factors that might result in adverse events or outcomes that are unrelated to the strategy under investigation.
Although success in achieving the aim of decreasing variation in the trial arms will increase the likelihood of a more definitive outcome, there is the serious danger that the results will have little relevance to the real world of cancer management where the majority of patients are elderly and commonly possess one or more comorbidities (eg, cardiac, renal, hepatic, pulmonary dysfunction). And, since the elderly are underrepresented in cancer trials and patients with comorbidities are frequently intentionally or unintentionally excluded from a large proportion of randomized studies, it may be unknown how a clinician should apply the outcomes in routine clinical practice.
Tumor Heterogeneity May Be Barrier
However, an equally distressing concern for phase III randomized trials is the recognized heterogeneity within the cancer itself. It has been appreciated for many years that patients with many malignancies (eg, breast, lung) must be treated based on unique molecular signatures present within the malignancy, and not simply based on the organ where the cancer originated. The phenomenon is relevant in an increasing number of tumors such colon cancer, melanoma, and ovarian cancer. Further, the percentage of patients with a clinically relevant molecular signature and the actual number of individuals who are realistically available to participate in answering a specific question in a phase III randomized trial is becoming a major barrier in the conduct of this specific clinical cancer investigative strategy.
Studies may begin, and then may not be able to complete accrual goals. That is not only a distressing waste of financial resources but, even more critically, a fundamental ethical failure since patients who previously entered the study had specifically been informed their contribution would result in a trial outcome that might potentially help future patients. Premature closure of a study without any reportable data would be breaking that promise to these individuals who had agreed to become research subjects.
Further, the reported results of studies following the often rather rigidly defined statistical format for phase III trials can lead to confusing and objectively unhelpful conclusions, despite the tremendous efforts that went into the design and conduct of the study in question.
Erlotinib Trial Illustrates Problem
Consider, for example, the recent report of a 973-patient randomized, double-blind, placebo- controlled study examining the efficacy of erlotinib, an EGFR inhibitor, as adjuvant therapy for stage IB to IIIA non–small cell lung cancer (NSCLC).1
The eligibility for this international trial included patients whose tumors demonstrated immunohistochemistry-positive EGFR expression or EGFR
amplification by fluorescence in situ hybridization.
The primary study endpoint of disease-free survival (DFS) was not different between the two study arms (HR, 0.90; P
= .324)1 However, for the 161 patients (16.5% of the total study population) who were found to possess an EGFR-activating mutation, there was a striking objective difference in DFS in favor of the erlotinib-treated patients—a median of 46.4 months versus 28.5 months. (HR, 0.61; P
Unfortunately, due to the limited number of patients in this subgroup and issues of study design and conduct, including lack of stratification for EGFR
mutations, there were imbalances in the study arms. The authors concluded that this difference between the mutation-positive patients treated with erlotinib and participants receiving placebo was not statistically significant.