Challenges facing clinical trial design in oncology

Abstract

Background: Clinical trials are designed to help patients to live better and/or longer. Each component considered in the design of the trial has the potential to shape and impact the ultimate result. Interpretation of a trial as positive or negative hinges on the endpoint of interest and the trial design. This thesis will consider each decision made during the initial trial design, and their impact on the success of a trial and translation to practice. Methods: Multiple research tools were utilised including systematic reviews, questionnaire based research and prospective studies to address: (i) confounding issues in patient population selection; (ii) the clinical relevance and validity of survival endpoints; (iii) effective integration of quality of life (QOL) tools; and (iv) the feasibility and safety of correlative tests. The results of the individual projects are summarised herein. Results: A systematic review of 166 phase III clinical trials in ovarian cancer demonstrated intermediate endpoints have been increasingly used as the primary endpoint. A relationship was seen between the median progression free survival (PFS) and overall survival (OS) (Pearson correlation +0.78) but very few studies had statistically significant improvements in both PFS and OS. Overall survival was more commonly underestimated in the statistical design than PFS (p<0.001) in ovarian cancer and in comparison to predictions in other cancer sub-types with shorter survival post-progression (p=0.02). The net gain in PFS has fallen with time (5.9 months 1980s; 1.9 months 2010+). Quality of life measures have been increasingly included in studies (2% of trials in 1980s, 64% of trials published since 2010; p<0.001) but were the primary endpoint in only one study. A predefined hypothesis was specified in 8 of 35 studies (23%). A mean rather than an individual score was more commonly reported. A literature review found correlative studies are incorporated in early phase trials to understand disease biology and treatment effect. A sub-study of ARIEL2 (a phase II clinical trial), demonstrated mandatory biopsies were feasible and safe in ovarian cancer. Higher rates of biopsies were achieved in studies with a mandatory rather than an optional biopsy (median 100% vs. 16%). Complications and successful procurement were poorly reported (5% and 23% of studies respectively). A retrospective review of 77 patients with endometrial cancer showed tolerability and the presence of multiple mutations (52% had ≥2 mutation) were barriers to the implementation of personalised medicine. 26% stopped targeted therapy due to toxicity rather than lack of effect. Time to progression on targeted therapy was not better than prior systemic therapy (HR=1.5). Conclusions: Tremendous progress has been seen in oncology trial design but more is needed to optimise resource utilisation and improve patient outcomes. Similar rigour to that used for response evaluation criteria in solid tumours (RECIST) needs to be applied to quality of life tools and the inclusion of correlative tests. It is important to consider and specify the statistical rationale for these aspects of trial design. As investigators we need to remain accountable for each test requested and ensure it is optimally integrated to ensure the study hypotheses are achievable and that these hypotheses reflect benefit for the patient.