Modeling any potential change in size, cost or patient benefit, resulting from incorporation of ctDNA, or other “optimizing diagnostic,” as part of a novel “SMARTer” oncology registration trial design.

Abstract

e23027 Background: One limitation in speeding more effective new cancer therapies to patients, and saving lives, is the time and expense of clinical trials. A second limitation is cancer is often a chronic disease, sequentially treated with multiple therapies; unlike in most registration trials, patient progress through a series of therapies. Building on previously published Sequential Multiple Assignment Randomized Trial (SMART) design, herein we model a new design to enhance registration (“SMARTer”). An “optimizing diagnostic” (often a ctDNA test) is used to determine, soon after initial use, if the new therapy is working. Those failing it are switched to the Standard of Care (SOC) for analysis they remain in the “new therapy first” arm, which is compared to the SOC arm. The aim is for all patients who benefit from the new therapy to remain on it, and, in contrast, if they are doomed to fail it, switch to SOC promptly to receive benefit from SOC early in the therapy path. Methods: To determine the value of the “new therapy first” arm we compared it to SOC by modeling assumptions of the accuracy of the optimizing test and the level of benefit of SOC after new therapy. The benefits of a new therapy first and switching, vs. SOC, was compared to a standard new therapy vs SOC “head-to-head”, without switching. Six scenarios, varying the effectiveness of the therapies, each with 16 combinations of sensitivity and specificity, were modeled. Results: Surprisingly, a good optimizing test reduces the number of patients by as much as 84% (from 710 to 114 in total) when new therapy alone had a 40% response rate and SOC a 30% rate (see table). When new therapy alone was no better than SOC, a study of 104 patients was sufficient to document that using the new therapy first added to patient benefit and could save lives, assuming independent mechanisms. Conclusions: SMARTer design thus significantly increases chances of approval, better mimics patients’ real world therapy path, and would reduce both time and cost significantly. An Optimizing Diagnostic, predicting appropriateness of a second dose/round of therapy based on the benefit of the first dose/round, thus facilitates clinical trials in a manner analogous to a Companian Diagnostic predicting the appropriateness of starting the first dose of a new therapy. [Table: see text]