Statistical challenges posed by uncontrolled master protocols: sensitivity analysis of the vemurafenib study.

Within the evidentiary hierarchy of experimental inquiry, randomized trials are the gold standard. Oncology patients enter clinical studies with diverse lifestyles, treatment pathways, host tissue environments, and competing comorbidities. Randomization attempts to balance prognostic characteristics among study arms, thereby enabling statistical inference of 'average benefit' and attribution to the studied therapies. In contrast, interpretations of uncontrolled trials require additional scrutiny to attempt to place the findings in the context of external evidence. Counter-factual reasoning and speculation across trials may be obscured by the disproportionate enrollment of prognostic subpopulations which may be unknown from publications of trial reports. Recent modifications to the regulatory environment (Food and Drug Administration Safety and Innovation Act) have elevated the importance of non-comparative trials. Moreover, the emergence of recent innovations in precision medicine have yielded trial designs that partition potentially heterogeneous subpopulations into 'statistically exchangeable' cohorts by histologies, or genetic alterations, further elevating the importance of single-cohort analyses. As patient cohorts become ever more refined into smaller targeted subsets, consumers of reports of uncontrolled trials should be further empowered with improvements in reporting practices that better describe the enrolled prognostic subpopulations and importantly their association with study end points. This article demonstrates the issue with a sensitivity analysis of the findings reported in a recent trial that was devised to evaluate the preliminary clinical efficacy of vemurafenib in BRAF V600 mutation-positive nonmelanoma cancers.

INTEGRATIVE BAYESIAN ANALYSIS OF NEUROIMAGING-GENETIC DATA THROUGH HIERARCHICAL DIMENSION REDUCTION.

Advances in neuromedicine have emerged from endeavors to elucidate the distinct genetic factors that influence the changes in brain structure that underlie various neurological conditions. We present a framework for examining the extent to which genetic factors impact imaging phenotypes described by voxel-wise measurements organized into collections of functionally relevant regions of interest (ROIs) that span the entire brain. Statistically, the integration of neuroimaging and genetic data is challenging. Because genetic variants are expected to impact different regions of the brain, an appropriate method of inference must simultaneously account for spatial dependence and model uncertainty. Our proposed framework combines feature extraction using generalized principal component analysis to account for inherent short- and long-range structural dependencies with Bayesian model averaging to effectuate variable selection in the presence of multiple genetic variants. The methods are demonstrated on a cocaine dependence study to identify ROIs associated with genetic factors that impact diffusion parameters.