Genotype-unmatched controls are feasible for drug development in Duchenne Muscular Dystrophy (DMD)

Topic:

Clinical Trials

Poster Number: 94

Author(s):

James Signorovitch, PhD, Analysis Group, Inc., Francesco Muntoni, MD, UCL Institute of Child Health and Great Ormond Street Hospital for Children, Molly Frean, PhD, Analysis Group, Inc., Mirko Fillbrunn, PhD, Analysis Group, Inc., Gautam Sajeev, ScD, Analysis Group Inc., Susan Ward, PhD, cTAP, Craig McDonald, MD, University of California, Davis, Nathalie Goemans, MD, PhD, University Hospitals Leuven, Erik Niks, MD, Leiden University Medical Center, Brenda Wong, MD, University of Massachusetts Medical School, Laurent Servais, MD, MDUK, Oxford, UK, Volker Straub, PhD, Newcastle University, Imelda JM de Groot, MD, Radboud University Nijmegen Medical Center, Mary Chesshyre, Dubowitz Neuromuscular Centre, Cuixia Tian, MD, CCHMC, Adnan Manzur, FRCPCH, Dubowitz Neuromuscular Centre, Eugenio Mercuri, MD, PhD, Department of Paediatric Neurology and Nemo Clinical Centre, Catholic University, Annemieke Aartsma-Rus, PhD, Leiden University Medical Center

Clinical trials of genotype-targeted therapies in DMD typically compare treated patients to genotype-matched controls. While this avoids confounding by genotype effects, the pool of eligible controls is further limited in an already rare disease. We hypothesized that using genotype-unmatched controls can be feasible when 1) DMD genotype class effects are small and precisely known and 2) stronger baseline prognostic factors are accounted for. We evaluated this prospect in a hypothetical clinical trial for effects on 1-year change in the North Star Ambulatory Assessment total score (NSAA). The hypothetical ‘treatment’ group (n=58) was exon 51 skip-amenable patients from the DEMAND III trial placebo arm. The genotype-unmatched ‘control’ group (n=58) included patients amenable to skipping exons other than 44, 45, 51 or 53 drawn from five natural history studies using similar age, function, and steroid inclusion criteria. Mean difference in NSAA for the treatment vs. control group indicated a bias of -1.6 NSAA units. After statistically accounting for genotype differences and baseline prognostic factors (e.g., motor function and steroid type), the bias was reduced to -0.3 units. As a statistical cost of genotype adjustment, the standard error of the treatment effect increased by 35%. Power analysis indicated that detecting a 5 (or 4) unit treatment effect using unmatched vs. matched controls would require 1.2 (or 1.4) times more enrolled patients. When genotype adjustment was integrated into a hypothetical multi-genotype, randomized, double-blind platform trial, overall sample size requirements were reduced vs. conducting separate trials with matched controls for each genotype class. We conclude that comparisons with genotypically-unmatched controls, with careful consideration of baseline prognostic factors, are feasible in DMD drug development and could help limit placebo exposure among patients eligible for genotype-targeted treatments.