Lead Investigator: Kelley Kidwell, University of Michigan
Title of Proposal Research: Use of external control with longitudinal data for drug evaluation in Duchenne muscular dystrophy a Bayesian small sample, sequential, multiple assignment randomized trial design (snSMART)
Vivli Data Request: 8445
Funding Source: None
Potential Conflicts of Interest: None
Summary of the Proposed Research:
Duchenne muscular dystrophy (DMD) is a rare, deadly inherited genetic disease with a prevalence of 19.8 per 100,000 live male births. Patients progressively lose the ability to walk or function independently and often die young from lung or heart problems. In DMD and other rare diseases, it is challenging to recruit patients to participate in clinical trials. Moreover, assigning patients to long-term, multi-year placebo arms can cause ethical issues. Use of the natural history or previous trial placebo data to complement or enrich the current trial placebo information is appealing to the researchers. We propose a small sample, sequential, multiple assignment, randomized trial (snSMART) design that integrates natural history data into the placebo arm. The proposed approach is a multi-stage design evaluating multiple doses of a promising drug to placebo. In stage 1, participants are randomized in greater proportion to receive low dose or high dose over placebo. In stage 2, participants are re-randomized across treatments depending on their stage 1 dose and response. We assume outcome data (e.g. North Star Ambulatory Assessment (NSAA) total score – a 17-item rating scale that is used to measure functional motor abilities in ambulant children with DMD, or six minute walk distance) is captured longitudinally (>2 times) throughout the SMART design and propose Bayesian methods to use all available data for efficient estimation of treatment effects in a snSMART design ((Bayesian methods uses Bayes’ theorem to conduct data analysis. The estimations of parameters are usually determined by the prior distribution of the parameters and the observed data). The framework is flexible enough to include external placebo information from the natural history studies. We illustrate our methods in the setting of DMD. We anticipate the proposed design and methods using longitudinal data are promising alternative tools for drug development in DMD and other rare diseases.
Statistical Analysis Plan:
In our project, we will use data from one or more studies to explore the potential biomarkers, which can predict therapeutic outcomes in UC. Further, we will use data from other studies to validate the biomarker’s predictive ability. In this step, we will perform meta-analysis to pool the odds ratio/hazard ratio of the biomarkers from different studies. We pooled effect estimates for each predictor through inverse variance analysis using random effects meta-analysis. Statistical heterogeneity as assessed with the I2 statistic, whose values higher than 50% indicate a substantial level of heterogeneity.
Continuous and categorical variables are described as median (interquartile range, IQR) and proportion (percentage), respectively. The Mann-Whitney test and χ2 test were performed to evaluate the difference for continuous and categorical variables, respectively. A p-value less than 0.05 was considered as statistical significance. Univariate logistic or cox regression analysis will be conducted to analyse the association of candidate predictors and outcomes. Multivariate logistic or cox regression analysis will be performed to adjusted potential confounders (like disease duration, treatment allocation). The receiver operating characteristic (ROC) analysis is performed to calculate the area under ROC curve (AUROC). The cut-off value is determined by the Youden index. AUROC, sensitivity, specificity, positive predictive value and negative predictive value are used to assess the predictive capacity of the predictors for predicting specific outcomes. Furthermore, we will perform subgroup analyses by treatment allocation, disease activity at baseline, age and gender. Missing value for major outcome will be excluded from statistical analysis. Missing values for other variables will be imputed by simple imputation, using the mice package in R.
Taking account of differences among studies, we will (1) compare baseline characteristics, like age, sex, race, disease duration and disease activity, among different trails. If some covariates are found different among trials, we will adjust them in the multivariate regression model; (2) perform subgroup analysis stratified by different trials to assess the interaction between predictors and trials; (3) conduct sensitivity analyses in patients with same disease activity, range of age, disease duration or type of biologics (anti-TNF agents, vedolizumab or ustekinumab) to verify the consistency of our results.
Requested Studies:
A Prospective Natural History Study of Progression of Physical Impairment, Activity Limitation and Quality of Life in Duchenne Muscular Dystrophy.
Data Contributor: Cure Duchenne
Study ID: NCT01753804
Sponsor ID: PRO-DMD-01
A Study of Tadalafil for Duchenne Muscular Dystrophy
Data Contributor: Lilly
Study ID: H6D-MC-LVJJ
Sponsor ID: H6D-MC-LVJJ
Public Disclosures:
Zheng, J., Zhang, X., Zhang, L., Li, L., Chen, M., Chen, R. and Zhang, S., 2024. Serum Albumin and Its Trajectory Are Associated with Therapeutic Outcomes in Ulcerative Colitis. Clinical Gastroenterology and Hepatology. Doi: 10.1016/j.cgh.2024.10.036