Lead Investigator: Pietro Spitali, Leiden University Medical Center
Title of Proposal Research: Prediction of disease progression in Duchenne muscular dystrophy
Vivli Data Request: 5944
Funding Source: None
Potential Conflicts of Interest: None
Summary of the Proposed Research:
Duchenne muscular dystrophy (DMD) is a rare genetic condition caused by mutations in a gene located on the X chromosome. This condition affects primarily males as females have 2 copies of the X chromosome and one functional copy is in most cases able to compensate for mutations present on the other X chromosome. DMD affects young boys, causing delayed motor development, wheelchair dependency (on average between 10 and 12 years of age) and death due to cardio-pulmonary complications.
There is currently no cure for the disease and multiple clinical trials failed due to reduced drug potency but also sub-optimal trial design. Indeed, proper design on interventional studies with drugs has been challenging due to inter-patients variability and the inability to anticipate clinical benefit in short period of time matching the classical duration of clinical trials.
In this project we aim to use a statistical model that we developed, to assess whether we can improve the prediction of disease milestones based on longitudinal data. The model has been used successfully in the cohort of patients followed up in our institution and access to a larger cohort would enable to provide a first layer of validation that we can anticipate disease progression and clinically relevant disease milestones in DMD patients.
Statistical Analysis Plan:
Generalized linear mixed models (GLMM) will be used to model group effects (fixed effects) taking into account individual trajectories (random intercept and slopes). Simpler model (e.g. with no random slopes) will be considered in case of convergence problems. Age at baseline, time on study, study ID, treatment will be used as covariates. The parameter estimates will enable to describe individual trajectories that will then be used as predictors in a time to event analysis where inability to perform certain tasks will be used as events (milestones).
The occurrence of missing values will be evaluated based on type of missingness pattern, which can be completely at random (MCAR), at random (MAR) or not-at-random (MNR). In the first phase with GLMM, missing values will be managed directly by the mixed models, which can handle unbalanced designs with measurements at irregular time points. In the second phase (time to event analysis) the presence of censored survival outcomes will be modelled using a penalized Cox model.
A Prospective Natural History Study of Progression of Physical Impairment, Activity Limitation and Quality of Life in Duchenne Muscular Dystrophy
Study ID: NCT01753804
Sponsor ID: PRO-DMD-01
An Exploratory Study to Assess Two Doses of GSK2402968 in the Treatment of Ambulant Boys With Duchenne Muscular Dystrophy
Study ID: NCT01462292
Sponsor ID: 114876
A Phase III, Randomized, Double Blind, Placebo-controlled Clinical Study to Assess the Efficacy and Safety of GSK2402968 in Subjects With Duchenne Muscular Dystrophy
Study ID: NCT01254019
Sponsor ID: 114044
A Phase II, Double Blind, Exploratory, Parallel-group, Placebocontrolled Clinical Study to Assess Two Dosing Regimens of GSK2402968 for Efficacy, Safety, Tolerability and Pharmacokinetics in Ambulant Subjects With Duchenne Muscular Dystrophy
Study ID: NCT01153932
Sponsor ID: 114117
Summary of Results:
We have used the data that you have kindly shared to assess the prediction accuracy of disease milestone in patients with Duchenne muscular dystrophy using a method that developed called penalized regression calibration (pencal).
Analysis of the data was completed, and I have uploaded the report in September 2021. The prediction accuracy was only slightly improved by pencal. This small improvement does not warrant combining the results into a manuscript. Please note that results were only discussed internally and not presented in meeting procedures.
It is possible that the small samples size, the low number of events or the short time on study did not allow to identify significant improvement in the prediction accuracy. Therefore, we consider the analysis of the data complete.