- This paper describes the innovative statistical model used for primary analysis of efficacy of investigational drugs being tested in the HEALEY ALS Platform Trial, a randomized, placebo-controlled trial evaluating multiple investigational drugs for ALS in parallel and sequentially with sharing of control data among regimens
- The primary analysis method relies on an innovative Bayesian shared-parameter model that provides an integrated estimate of how well a given treatment slows disease progression as measured by both function and survival
- Potential differences in the shared control group are accounted for with Bayesian hierarchical modeling
- Clinical trial simulations showed that the model has good power and is robust to variation across regimens and between effects on function and survival
The Healey ALS Platform Trial, headquartered at Massachusetts General Hospital, is speeding the evaluation of investigational treatments for ALS by allowing the study of multiple drugs in parallel and sequentially.
Subscribe to the latest updates from Neuroscience Advances in Motion
The trial has operational and statistical efficiencies compared with a traditional randomized clinical trial, which is especially important for a disease with a dramatic unmet medical need and a large potential product pipeline. However, the complexities of the trial have required innovative statistical approaches.
Sabrina Paganoni, MD, PhD, of the Sean M. Healey & AMG Center for ALS, Eric A. Macklin, PhD, assistant in biostatistics at Mass General and a biostatistician at Harvard Medical School, and colleagues provide details in Annals of Neurology.
The Healey ALS Platform Trial is a multicenter, randomized, placebo-controlled, adaptive, perpetual platform trial. As previously described in Annals of Neurology, a master protocol specifies the global trial design. Regimen-specific appendices specify the uniquely tailored designs used to evaluate each investigational product.
Participants are screened and randomized in two stages: regimen and treatment group. Participants and investigators know the regimen to which a participant was assigned but are blind to whether the participant is receiving active treatment or placebo.
Shared Control Data
In the Healey ALS Platform Trial, primary analyses of each regimen incorporate data on all participants randomized to the active treatment arm within that regimen, all participants concurrently randomized to the control arm within that regimen, and control participants in other regimens. The sharing of control data increases statistical efficiency and reduces the need to randomize patients to placebo.
However, analyses involving shared controls require careful consideration to avoid bias. Regimens may differ for enrollment time ("time-trend effects") or mode of drug administration and eligibility criteria related to safety ("regimen-specific effects"). The statistical methodology of the platform trial, explained below, addresses these concerns.
Food and Drug Administration guidance states primary analyses in trials of ALS treatments should integrate measures of function and survival. However, the statistical approaches mentioned in the guidance don't accommodate the complexities of shared control data and don't provide a clinically interpretable or meaningful estimate of treatment effect.
The Healey ALS Platform Trial group developed an innovative method for primary analysis, a Bayesian shared parameter model that provides an integrated estimate of treatment effect on both function and survival. The functional component is a repeated measures model of total score on the Revised ALS Functional Rating Scale from baseline through week 24.
Within the Bayesian framework, the likelihoods of both the functional outcome and survival are used to estimate slowing in the rate of disease progression, called the disease rate ratio (DRR). The DRR is akin to a hazard ratio in proportional hazards survival analysis and has the same meaningful clinical interpretation.
For example, DRR of 0.5 indicates 50% slowing in disease progression, where a treated participant would be expected to progress to the same state as a control participant in double the amount of time.
To allow dynamic sharing of control data across regimens, a Bayesian hierarchical modeling framework considers regimen-specific effects and time-trend effects.
Clinical Trial Simulations
In a trial as complex and novel as the Healey ALS Platform Trial, no simple mathematical formula can capture the operating characteristics of the design and primary analysis.
Instead, the research group used the Bayesian model to conduct clinical trial simulations for optimization of the inclusion/exclusion criteria, sample size, length of follow-up, baseline covariates and adaptive rules for stopping early for futility or success.
Additional simulations have confirmed that the innovative statistical analysis methods integrate function and survival and account for differences in shared controls across regimens.
Learn more about the Sean M. Healey & AMG Center for ALS
Refer a patient to the Sean M. Healey & AMG Center for ALS