Modeling lesion counts in multiple sclerosis when patients have been selected for baseline activity.

Area of Science:

• Neurology
• Biostatistics
• Clinical Trials

Background:

• Gadolinium-enhancing lesions on MRI are key outcomes in multiple sclerosis (MS) trials.
• Participant selection based on baseline lesion activity is common in MS studies to enhance statistical power.
• The impact of this screening on statistical inferences, particularly parameter estimation and interval coverage, remains underexplored.

Purpose of the Study:

• To evaluate the performance of the negative binomial distribution for modeling MS lesion counts in patients selected for baseline activity.
• To investigate how screening for baseline activity influences parameter estimation and interval coverage in MS clinical trials.

Main Methods:

• Computer simulations were employed to assess the effects of screening on inferences from a negative binomial model.
• The study analyzed treatment effects in two independent samples under screened and unscreened conditions.
• Methods for incorporating screening properties into trial design were demonstrated.

Main Results:

• Screening for baseline activity improves point estimation accuracy for lesion counts modeled with a negative binomial distribution.
• However, this screening practice can lead to decreased interval coverage, potentially underestimating true treatment effects.
• Screening also demonstrated a tendency to inflate the Type I error rate, increasing the risk of false positive findings.

Conclusions:

• While screening for active lesions in MS trials can enhance cost-effectiveness, it presents a trade-off with statistical validity.
• The practice may increase the false positive rate, necessitating careful consideration during the planning of Phase II trials.
• Optimizing trial design requires balancing the benefits of screening with its potential impact on statistical inference and error rates.