Comparison of Finkelstein’s Method With the Conventional Approach for Interval-Censored Data Analysis

Interval-censored time-to-event data occurs frequently in randomized clinical trials. Many new methods for analysis of interval-censored time-to-event data have been proposed in the last two decades. However, most of these methods either rely on assumptions that are hard to verify in practice or are computationally challenging. As a result, none of them has been accepted by the pharmaceutical industry as a standard method. Taking advantage of modern computational power, we did an extensive simulation to compare the performance of conventional imputation-based analysis methods for such data with an existing nonparametric interval-censored data analysis method by Finkelstein (1986) in a typical setting for confirmatory clinical trials. The simulation results clearly favor the nonparametric method, which provides credible inference even under extreme conditions. This helps ease a key concern of the regulatory agency and paves the way for wide acceptance of interval-censored clinical endpoint as a primary endpoint in confirmatory trials. An approximate formula for event size calculation is provided for the design of clinical trials with such an endpoint.