Power and sample size calculation for non-inferiority trials with treatment switching in intention-to-treat analysis comparing RMSTs
Year: 2025
Type: article
Abstract: BACKGROUND: Difference in Restricted Mean Survival Time (DRMST) has attracted attention and is increasingly used in non-inferiority (NI) trials because of its superior power in detecting treatment effects compared to hazard ratio. However, when treatment switching (also known as crossover) occurs, the widely used intention-to-treat (ITT) analysis can underpower or overpower NI trials. METHODS: We propose a simulation-based approach, named nifts, to calculate powers and determine the necessary sample size to achieve a desired power for non-inferiority trials that allow treatment switching, in ITT analysis using DRMST. RESULTS: The nifts approach offers three options for a non-inferiority margin, assumes three entry patterns and generalized gamma distributions for event time, incorporates two distributions for dropout censoring, and provides five distribution options for switching. Real-world and simulated examples are used to illustrate the proposed method and examine how switching probability, switching time, the relative effectiveness of treatments, allocation ratio, entry patterns, and event time distribution influence powers and sample sizes. nifts adjusts the non-inferiority margins intended for NI trials without treatment switching to accommodate the presence of treatment switching in the designs. With the adjusted margins, the type I errors are well-controlled. The ratios of sample sizes with treatment switching to those without switching are close to 1, indicating no significant change in power at sample sizes without switching when using adjusted margins. The performance on power and sample sizes is not sensitive to the choice of switch time distributions. CONCLUSIONS: This simulation-based approach provides power and sample size calculation in NI trials with treatment switching, when comparing the RMSTs of two treatment groups in ITT analysis. With its comprehensive parameter settings, nifts will be useful for designing NI trials that allow for treatment switching. nifts is freely available at https://github.com/cyhsuTN/nifts .
Source: BMC Medical Research Methodology
Authors: Austin Shih, Chih-Yuan Hsu, Yu Shyr
Institutions: Vanderbilt University, Vanderbilt University Medical Center
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $3,090
OpenAlex ID: W4411111602