Robust analyses for radiographic progression in rheumatoid arthritis

Area of Science:

• Rheumatology
• Biostatistics
• Clinical Trials

Background:

• Assessing joint structural damage via radiographic progression (e.g., van der Heijde modified Total Sharp Score, mTSS) is crucial in rheumatoid arthritis (RA) trials.
• The analysis of covariance with linear extrapolation (ANCOVA+LE) is a common statistical method for analyzing mTSS, but may be suboptimal for studies with missing data.

Purpose of the Study:

• To compare the performance of the random coefficient (RC) model against ANCOVA+LE and ANCOVA with last observation carried forward (ANCOVA+LOCF) for analyzing mTSS in simulated RA trial data.
• To evaluate statistical models based on bias, root mean square error (RMSE), statistical power, and type I error rates under various missing data scenarios.

Main Methods:

• Simulations were conducted in a two-arm (active treatment vs. placebo) setting over 44 weeks.
• The study compared the RC model, ANCOVA+LE, and ANCOVA+LOCF across different missing data scenarios.

Main Results:

• The RC model demonstrated superior performance over ANCOVA+LE in terms of bias, RMSE, power, and type I error rate.
• When data were missing, the RC model with observed data (RC+OBS) showed comparable or better power and bias reduction than ANCOVA+LE.
• Both ANCOVA and RC models performed similarly when no data were missing.

Conclusions:

• The RC model is a more sensitive and precise alternative to ANCOVA+LE for analyzing mTSS in long-term extension and pediatric RA studies, particularly those with a high likelihood of missing data.
• The RC model can estimate mTSS changes at time points with missing data by calculating a slope.
• ANCOVA+LE is recommended for sensitivity analyses.