# Adjusting for truncated study duration in recurrent event analysis: A weighting approach for clinical trials

## TL;DR

This paper introduces a weighting method to reduce bias in risk estimates caused by early dropout or study termination in clinical trials with recurrent events.

## Contribution

A novel time-based weighting approach is proposed to adjust for truncated follow-up in recurrent event analysis.

## Key findings

- The weighted PWP-GT model showed lower bias (1.0% vs. 1.3%) and improved precision in simulations.
- Weighting reduced standard errors and produced more conservative hazard ratios in real trial data.

## Abstract

In recurrent event analysis with fixed follow-up intervals, truncated follow-up due to early dropout or study termination introduces bias and reduces precision in risk estimates, particularly in clinical trials where shorter observation periods may underestimate event risks.

We propose a time-based weighting approach using the ratio of observed-to-expected follow-up duration in the Prentice-Williams-Peterson Gap Time (PWP-GT) model. The method was evaluated in simulations and applied to a double-blinded trial (N = 4000) comparing 500 mg vs. 1500 mg daily calcium supplementation for preeclampsia prevention. For demonstration of the problem and application of the weighting method, drug non-adherence at follow-up visits was considered as the recurrent event.

Simulations showed the weighted PWP-GT model had lower bias (1.0% vs. 1.3%) and improved precision compared to the unweighted model, with coverage probabilities >94%. In the trial data, weighting yielded smaller standard errors and a more conservative hazard ratio for hypertension family history (weighted HR = 1.14, SE = 0.054 vs. unweighted HR = 1.23, SE = 0.065).

Unaccounted truncated follow-up in recurrent event studies can bias the risk estimation if unaccounted for. Our findings demonstrate that total time-based weighting effectively addresses this bias and enhances precision in both simulated and real datasets.