[The principles of trend-in-trend design and its applications in real-world causal inference]

The trend-in-trend (TIT) design is an emerging hybrid ecological- epidemiological research method. Based on cumulative exposure probability stratification and cross-stratum temporal trend differentiation, TIT compares differences in outcome temporal slopes across populations with distinct exposure rates. When its core assumptions are met, TIT effectively controls unmeasured confounding by indication and time-varying confounding, thereby inferring causal associations between exposure and outcomes and yielding more robust effect estimates than traditional cohort studies. This article systematically elaborates the statistical principles and core assumptions of the TIT design, compares its methodological characteristics with quasi-experimental methods including difference-in-differences and interrupted time series, and illustrates its implementation procedures through a case study evaluating the cardiovascular risk of rofecoxib. As a reliable methodological tool for real-world study, TIT is particularly applicable to post-marketing drug re-evaluation and scenarios with strong temporal trends and unmeasured confounding.