A simulation study published in BMC Medical Research Methodology has identified significant biases in several self-controlled designs used for drug safety assessments. The research, conducted by Masatomo Omiya and Tosiya Sato, evaluates how exposure time trends affect the accuracy of estimators when using electronic healthcare databases. The study focuses on the challenge of assessing associations between drugs and adverse drug reactions, where information on confounding factors is often incomplete. Self-controlled designs are commonly employed to mitigate confounding by time-invariant factors, but the authors demonstrate that changing drug use frequencies over time can distort the resulting estimates.
The primary objective of the simulation was to evaluate the performance of five specific self-controlled designs under various exposure time trends and scenarios. The designs examined were the case-crossover, case-time-control, case-case-time-control, symmetric bidirectional case-crossover, and self-controlled case series.
The simulation incorporated four distinct patterns of exposure time trends to test the resilience of the estimators. These patterns included a constant exposure rate, a linear increase, a cosine function-pattern temporal change, and a mountain-pattern temporal change.
Three specific scenarios were investigated to understand the broader context of these biases. The first scenario examined the impact of exposure time trends alone on the estimators. The second scenario considered situations where adverse drug reaction events restrict the length of the observation periods. The third scenario explored cases where the occurrence of an adverse drug reaction influences subsequent exposures.
The results indicated that the case-crossover design exhibited the most pronounced positive bias in estimates. This bias was particularly evident when the exposure time trend demonstrated a linear increase of more than 20% per year or when a mountain-pattern temporal change was present. In contrast, the symmetric bidirectional case-crossover design showed a negative bias specifically under the cosine function-pattern temporal change.
Other designs, including case-time-control, case-case-time-control, and self-controlled case series, showed little bias from exposure time trends. However, the study noted that case-time-control and case-case-time-control were less precise than the other self-controlled designs.
The study also found that when adverse drug reaction occurrence affected the observation period or subsequent exposure, the estimation performance of symmetric bidirectional case-crossover and self-controlled case series was substantially affected.
The authors conclude that design selection for drug safety assessments should first consider whether adverse drug reaction occurrence may affect the observation period or subsequent exposure. Following this initial assessment, researchers should consider the anticipated exposure time trends and the available data structure.
Subject of Research: Impact of exposure time trends on self-controlled design estimators in drug safety simulation studies
Article Title: Impact of exposure time trends on estimators of self-controlled designs in drug safety assessment using electronic healthcare databases: a simulation study
Article References: Omiya, M., & Sato, T. (2026). Impact of exposure time trends on estimators of self-controlled designs in drug safety assessment using electronic healthcare databases: a simulation study. BMC Medical Research Methodology. https://doi.org/10.1186/s12874-026-03037-2
Image Credits: AI Generated
DOI: 10.1186/s12874-026-03037-2
Keywords: drug safety, simulation study, self-controlled designs, case-crossover, exposure time trends, pharmacoepidemiology, electronic healthcare databases, statistical bias, adverse drug reactions, Impact, exposure, time
News Source: Louis Brooks. (November 9, 2026). Simulation Study Reveals Bias in Self-Controlled Drug Safety Designs. Scienmag.
