[Case-crossover design: Basic essentials and applications]

Case-crossover analysis is an observational epidemiological design that was proposed by Maclure in 1991 to assess whether a given intermittent or unusual exposure may have triggered an immediate short-term, acute event. The present article outlines the basics of case-crossover designs, as well as their applications and limitations. The case-crossover design is based on exclusively selecting case subjects. To calculate relative risk, exposure during the period of time prior to the event (case period) is compared against the same subject's exposure during one or more control periods. This method is only appropriate when the exposures are transient in time and have acute short-term effects. For exposures in which there is no trend, a unidirectional approach is the most frequent and consists of selecting one or more control periods prior to the case period. When the exposure displays a time trend (e.g., air pollution), a unidirectional approach will yield biased estimates, and therefore bidirectional case-crossover designs are used, which select control time intervals preceding and subsequent to that of the event. The case-crossover design is being increasingly used across a wide range of fields, including factors triggering traffic, occupational and domestic accidents and acute myocardial infarction, and those involved in air pollution and health and pharmacoepidemiology, among others. Insofar as data-analysis is concerned, case-crossover designs can generally be regarded as matched case-control studies and consequently conditional logistic regression can be applied. Lastly, this study analyzes practical examples of distinct applications of the case-crossover design.