Fig. 1

The case-crossover study design. The self-controlled design compares, among individuals experiencing an outcome, the number of individuals exposed to NSAID use at the outcome date, but not at any reference date, with the number of individuals exposed at a reference date, but not at the outcome date. Patient #1 is exposed at the second reference point from the outcome. Patient #2 is exposed at the outcome date. Patient #3 is not exposed at any reference day or outcome date and is dropped from the analysis. Using Mantel–Haenszel method, the odds ratios of being exposed on the outcome date vs. the reference date is calculated by dividing the first pattern with the second pattern