Fig. 3.

Estimates from the multivariable CLR plotted against the estimates from the case–control study (Mockenhaupt et al. [7]). The line marks the 45-degree line of equality (note the different scales on the x- and the y-axis). Ideally, the estimates from the multivariable CLR would resemble the estimates from the case–control study; however, some very large discrepancies are apparent. Displayed next to the estimates is the minimum of the two possible discordances between case and reference period of the risk factor. It can be seen that drugs with extreme estimates have only one or zero such discordances, while drugs with at least two discordances have plausible estimates