Recommendation	Rationale
Statistical interaction measures with additive baseline models should be preferred over those with multiplicative baseline models for detecting signals of DDIs in spontaneous report databases	Statistical interaction measures with additive baseline models provided better sensitivity and equal or better specificity for both established and emerging DDIs [55]
Future research should explore how statistical interaction measures with additive baseline models can best be incorporated in broader predictive models of adverse drug interactions, and in routine signal detection	This was out of scope for the PROTECT study, but recent research has found that predictive models accounting for multiple aspects of strength of evidence perform better than statistical measure of interaction alone [54]