Data reduction when aggregating information about harms associated with medical interventions

In interpreting and aggregating data in published reports, readers and authors must be aware that some data loss and transformation are inevitable in the process (figure 1).1 Kamp and colleagues recently examined the beneficial and adverse event (AE) profiles of tricyclic antidepressants in a systematic review of available evidence from randomised controlled trials. The authors identified 103 trials randomising 10 590 participants, concluding that in the short term these medications may reduce depressive symptoms (mean difference on the 17-item Hamilton Rating Scale for Depression of –3.77, 95% CIs –5.91 to –1.63; 17 studies; low certainty of evidence) and increase the chances of ‘serious AEs’ (SAEs) (OR 2.78, 95% CI 2.18 to 3.55; 35 trials; very low certainty of evidence) compared with placebo.2

The International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH), the European Medicines Agency, and the Food and…