Adverse Drug Reactions in Hospitalized Patients: A Critique of a Meta-analysis

Marion Kvasz, MD, MPH; I. Elaine Allen, PhD; Matthew J. Gordon, BA; Eric Y. Ro, BA; Rhonda Estok, RN; Ingram Olkin, PhD; and Susan D. Ross, MD, FRCPC

In This Article


Our plan was to replicate independently the meta-analysis, following the published methods as closely as possible. We evaluated the meta-analysis methodology in light of appraisal methods for systematic reviews suggested by Oxman and Guyatt.[2] The stated goal of the meta-analysis was to "estimate injuries incurred by drugs that were properly prescribed and administered." For this reason, the meta-analysis authors used the World Health Organization (WHO) definition of ADR, which refers to any noxious, unintended, and undesired effect of a drug that occurs at doses used in humans for prophylaxis, diagnosis, or therapy. This definition excludes therapeutic failures, intentional and accidental poisoning (ie, overdose), and drug abuse. Also, this definition does not include adverse reactions caused by errors in drug administration or noncompliance. For studies in which another definition of ADR was used, reactions caused by errors in administration or noncompliance were to be removed from the results to make the data comparable to WHO ADR definitions for incidence calculations.

A detailed protocol was prepared in advance. This described our approach to replicating and evaluating the meta-analysis in terms of explicit objectives, literature search methods, study eligibility criteria, data elements to be extracted from each eligible paper, and statistical analysis methods. Our protocol questions included the following: (1) Were ADRs defined with sufficient precision and consistency across source studies to permit pooling of study results? (2) Were preventable ADRs reported separately from all adverse events in each study? (3) Were the correct numbers (ie, number of patients with probable/definite ADRs) used for computing ADR incidence rates? (4) In pooling data, were appropriate adjustments made for differences in studies, patients, and outcomes?