Meta-analytic Techniques to Assess the Association Between N-acetylcysteine and Acute Kidney Injury After Contrast Administration: A Systematic Review and Meta-analysis

IMPORTANCE: The most suitable analytic method to systematically analyze numerous trials with contradictory results is unclear. Multiple trials assessing the use of N-acetylcysteine (NAC) for prevention of contrast-induced acute kidney injury (CI-AKI) have had contradictory results with recent trials confirming a lack of benefit. OBJECTIVE: To systematically review the literature on NAC for the prevention of CI-AKI, and to explore the heterogeneity, publication bias, and small-study effect to determine the most suitable analytic method in a setting where the literature is contradictory. DATA SOURCES: Medline, Embase, and Cochrane Central Register of Controlled Trials databases were used to find randomized clinical trials (RCTs) comparing NAC with any other prophylactic agent or placebo in adults. STUDY SELECTION: The search included studies published in English from database inception to January 2020. Two independent reviewers screened the studies, extracted data, and performed the risk of bias assessment. DATA EXTRACTION AND SYNTHESIS: A meta-analysis was conducted about the effect of NAC on CI-AKI, the need for dialysis, and mortality. Fixed and random effects analyses were also performed. Funnel plots and the trim and fill method were used for assessment of publication bias. Metaregression was performed to explore the heterogeneity and subgroup analysis to examine the association between NAC and CI-AKI when studies were categorized according to sample size and number of events. RESULTS: A total of 101 trials were included in this meta-analysis. The median sample size was 112 (range, 20 to 4993). Twenty-nine trials had a sample size of 200 or more, and only 3 trials had a sample size of 500 or more. Forty-five trials reported the need for kidney replacement therapy, and 41 trials reported mortality as an outcome. NAC seemed to show a benefit, with a pooled OR of 0.72 (95% CI, 0.63-0.82) using random effects model and a pooled OR of 0.82 (95% CI 0.76-0.90) using a fixed effects model. However, there was significant heterogeneity (I(2) = 37.6; P < .001) and significant publication bias, which was reduced only when restricting to large RCTs (N ≥ 500). The clinical outcomes (ie, the need for kidney replacement therapy and mortality) revealed little heterogeneity and no publication bias, and each provided a robust neutral summary result. CONCLUSIONS AND RELEVANCE: In this meta-analysis, NAC was associated with a benefit in the prevention of CI-AKI. However, because of substantial publication bias and other biases, standard meta-analytic techniques resulted in significant heterogeneity and a spurious, or factitious, association, even when using a random effects model. When the analysis was restricted to RCTs with a large sample size to account for publication bias or restricted to trials with clinical outcomes, this issue was reduced and resulted in more robust and neutral effect sizes.