Methods used to assess the performance of biomarkers for the diagnosis of acute kidney injury: a systematic review and meta-analysis

Purpose: Methods used to explore biomarkers for acute kidney injury (AKI) might have a major impact on the results and the use of these biomarkers. We evaluated the methods used to investigate biomarkers of AKI. Materials and methods: A systematic review and meta-analysis were performed using a computerized search of the MEDLINE and the EMBASE databases (PROSPERO CRD42017059618). Articles reporting biomarker's performance to diagnose AKI were included. The outcome included a description of the methods used to assess the performance of biomarkers to diagnose AKI. Results: Among the 295 included studies, assessment of biomarkers was the primary endpoint in 284 with sample size calculation in only 8% of cases. Eighty-five percent of the studies summarized the performance of biomarkers with receiver operating characteristic (ROC) curves; however, 74 studies (25%) did not provide the threshold, sensibility or specificity. A total of 176 studies evaluated more than one biomarker, and only 25% combined biomarkers to increase diagnostic performance. We determined that the definition of AKI and study design impacted the diagnostic performance using uNGAL (urinary neutrophil gelatinase-associated lipocalin) as an example. Major publication bias was identified. Conclusions: Most articles that reported biomarkers of AKI performance present methodological weaknesses. Basic rules should be provided to increase the quality of reporting in this area.