reports the database searches might have missed. We also searched ClinicalTrials.gov to identify ongoing studies. We searched for publicly available data held by the U.S. Food and Drug Administration. Study Eligibility Criteria, Participants, and Interventions We followed the population, interventions, comparators, outcomes, timing, and setting (PICOTS) framework in developing the criteria for including studies in the review, and included studies of patients of all ages with low, moderate, or high risk of developing CIN. We included randomized controlled trials (RCTs) of any intervention to prevent CIN (including administration of N-acetylcysteine, sodium bicarbonate solution, sodium chloride solution, statins, adenosine antagonists, diuretics, vasoactive drugs, antioxidants, dopamine, and renal replacement therapy) in which the study groups received either IOCM or LOCM via IV or intraarterial injection. Studies had to report on at least one of the outcomes listed in the Key Question. We included observational studies where available for all comparisons of interest. ES-3 Study Appraisal and Synthesis Methods The titles and abstracts were independently screened by two reviewers. Inclusion at the titlescreening level was liberal; if a single reviewer believed an article might contain relevant information, the article was moved to the abstract level for further screening. When reviewing abstracts followed by the full text of articles, both reviewers had to agree on inclusion or exclusion. Disagreements that could not be resolved by the two reviewers were resolved by a third expert member of the team. At random intervals during screening, senior team members performed quality checks to ensure that eligibility criteria were applied consistently. We performed de novo meta-analyses of all studies on a given comparison if the studies were similar by qualitative or statistical criteria. Pooled risks for large comparisons (18 or more studies) were calculated using a random-effects model using the method of DerSimonian and Laird.6 For comparisons with fewer than 18 studies, we used the Knapp-Hartung small sample estimator approach. This method allows for small sample adjustments to the variance estimates and forms confidence intervals (CI) based on the t distribution with k - 1 degrees of freedom.7 Statistical heterogeneity was assessed using the I-squared statistic. Two reviewers independently assessed each study’s risk of bias using five items from the Cochrane Risk of Bias tool for randomized studies:8 • Was the allocation sequence adequately generated? • Was allocation adequately concealed? • Was knowledge of the allocated intervention adequately prevented during the study? • Were incomplete outcome data adequately addressed? • Are reports of the study free of suggestion of selective outcome reporting? When assessing the risk of bias, we focused on the main outcome of interest, CIN, an outcome that is objectively measured by laboratory testing. Study limitations were determined for each comparison group for CIN and other reported outcomes. Study limitations were determined using the following algorithm for a body of evidence. A body of evidence was assessed as having high study limitations if greater than 50 percent of the studies scored negative in one or more of the criteria. A body of evidence was assessed as having low study limitations if most (51% or greater) of the studies scored positive in all five domains. Bodies of evidence not meeting one of the above criteria were assessed as having medium study limitations. The team graded the strength of evidence (SOE) on comparisons of interest for the key outcomes. We used the grading scheme recommended in the Agency for Healthcare Research and Quality Methods Guide for Effectiveness and Comparative Effectiveness Reviews9 and considered all domains: study limitations, directness, consistency, precision, reporting bias, and magnitude of effect.9 Following the guidance of the GRADE (Grading of Recommendations Assessment, Development and Evaluation) Working Group,10 we rated evidence as precise if the total number of patients exceeded an optimum information size and the 95% (CI) excluded a risk ratio of 1.0. If the total number of patients exceeded the optimum information size and the 95% CI did not exclude the possibility of no difference (i.e., risk ratio of 1.0), we rated the evidence as precise only if the 95% CI excluded the possibility of a clinically important benefit or harm (i.e., risk ratio less than 0.75 or greater than 1.25). For the main outcome of interest, CIN, we used an optimum information size of 2,000 based on an expected 0.1 probability of CIN in the comparison group and a minimally important relative risk difference of 25 percent. For less frequent adverse outcomes, we used an optimum information size of 10,000 based on an expected 0.02 probability in the comparison group and a minimally important relative risk ES-4 difference of 25 percent. If only one study was available for a given comparison, we downgraded the evidence for having unknown consistency. We classified the SOE pertaining to each comparison into four category grades: high, moderate, low, and insufficient. The body of evidence was considered high grade if study limitations were low and there were no problems in any of the other domains, and it was subsequently