effects model for low-dose N-acetylcysteine (1200 mg/day or less) was 0.75 (95% CI: 0.63 to 0.89), indicating that, on average, the small effect is consistent with a borderline clinically important reduction in CIN (Figure 4). The statistical heterogeneity of the studies was low, with an I-squared of 0%. The pooled risk ratio using the Knapp-Hartung method for the studies using IV contrast media and low-dose Nacetylcysteine was 0.62, but in this small subset of five studies, the confidence interval was so wide that we cannot rule out a clinically important increased risk (95% CI: 0.18 to 2.10). For studies using intra-arterially administered contrast media and low-dose N-acetylcysteine, the pooled risk ratio was 0.77 (95% CI: 0.66 to 0.91) indicating that, on average, the benefit is at a level consistent with a clinically unimportant reduction in CIN. Using Harbord’s modified test for small study effects, we did not find evidence of asymmetry in results by study precision (bias coefficient of -0.70, standard error of 0.44, p=0.123). Overall, the strength of evidence was low that low-dose N-acetylcysteine with IV saline had a small clinically unimportant effect in preventing CIN compared with IV saline with or without a placebo (Table 3; see Appendixes F and G for study limitations). We performed stratification analyses to investigate the influence of contrast media osmolality on the effect of N-acetylcysteine. The pooled risk ratio of CIN, using a random effects model, for studies using LOCM was 0.69 (95% CI: 0.58 to 0.84), indicating that, on average, the difference is consistent with a clinically important reduction in CIN with N-acetylcysteine in patients receiving LOCM, but the confidence interval does not rule out a clinically unimportant difference (Figure 5).The statistical heterogeneity across studies was low, with an I-squared of 19 percent. The strength of the evidence was moderate that in patients receiving LOCM, Nacetylcysteine with IV saline had a clinically important reduction in CIN. The pooled risk ratio 17 for CIN from studies of N-acetylcysteine using IOCM was 1.12 (95% CI: 0.74 to 1.69). The confidence interval was wide enough for N-acetylcysteine when IOCM was used to suggest possible harm without any indication of a clinically important benefit (Figure 6). The strength of the evidence was low that in patients receiving IOCM, N-acetylcysteine with IV saline did not have a clinically important decrease in CIN. The estimates of effect are remarkably stable across different types of studies with a 20 to 30 percent reduction, which is near the edge of what we defined to be a minimally important difference. The variation is mainly in the CIs, which is likely due to variation in the number of people in the different studies. We also performed stratification analyses to investigate the influence of the route of Nacetylcysteine administration. The pooled risk ratio for CIN, using a random effects model, for patients who received oral N-acetylcysteine was 0.77 (95% CI: 0.65 to 0.92), indicating that, on average, the difference is not clinically important. The pooled risk ratio for CIN for patients who received IV N-acetylcysteine (run with the Knapp-Hartung method) was 0.90 (95% CI: 0.72 to 1.12), indicating that the difference is not clinically important (Figure 7). Our sensitivity analysis, which removed one study at a time, did not show any significant impact on the estimated effect of N-acetylcysteine. When we examined the variation of risk ratio estimates according to baseline characteristics of the study population, we did not observe any meaningful difference by age, sex, baseline renal function, or the presence or absence of diabetes mellitus. There was no trend in the effect size by year of the study publication (Figure 7). When we examined how the results of studies of N-acetylcysteine varied in forest plots organized by the number of study limitations, we did not see any pattern indicative of a trend by study quality. Thirteen of the 67 RCTs reporting on CIN were not included in the meta-analyses for a variety of reasons, including missing data, dosage differences, and inclusion criteria differences (see Appendix E, Evidence Table E-5).67,82-90 In addition to the studies that reported on the incidence of CIN, three studies reported on changes in serum creatinine (Appendix E, Evidence Table E-6) and/or glomerular filtration rate (Appendix E, Evidence Table E-7) without reporting the incidence of CIN.91-93 In those nine studies, the mean change in serum creatinine or glomerular filtration rate did not differ enough between groups to meet the definition of CIN. Eleven observational studies were included in the studies we reviewed.94-104The results of the observational studies were similar to those reported in the RCTs. Other Outcomes Of the 77 studies investigating development of CIN when comparing N-acetylcysteine plus IV saline with a placebo with or without IV saline, 35 also included data on secondary outcomes. Twenty eight reported patients’ needs for renal replacement therapy,2 Of the 20 studies that examined the need for