Twenty-two RCTs (n=2,746) were included.
The Jadad scores ranged from 1 to 5 out of 5. Eleven RCTs scored 3 or more points.
Statistically significant heterogeneity was found when all studies were pooled (I2=37%; p=0.04); no pooled estimate was provided. Visual inspection of the funnel plot suggested publication bias may be present, but none of the statistical tests were statistically significant. None of the patient or study characteristics were significantly associated with the efficacy of NAC. The sensitivity analysis identified 4 small studies that contributed most to the heterogeneity. The same 4 studies were identified by the L’Abbe plot as being outliers. These studies had relatively large increases in creatinine in control groups and relatively large decreases in creatinine in NAC groups.
The cluster analysis identified 2 clusters which showed statistically significantly different treatment effects (p<0.0001); the individual clusters were statistically homogeneous (I2=0% and p>0.5 for both). Cluster 1 comprised 18 studies (89% of patients) and showed no significant difference in CIN between treatment groups (RR 0.87, 95% CI: 0.68, 1.12, p=0.28). Cluster 2 comprised 4 studies (representing 11% of patients) and showed a large reduction in CIN with NAC compared with control (RR 0.15, 95% CI: 0.07, 0.33, p<0.0001). Studies in cluster 2 used low or moderate doses of NAC, were published earlier, were smaller and of lower quality, and control groups reported more CIN compared with cluster 1 groups. In addition, the 4 studies showing a benefit in cluster 2 used half-normal saline infusions.
There was no significant difference between NAC groups and control in the occurrence of dialysis in either of the clusters.