Nine RCTs (n=849) were included in the review. Three trials used frusemide to prevent acute renal failure (325 enrolled) and six used frusemide to treat acute renal failure (623 enrolled).
Four trials received a Jadad score of 1, two trials received a score of 5, and the remaining three trials received scores of 4, 3 and 2. Four trials were judged to have adequate allocation concealment.
No significant reduction in in-hospital mortality (RR 1.11, 95% CI: 0.92, 1.33) was shown when frusemide was compared with placebo. When subgrouped by treatment trials (RR 1.09, 95% CI: 0.90, 1.31; based on 4 trials, n=574) and prevention trials (RR 2.33, 95% CI: 0.75, 7.25; based on 2 trials, n=202), frusemide was not shown to reduce in-hospital mortality in comparison with placebo.
No significant reduction in the requirement for renal replacement therapy (RR 0.99, 95% CI: 0.80, 1.22) was shown when frusemide was compared with placebo. Statistically significant heterogeneity was found because of differences in the treatment trials (P<0.0001). When subgrouped by treatment trials (RR 0.94, 95% CI: 0.71, 1.26; based on 4 trials, n=204) and prevention trials (RR 4.08, 95% CI: 0.46, 35.96; based on 3 trials, n=255), frusemide was not shown to reduce the requirement for renal replacement therapy in comparison with placebo.
No significant difference in the number of dialysis sessions required (WMD -0.48, 95% CI: -1.45, 0.50), or the proportion of participants with persistent oliguria (RR 0.54, 95% CI: 0.18, 1.61), was shown between frusemide and placebo. Statistically significant heterogeneity was found in the latter analysis. Frusemide was associated with an increase in hospital stay (WMD 3.57 days, 95% CI: 0.02, 7.12; P=0.049), based on 2 preventive studies. High-dose frusemide (1 to 3.4 g/day) was associated with an increased risk of deafness or tinnitus being reported (RR 3.97, 95% CI: 1.00, 15.78; P=0.05); no statistical heterogeneity was found.
Sensitivity analyses, including only studies with adequate allocation concealment or excluding the study using a single bolus of frusemide, did not change the magnitude or direction of the results obtained.
The absence of small studies showing a reduction in mortality after treatment with frusemide indicated a small possibility of publication bias.