Vancomycin-Induced Acute Kidney Injury in Critically Ill Adult Patients: a Multicenter Target Trial Emulation Study

Background: Drugs frequently cause acute kidney injury (AKI) in patients admitted to the intensive care unit (ICU). Although vancomycin is often regarded as a potential cause of AKI, this view is highly debated. Two recent meta-analyses identified weak effects. One included seven randomized controlled trials and found moderate quality of evidence, and a second included 60 observational studies of which most suffered from a moderate to high risk of bias.

Objectives: We aimed to estimate the effect of initiating vancomycin versus initiating one of several alternative minimally nephrotoxic antibiotics on 14-day risk of AKI in adult non-dialysis dependent AKI free ICU admissions.

Methods: We applied the target trial emulation framework to design and emulate a hypothetical trial comparing the 14-day risk of AKI between two treatment options: initiation of vancomycin and initiation of an alternative, minimally nephrotoxic antibiotic: clindamycin, linezolid, teicoplanin, meropenem, cefazolin or daptomycin. An observational cohort comprising of ICU admissions to 15 Dutch ICUs between 2010 and 2019 was used to emulate the hypothetical trial. In our emulation we included adult ICU admissions that initiated vancomycin or one of the alternative antibiotics after the first, and before the seventh day of the ICU admission, who were non-dialysis dependent and AKI free. ICU admissions without a baseline serum creatinine (SCr) measurement on the first ICU admission day were excluded. Follow up started at initiation of the treatment and ended at AKI diagnosis or 14 days after initiation, whichever occurred first. ICU admissions were censored at ICU discharge, death in the ICU, the initiation of another treatment option and the initiation of kidney replacement therapy without a preceding AKI. AKI was diagnosed according to the KDIGO SCr criteria. We estimated the hazard ratio (HR) through a marginal structural Cox model, applying inverse probability of treatment weighting to adjust for 35 potential confounders. A robust sandwich variance estimator was used to obtain a valid 95% confidence interval (CI).

Results: The entire cohort consisted of 176,489 ICU admissions, of which 1,809 were included in the emulation. Of these, 887 initiated vancomycin and 922 initiated an alternative antibiotic. AKI incidences in the two groups were 18.8% and 13.6%, respectively. Initiation of vancomycin was not significantly associated with AKI without (HR 1.24, 95% CI 0.91 – 1.70, p = 0.17) or with (HR 1.18, 95% CI 0.83 – 1.67, p = 0.36) adjustment for confounding.

Conclusions: We did not find a significant association between initiation of vancomycin and 14-day risk of AKI, but the relatively wide CIs may indicate a lack of power. Future research may investigate whether the association depends on the vancomycin dose.