A Quasi-Experiment To Study the Impact of Vancomycin Area under the Concentration-Time Curve-Guided Dosing on Vancomycin-Associated Nephrotoxicity

Study design and population.This was a retrospective, quasi-experimental study including hospitalized patients receiving intravenous vancomycin at four adult hospitals within DMC. Both the Wayne State University Institutional Review Board and the DMC Research Review Committee approved the study. All patients who received vancomycin for at least 72 h for a documented or suspected infection from January 2014 through December 2015 were eligible for inclusion. In 2014, vancomycin was dosed to target trough concentrations of between 15 and 20 mg/liter in patients with disease states for which it was indicated. In January 2015, the vancomycin dosing guidelines for the treatment of invasive infections were revised to target an AUC₂₄ of 400 to 600 mg · h/liter. Details related to vancomycin dosing during each time period are described below. At least one vancomycin trough concentration obtained near steady state and within the first 96 h of therapy was required for inclusion in the trough concentration-guided dosing group. For inclusion in the AUC-guided dosing group, a minimum of two vancomycin concentrations in samples drawn during the same dosing interval within the first 96 h of therapy were required. Patients with a preexisting need for renal replacement therapy (RRT) or a baseline serum creatinine concentration (SCr) of ≥2 mg/dl were excluded. Patients who received vancomycin to treat meningitis, skin and soft tissue infections without concomitant bacteremia, urinary tract infections, or surgical prophylaxis were excluded, as the current DMC vancomycin monitoring guidelines do not recommend AUC₂₄ targets for these indications. Due to the emerging body of evidence correlating use of the combination of vancomycin and piperacillin-tazobactam with increased nephrotoxicity and a piperacillin-tazobactam shortage, piperacillin-tazobactam use sharply declined during the study period (23, 25, 26). To account for this potential confounding variable, patients receiving concomitant piperacillin-tazobactam were excluded.

Patient data.Patient clinical data, including demographics, comorbidities, medication administration records, laboratory values, physiologic parameters, and indication for vancomycin therapy, were obtained from the electronic medical record by querying the organization's business intelligence software. Comorbid conditions were defined using the codes from the International Classification of Disease, 9th revision (ICD-9), and the International Classification of Disease, 10th revision, Clinical Modification (ICD-10-CM). The degree of patient comorbidity was quantified using the Elixhauser comorbidity index (27). The severity of illness was quantified using the acute physiology and chronic health evaluation II (APACHE II) score and the worst physiological parameters within 24 h of vancomycin initiation (28). The indication for vancomycin therapy was obtained from the electronic antibiotic indication field completed by the prescriber at the time of initial order entry. The dates and times of administration of vancomycin and other potentially nephrotoxic medications were identified from the electronic record of barcoded medication administration. The following medications or medication classes were considered potential nephrotoxins: aminoglycosides, angiotensin-converting enzyme inhibitors (ACEi), angiotensin receptor blockers (ARB), amphotericin B, calcineurin inhibitors, intravenous contrast dye, loop diuretics, polymyxins, and vasopressors. Renal function was assessed by determination of the serum creatinine concentration (SCr) and creatinine clearance (CL_CR), which was estimated by use of the Cockcroft-Gault formula (29).

Vancomycin dosing at the Detroit Medical Center.Prior to 2015, vancomycin was dosed to target serum vancomycin trough concentrations of 15 to 20 mg/liter on the basis of basic pharmacokinetic principles and published pharmacokinetic equations for vancomycin (30, 31). Trough concentrations were measured near steady state (i.e., after 3 to 5 doses), and the dose and frequency were adjusted as necessary. In January 2015, the institutional guidelines for vancomycin dosing for the treatment of invasive infections were revised to target an AUC₂₄ of 400 to 600 mg · h/liter. The upper end of the AUC₂₄ target range was selected on the basis of data suggesting an association between an AUC₂₄ of >600 mg · h/liter and an increased risk of nephrotoxicity (11). In order to calculate patient-specific AUC₂₄ values, two postinfusion serum vancomycin concentrations were measured. AUC₂₄ was calculated by estimating the AUC during the infusion by use of the trapezoidal rule (AUC_inf = C_max + C_min/2 · time of infusion, where AUC_inf is the AUC from time zero to infinity and C_max is the maximum concentration) and the AUC during the elimination phase (AUC_elim) via the logarithmic trapezoidal rule (AUC_elim = C_max − C_min/elimination rate constant). The AUC for a given dose was then calculated by adding AUC_inf and AUC_elim. Finally, AUC₂₄ was calculated by multiplying this number by the number of daily doses.

Outcomes.The primary outcome in this analysis was the comparative rate of acute kidney injury, assessed by three different definitions of vancomycin nephrotoxicity (Table 4). The baseline SCr was defined as the SCr immediately preceding the first dose of vancomycin, if available. In cases in which the sample in which SCr was determined was drawn after the first vancomycin dose, the first SCr immediately after the initial vancomycin dose was considered the baseline SCr. Secondary endpoints were measures of vancomycin exposure, including the total daily vancomycin dose, AUC, and trough concentration.

Data analysis.In the primary analysis, the independent association between the vancomycin dosing strategy and nephrotoxicity, as defined by the vancomycin consensus guidelines, was examined. Patients who developed acute kidney injury were first compared to those who did not in a bivariate analysis using the chi-square/Fisher exact test or the Student t test/Mann-Whitney U test as appropriate. Multivariable logistic regression analysis was then used to evaluate the independent predictors of nephrotoxicity. The variables associated with nephrotoxicity with biological plausibility in bivariate analysis (P < 0.1) were simultaneously entered into candidate regression models and removed individually using a backward elimination procedure until only variables with an adjusted P value of <0.1 remained. Vancomycin exposure variables (total daily dose, AUC, trough concentration) were not candidates for the regression analysis because they lie in the causal pathway between the monitoring strategy and nephrotoxicity. Model fit was assessed with the Hosmer-Lemeshow goodness-of-fit test; models with a nonsignificant result were considered adequate. The multicolinearity of candidate regression models was assessed by use of the variance inflation factor, with values of <3 considered acceptable. The association between the monitoring approach and the time to nephrotoxicity was also assessed via Cox proportional hazards regression, constructed in the same backward elimination approach as the logistic regression. Due to unanticipated differences in the severity of illness between the treatment groups, post hoc matched analyses were also performed with patients matched on APACHE II score ±3.

As a secondary analysis, the vancomycin exposures achieved with the two monitoring approaches were compared. First, the total daily vancomycin dose and clinically measured steady-state trough concentrations were compared between the two groups in the entire cohort. AUC₂₄ values could not be directly calculated for the trough concentration-guided dosing period due to a lack of multiple serum vancomycin concentrations per dose interval. As such, it was not possible to directly compare the calculated AUC₂₄ values for the entire cohort. In order to allow a comparison of AUC exposures between the two dosing strategies, a subgroup analysis was performed using a Bayesian estimation on 300 patients with a bacteremia or pneumonia indication. Patients with these indications from each group were first matched on indication and APACHE II score ±3. One hundred fifty pairs were then randomly selected for comparison. The concentration-time profiles for these patients were estimated via the maximum a posteriori probability (MAP) Bayesian function of the ADAPT V program using a previously published 2-compartment population pharmacokinetic model as the Bayesian prior (32, 33). This approach has been previously validated for estimation of the vancomycin AUC using trough-only serum concentration sampling (3). Day 1 and 2 AUC values (AUC_0–24 and AUC_24–48, respectively) and day 1 and 2 trough concentrations ([C_min24] and [C_min48], respectively) were then compared between groups using the Mann-Whitney U test.

All statistical tests were two-sided, and P values of ≤0.05 were considered statistically significant. Statistical analyses were performed using SPSS (version 24.0) software (SPSS, Armonk, NY).