Incidence Evaluation of Acute Kidney Injury and Pharmacoeconomic Analysis Among Critically Ill Patients With Use of Antipseudomonal β-Lactams and Vancomycin

Background: Whether vancomycin (VAN) plus piperacillin-tazobactam (PTZ) could increase the risk of acute kidney injury (AKI) is still controversial in critically ill patients. The purpose of this study was to compare the risk of developing AKI and risk of developing AKI and treatment cost among this population receiving VAN/PTZ to a matched group receiving VAN/other antipseudomonal β-lactams. Methods: This multicenter, retrospective, matched study included 700 critically ill patients who received ≥ 48 hours of VAN/PTZ or VAN/other antipseudomonal β-lactams. The risk of developing AKI was compared between these two combination therapies using propensity-adjusted analysis. Furthermore, a pharmacoeconomic decision-analytic model was performed. Results: According to three AKI-dened criteria, VAN/PTZ was associated with signicantly higher incidence of than VAN/other antipseudomonal β-lactams (all P < 0.001). In multivariate analysis, regardless of any VAN/other antipseudomonal β-lactams, VAN/PTZ was an independent predictor for stage 2 or 3 AKI. In the empiric treatment, the incremental cost-effectiveness ratios per additional nephrotoxic episode of 1147.35$, 1845.11$, and 3989.95$ were found for VAN/PTZ relative to, vancomycin plus imipenem-cilastatin, vancomycin plus meropenem, and vancomycin plus cefoperazone-sulbactam, respectively. Conclusion: In critically ill patients, VAN/PTZ was associated with both higher AKI risk and treatment cost when considering AKI occurence compared to VAN/other antipseudomonal β-lactams. Trial retrospectively registered, ClinicalTrials.gov number: NCT03776409.


Introduction
Critically ill patients usually present with bacteremia and with hospital-acquired infections, thus empirical antimicrobial therapy is inevitable. Vancomycin in combination with antipseudomonal β-lactams (VAN/BL) is a common method in clinical practice (1).
Traditionally, hospitals have selected piperacillin-tazobactam (PTZ) or other β-lactam as the "workhorse" antipseudomonal antibiotics based on institutional susceptibility trends, acquisition costs, and other prescription considerations (2,3). Nevertheless, recent studies show that an increased risk of acute kidney injury (AKI) is found for vancomycin plus piperacillin-tazobactam (VAN/PTZ) (4)(5)(6)(7). Moreover, AKI induced by VAN/PTZ may lead to an increase in mortality, hospital duration, and treatment costs. Although current literature supports the association of AKI in non-critically ill patients, limited studies of critically ill patients have failed to demonstrate that VAN/PTZ increases the risk of AKI. Only one study has shown such association in critically ill patients (8). However, this large sample study includes severe renal insu ciency patients (baseline estimated glomerular ltration rate [eGFR] level of ≤ 30 mL/min), which might increase the bias of identifying the nephrotoxicity of treatment. Another study that evaluated the brief empiric use of VAN/PTZ therapy (< 72 hours) suggests that the use of VAN/PTZ therapy in critically ill patients does not confer a risk of AKI (5). Nevertheless, previous studies showed that duration of VAN/PTZ usually exceeds 72 hours and the onset of drug-induced AKI typically occurs after about 4-8 days of therapy in clinical practice (9)(10)(11). Therefore, it is necessary to evaluate the the safety of combination therapy on combination duration of ≥ 48 hours.
Given the controversy results of previous studies, this study is designed to explore whether the critically ill patients receiving VAN/PTZ have a greater risk of AKI compared with patients receiving VAN/BL. A pharmacoeconomic analysis was to investigate the difference among VAN/PTZ and VAN plus cefoperazonesulbactam (VAN/CPZ-SBT) or VAN plus meropenem (VAN/MEM) or VAN plus imipenem-cisastatin (VAN/IPM-CIS) of treatment costs in this population using clinical safety data generated by the cohort study, epidemiological data, and local medical data.

Study design
This was a multicenter, retrospective study. Eligible patients were adults hospitalized in any of the ICUs at 16 medical centers between January 1 2008 and December 31 2018. For patients admitted on multiple times during the study time, the last admission was used. This study was approved by all medical centers with a waiver of informed consent. This study has been registered with the Clinical Trials Registration Center (ClinicalTrials.gov) (NCT03776409).

Data Collection
Data were collected using an electronic recording form using the Research Electronic Data Capture (REDCap) software (12), which was developed by the Department of Medicine of the University of Padova, Italy (see details of study oversight in Additional le). Data were extracted from electronic records included demographic information, comorbidities, APACHE II score, SOFA score, mechanical ventilation, infection types, AKI risk scores, and microbiological data. Antibiotic therapy covariates were collected including dose and duration of treatment (see details of data collection in Additional le).

Outcomes
The primary outcomes were the incidence of any AKI de ned by the three de nitions (the Acute Kidney Injury Network [AKIN] criteria (13), the Risk-Injury-Failure-Loss-End Stage Renal Disease [RIFLE] criteria (14) and vancomycin consensus guideline (15)) and the incidence of stage 2 or 3 AKI de ned by the AKIN criteria. The occurrence of AKI was assessed beginning 24 hours after initiation of combination therapy. Follow-up continued until 7 days after 1 or both antibiotics in the combination were discontinued (see Additional le

Statistical analysis
For the primary outcome of AKI de ned by the AKIN criteria, we determined a sample size of at least 149 patients in each group (VAN/PTZ vs VAN/BL) to achieve a statistical power of 80% using an α of 0.05 based on a previous study (8) which reported the incidence of VAN/PTZ and VAN/BL (39.3% vs 24.2%).
For categorical data, the Pearson χ 2 test or Fisher's exact test was used. For continuous data, the Student's ttest and Wilcoxon rank-sum test were used. To control for non-randomized potential bias, a propensity scorematched analysis was conducted. Patients were divided into two combination therapy groups (VAN/PTZ and VAN/BL groups). A 1:3 ratio with a caliper length of 0.2 SDs was used to match the patients in the VAN/PTZ group and those in VAN/BL group. The matching was performed based on covariates with signi cant differences in the baseline characteristics between the two groups ( Table 1). In the subgroup analysis, we selected the top three combination treatments to evaluate the risk of AKI and perform pharmacoeconomic analysis. The patients in the VAN/PTZ group were matched to VAN/CPZ-SBT, VAN/MEM, and VAN/IPM-CIS groups in a 1:1 ratio. To evaluate risk factors of AKI between VAN/PTZ and VAN/BL groups, all variables with a P value < 0.2 in the bivariate matched analysis were included in multivariate analysis. For the secondary outcome of time to AKI, comparisons were made with the Kaplan-Meier curve and the log-rank test. We used SPSS 22.0 (SPSS Inc., Chicago, IL, USA) for statistical analysis. P < 0.05 was considered statistically signi cant. If a covariate was completely missing for an individual, the population median was used.  and 92 cases in the other combination groups) (Fig. 1).

Primary outcomes
In the matched cohort, the incidence of any AKI was higher among patients receiving VAN/PTZ compared to those receiving VAN/BL (   Table 2).
In addition, the Kaplan-Meier curve showed that the risk of any AKI in the VAN/PTZ group was signi cantly higher increased compared with the VAN/BL group based on AKIN criteria (Fig. 2).

Second outcomes
The intensive care unit (ICU) length of stay was signi cantly longer in the VAN/BL group than in the VAN/PTZ (20 vs 18 days, P = 0.039). Most patients in both groups (98.7% vs 99.7%) were unable to return to baseline renal function. There were no signi cant differences between the two groups in in-hospital mortality, the need for RRT, microbiological response, and day of AKI occurrence (Table 2).

Risk factors analysis for AKI
A total of 85 (17.5%) patients developed any AKI, and 46 (9.5%) patients developed stage 2 or 3 AKI. In multivariate regression analysis, the following variables were independent predictors of any AKI: VAN (Table 3).  Our study had several strengths. First, our study excluded severe renal insu ciency patients (eGFR ≤ 30 mL/min/1.73m 2 ), which reduced the risk of bias in the results. The results of a previous study(8) may draw a vague conclusion since the study did not exclude severe renal insu ciency patients who are assumed to be at an increased risk of AKI. Second, critically ill patients have a high baseline risk of AKI due to the severity of diseases and comorbidities. In contrast to previous studies that had inadequately adjusted for AKI risk, we used propensity matching scores to control potential bias to compare differences in incidence of AKI between the two combination therapies. Third, previous studies evaluated any stage of AKI, assuming that all stages are equally deleterious (18,20). In the current analysis, we focused on moderate to severe AKI, which is strongly associated with increased risk of morbidity and mortality (21). Fourth, VAN/PTZ may have a signi cant impact on kidney outcomes, resulting in increased mortality, hospital length of stay, and medical cost (22). Therefore, we performed a pharmacoeconomic analysis to further clarify the clinical applicability of combination therapy.
Although the AKI of VAN/PTZ is frequently reversible, short-term AKI may increase mortality for critically ill patients (23). Meanwhile, the duration of VAN/PTZ could be limited to reduce the incidence of AKI (24). Currently, studies have proved that selecting alternative combination therapy or limiting the combined duration of VAN/PTZ could effectively reduce the incidence of AKI (25). The higher risk of stage 2 or 3 AKI in the VAN/PTZ group emphasizes the need for individualized treatment in the selection of an appropriate antibiotic. The rational use of antibiotics can not only effectively prevent bacterial resistance, but also improve the clinical outcomes of patients. The latter includes prevention the adverse events, such as AKI, which may have a certain impact on the kidney outcomes of patients. In particular, stage 2 or 3 AKI are associated with CKD progression, prolonged hospital stay and increased mortality (26). In addition, our study indicated that all combination treatments were at least 48-72 hours and AKI occurred within 7 days. In clinical practice, rapid diagnosis might help to early discontinue VAN/PTZ, thereby preventing the incidence of AKI.
PTZ monotherapy does not cause kidney injury in a prospective study (27). If the use of VAN/PTZ dose increase AKI risk, whereas, the mechanism of nephrotoxicity induced by VAN/PTZ remains unclear. People have hypothesized that subclinical interstitial nephritis caused by PTZ that is exacerbated by oxidative stress (reactive oxygen species production) may induce kidney injury (28). Neveretheless, interstitial nephritis is usually a rare event. It seems unlikely that a large increase of 9% AKI (de ned by Scr) would occur since VAN/PTZ increases AKI risk through the mechanism of interstitial nephritis (10). Another hypothesis is that PTZ may reduce vancomycin clearance, resulting in the accumulation of vancomycin in the nephron (29).
However, there is no mechanistic evidence to support these two hypotheses.
This study has several limitations. First, this study was a retrospective analysis, unpredictable factors may in unce the results. We tted several multivariate models to adjust for factors to avoid the in uence of other AKI risk factors on the results. Second, our study used Scr to determine the degree of kidney injury. However, Scr is not a direct indicator of kidney injury, rather it is a surrogate of glomerular function. More speci c and sensitive biomarkers (such as kidney injury molecule-1 [KIM-1], osteopontin, etc.) are needed to determine whether they are associated with higher AKI when combination of VAN/PTZ. Third, we fail to assess the correlation between vancomycin trough concentration and the incidence of AKI in combination therapy due to lack of concentration data. Some researchers have found that there was no signi cant difference in the proportion of patients with vancomycin trough concentration > 15 mg/L or > 20 mg/L in the vancomycin/cefepime and the VAN/PTZ groups (30). Fourth, we fail to consider time-varying confounding.
In critically ill patients, the risk of AKI potentially changes daily because exposures associated with AKI may change daily (e.g. nephrotoxins, blood pressure, organ function). The potential for AKI risk to change after ICU admission is not addressed.
Future studies evaluating AKI associated with combination therapy should consider strati cation by baseline renal function in order to further explore the characteristics of this combination therapy. In addition, optimal strategies for managing VAN/PTZ-associated AKI should be determined for renal insu ciency patients, which could provide the possibility of preventing the incidence of AKI. Clinicians should recognize the risk factors of drug-induced AKI and closely monitor clinical response of patients during the course of treatment.
Moreover, the mechanism of nephrotoxicity caused by VAN/PTZ has not been well characterized.
Investigating the mechanism of nephrotoxicity has great signi cance to prevent the risk of nephrotoxicity and improve the clinical outcome of patients.

Conclusion
VAN/PTZ was associated with signi cantly higher incidence of AKI compared to VAN/BL in critically ill patients. In addition, patients receiving VAN/PTZ compared to VAN/CPZ-SBT, VAN/MEM, VAN/IPM-CIS could increase treatment cost when considering AKI occurrence. For critically ill patients who require empirical use of vancomycin and an antipseudomonal drug for antimicrobial therapy, clinicians should take a full consideration for institutional antimicrobial resistance patterns, possible pathogens, and the risk of AKI.
Alternative combination therapies other than VAN/PTZ should also be considered. Further research is needed to illuminate the mechanism of AKI caused by VAN/PTZ.