Incidence and risk factors of vancomycin-associated acute kidney injury in a single center: Retrospective study

Methods: Patients aged 18 and above who received a minimum of 24 hours of intravenous vancomycin and who had serial creatinine measurements over a 13-month period were identifi ed through electronic records. Patients with pre-existing AKI, or eGFR of less than 30ml/min, and patients with end stage kidney disease were excluded. Results were analyzed using t-test and Fisher’s test. A logistic regression model was used to identify the predictors for VA-AKI.


Introduction
Vancomycin is a glycopeptide antibiotic widely used for the management of methicillin-resistant Staphylococcus aureus (MRSA) infections. Earlier preparations of vancomycin in the 1950s were relatively impure and, as a result, was referred to as "Mississippi Mud". Previous reports showed that vancomycin could be both nephrotoxic and ototoxic [1]. Despite improvements in the puri ication process, literature suggests that vancomycin is still associated with 5% -40% increased incidence of acute kidney injury (AKI) [2][3][4][5][6]. A recent systematic review and meta-analysis suggested that use of vancomycin has a relative risk of 2.45 for development of AKI [7].
The exact mechanism of AKI with vancomycin use (VA-AKI) remains unclear; however, studies suggest that certain risk factors may increase the risk. Obesity [8,9] intensive care unit (ICU) stay [10] higher doses of vancomycin, prolonged duration of treatment, elevated serum vancomycin levels https://doi.org/10.29328/journal.jcn.1001067 [11] and concomitant treatment with nephrotoxic agents [12], have been implicated as risk factors which may have a role in the development of VA-AKI [2,3,10,13]. Hirai and colleagues observed that patients in their study showed poor recovery from vancomycin nephrotoxicity although vancomycin trough concentrations improved. In cases where vancomycin was continued, nephrotoxicity recovered in 40% of the patients [14]. Although VA-AKI is usually considered a reversible phenomenon, the consequences of AKI can be severe, ranging from increased length of hospitalization, long term dialysis, and transplantation.
Other associated risk factors include concomitant treatment with aminoglycoside [15], or piperacillintazobactam [16][17][18]. We believe that further systematic studies are needed to con irm and quantify the risk of VA-AKI and to identify the potential risk factors for the development of VA-AKI. We have, therefore, conducted a retrospective review of vancomycin use in our hospital and investigated the incidence of VA-AKI, in order to identify risk factors for the development of VA-AKI.

Inclusion and exclusion criteria
This is a retrospective cohort study of patients identi ied through a query of electronic health records (EHR) at a University-af iliated community hospital. We included all patients 18 years and older admitted between 01/01/2017 and 02/01/2018 who received greater than 24 hours of intravenous (IV) vancomycin. Vancomycin was initiated as deemed appropriate by the clinician for suspected or proven Gram-positive infection, either on admission or during hospital stay, and dosing was adjusted for renal function.
Patients with estimated glomerular iltration rate (eGFR) of less than 30 ml/min (based on the Modi ied Diet in Renal Disease [MDRD] equation) including those with end stage renal disease on renal replacement therapy were excluded. Patients with pre-existing AKI prior to vancomycin administration or who had less than four serum creatinine levels during their admission were also excluded.

Covariates
Data was reviewed from the EHR to extract information on age, body mass index (BMI), admission diagnosis, comorbidities, other medication use, baseline and peak serum creatinine, highest serum trough vancomycin level (or highest serum vancomycin level obtained when true trough not obtained), duration of vancomycin therapy and cumulative dosage, hypotension, admission hemoglobin, acknowledgement of AKI, and renal consultation during admission. Hypotension was de ined by mean arterial pressure less than 65 mmHg for more than three consistent readings or when vasopressors were used that admission. Baseline creatinine was the most recent serum creatinine value prior to vancomycin use. History of diabetes mellitus (DM), heart disease (congestive heart failure [CHF], arrhythmia [including supraventricular and ventricular tachycardia, as well as sick sinus syndrome and bradyarrhythmias and coronary artery disease [CAD]) and peripheral vascular disease (PVD) were identi ied based on chart review of medical notes. The use of antibiotics including amoxicillin-clavulanic acid, cefepime, ceftriaxone, piperacillin-tazobactam, caspofungin, azithromycin, doxycycline, ampicillin, aztreonam, sulfamethoxazoletrimethoprim, clindamycin, metronidazole, oseltamivir, and acyclovir were included. These antibiotics were included when used in combination with IV vancomycin at therapeutic doses adjusted for the renal function for at least 24 hours. Non-steroidal anti-in lammatory drugs (NSAIDs) included ibuprofen, ketorolac, diclofenac and celecoxib. Diuretics included furosemide or other loop diuretics, chlorthalidone, hydrochlorothiazide and spironolactone. Use of angiotensin converting enzyme inhibitor/angiotensin receptor blockade (ACEi/ARB) and iodinated contrast was also included.

Vancomycin Associated Acute kidney injury (VA-AKI)
AKI was de ined per the Kidney Disease Improving Global Outcomes (KDIGO) guideline, which for stage 1 is an increase in serum creatinine of greater than or equal to 0.3 mg/dL within 48 hours or an increase in serum creatinine to greater than or equal to 1.5 times baseline within the prior 7 days. Stage 2 AKI is de ined as an increase in serum creatinine 2-2.9 times the baseline, and stage 3 is de ined as an increase in serum creatinine greater than 3 times the baseline or initiation of renal replacement therapy. All cases who developed AKI were considered VA-AKI as every patient was receiving IV vancomycin.
All medical records, including all nephrology consultation notes, and other notes that documented identi ication and possible etiology of AKI were reviewed. Final decision regarding the etiology of the AKI was based on nephrology notes when present and/or chart review by the research investigator nephrologist who reviewed the medical records personally.

Endpoints
The primary endpoint was the incidence of AKI in patients who received vancomycin, de ined as VA-AKI. Acute nephrotoxicity solely related to vancomycin use was determined based on the absence of an alternative explanation of the acute kidney injury and on chart review by a nephrologist. Secondary endpoints included identi ication of risk factors which predispose patients to develop VA-AKI.

Statistical analysis
Descriptive statistics were conducted on all the variables. Baseline demographics and clinical characteristics were tabulated, mean with standard deviation for continuous variables and frequencies (proportions) with percentages for categorical variables. Statistical signi icance between those who developed VA-AKI and those who did not was obtained by t -test for the continuous variables, and from Fisher's exact test for the categorical variables reported as the p -value.
Univariate logistic regression analyses were performed for each exposure variable to compare the group with nephrotoxicity vs. those without, and to identify all the possible risk factors. A multivariate analysis was then constructed to test the association of identi ied risk factors while controlling the other variables. SAS statistical software (SAS Institute, Cary, NC) was used for statistical analysis.

Results
Our study was based on the 598 patients who met the inclusion criteria. 70 (11.7%) patients developed VA-AKI. The cause of AKI in 12 cases (17% of VA-AKI patients and 2% of total population) was deemed to be directly related to vancomycin toxicity, as determined by the nephrology consultation notes and/or chart review by the research investigator nephrologist. The cause of AKI in the rest of the population was mainly stated as multifactorial.
In those with VA-AKI, 48.6% (n = 34) had stage 1, 21.4% (n = 15) had stage 2, and 30% (n = 21) had stage 3 AKI. AKI was not documented in the medical records in 28 (40%) cases based on the changes in the creatinine per KDIGO de inition. Table 1 demonstrates the demographics of patients with and without VA-AKI. There were no signi icant differences in gender, mean age, and baseline serum creatinine between the two groups. Prevalence of comorbidities including anemia, diabetes mellitus, hypertension, peripheral vascular disease, heart disease, sickle cell disease and malignancy was not statistically different in the two groups.
Compared to those with no AKI, patients with VA-AKI had a longer duration of treatment with vancomycin (6.7 vs. 5.2 days) and were noted to have a higher mean serum vancomycin trough level (22.6 mg/L vs. 15 mg/L). Those with VA-AKI had a higher mean cumulative vancomycin dose of 13.3 mg/L compared to those with no AKI who had a mean cumulative vancomycin dose of 11.7 mg/L. However, this association was not statistically signi icant.
There was a signi icantly higher incidence of baseline CKD stage 3 (eGFR of less than 60 ml/min based on the MDRD Those with VA-AKI had a signi icantly higher incidence of concomitant use of piperacillin-tazobactam or metronidazole compared to the no AKI group. Total number of concomitant antibiotics was not statistically different in both the groups. We further subdivided those with VA-AKI into stages of AKI. No discernible associations were found when comparing stages of AKI with various risk factors.
The continuous variables are age, hemoglobin, baseline creatinine, highest serum vancomycin level, and vancomycin duration measured in terms of mean/standard duration. The rest of the variables are categorically reported as numbers and percentages.
We further studied the identi ied risk factors for VA-AKI and adjusted for possible confounding variables in table 2.
Univariate analysis revealed that serum trough vancomycin levels > 20 mg/L were associated with sixfold increased odds of VA-AKI when compared to those with highest serum trough vancomycin levels < 15 mg/L [6.46 (3.33, 12.5)]; this association remained robust even after adjusting for age, gender, comorbidities, diuretic use, iodinated contrast use, hypotension and piperacillintazobactam use [7.18 (3.37, 15.3)]. Duration of vancomycin use only marginally increased the odds of VA-AKI [1.05 (1.01,1.10)]. While increased duration of vancomycin use was associated with increased odds of VA-AKI in the univariate analysis, this association was not seen following adjustment.
VA-AKI was seen more frequently in group with eGFR < 60 ml/min, however, this association was not signi icant when adjusted for other variables.
Hypotension was associated with a four-fold increase in odds of VA-AKI in our population [  Data are expressed as odds ratio (95% con idence interval).

D iscussion
Vancomycin is often used as a irst line empiric therapy for many moderate-to-severe gram-positive infections and has been associated with AKI [4]. Given the widespread use of vancomycin, there are reports of VA-AKI ranging between 5% to 40% [2][3][4][5][6].
In our study, 11.7% of patients who received vancomycin and met the inclusion criteria developed VA-AKI. Only in 2% of the total 598 cohort patients was vancomycin considered the most likely cause. In 83% of cases of VA-AKI, the cause of AKI was determined as multifactorial with vancomycin being a contributing factor. In our study, patients with high vancomycin levels had a signi icantly higher incidence of AKI. This inding suggests that even when vancomycin was not considered the main cause of AKI, it could have played a role in the development of VA-AKI.
We observed signi icant differences in the serum trough vancomycin levels in the groups with VA-AKI compared to those without AKI (mean level of 22.6 mg/L vs. 15 mg/L respectively; p < 0.001). In our study, odds of VA-AKI was signi icantly increased in the group with serum vancomycin levels greater than 20 mg/L with odds ratio of 6.46 (3.33, 12.5). This observation remained robust after adjustment for other major risk factors such as hypotension. This inding is of particular importance as it suggests that the association seen between high serum vancomycin trough levels and VA-AKI is not confounded by severity of sepsis and hemodynamic instability. A recent study has also shown similar indings [3].
In a meta-analysis by Hal, et al. [6], odds of nephrotoxicity remained increased, at 3.12 (95% CI, 1.81-5.37; p < 0.01) for patients attaining initial trough levels of ≥ 15 mg/L, indicating that an augmented risk of toxicity occurs in individuals with troughs of ≥ 15 mg/L. In our study, we did not ind a higher risk of AKI when highest vancomycin levels were between 15-20 mg/L compared to those with levels < 15 mg/L despite adjustment for other variables. Due to the small demarcation in trough values between 15-20 mg/L and > 20 mg/L, it is possible that vancomycin trough values of > 20 mg/L could be the main factor driving nephrotoxicity in the > 15 mg/L strata in prior observational studies. It is also possible that our sample size did not have the power to detect differences at that range.
Earlie r studies reported that treatment with vancomycin for a duration of ≥ 14 days is associated with a higher risk of VA-AKI, de ined as a rise in serum creatinine of ≥ 0.5 mg/dL [19]. However, in our study, odds of VA-AKI was not increased by increased duration of vancomycin use; this is similar to a recent study showing AKI to be associated with high vancomycin trough level rather than dose or duration [20].
The mechanism of vancomycin induced nephrotoxicity is unclear. Several possible mechanisms of injury have been proposed. These include acute tubular injury and/or acute tubulointerstitial nephritis as a result of the induction of oxidative stress, complement activation with in lammatory injury, and mitochondrial damage [21]. Obstructing tubular cast formation has also been proposed as a novel possible cause of VA-AKI [22]. The multifactorial nature of VA-AKI is best con irmed by the indings of two randomized clinical trials which also support the notion that certain clinical factors can augment the severity of the nephrotoxicity [23]. One study [12] conferred that a higher number of nephrotoxic agents used with vancomycin predisposes to increased risk of AKI. Another study even argued that intravenous vancomycin at therapeutic levels as monotherapy may not be nephrotoxic when compared to other antibiotics [24].
Based on the above proposed mechanisms, our study has investigated possible risk factors for the development of VA-AKI. According to our study, the presence of hypotension was a strong and independent predictor of VA-AKI, following adjustment for variables that could be potential confounders. In our population, those with VA-AKI had a statistically signi icant higher incidence of hypotension when compared to those without AKI: 40% vs. 13% respectively. Of those with hypotension and VA-AKI, 36% of cases of AKI (stages 1 and 2) were not acknowledged in the medical notes by the providers. This pattern was mainly seen in patients who had a creatinine rise which remained within the normal limits set by the EHR (in our system 0.44 -1.03 mg/dL).
Similarly, simultaneous iodinated contrast use also was associated with increased risk of AKI. The exact mechanism of renal injury with vancomycin in the presence of hypotension or iodinated contrast remains unclear. It is likely that the combination of vancomycin and hypotension or iodinated contrast augments the prerenal state leading to impaired renal perfusion, increasing the risk of oxidative stress and tubular necrosis. This may indicate that, in patients who receive vancomycin, special attention needs to be considered for maintaining hemodynamic stability and preventing hypotension while also trying to avoid iodinated contrast use if possible. The indings from this study do not however prove a causal relationship between vancomycin use in hypotension and AKI. It merely could be that critically ill patients are more likely to receive vancomycin.
In our analysis, increased odds of VA-AKI were not https://doi.org/10.29328/journal.jcn.1001067 observed with concomitant NSAID and vancomycin use although NSAIDs do cause afferent renal artery vasoconstriction. It is possible that the doses of NSAIDs received were not signi icant or given for a prolonged period to impact the renal function.
Several studies and a systematic review have reported an association between use of concomitant piperacillintazobactam and VA-AKI by mechanisms not well understood [16][17][18]. Decreased clearance of vancomycin by piperacillintazobactam, resulting in vancomycin accumulation, sometimes muted the impact of vancomycin trough concentrations on AKI [18,[25][26][27]. In our study, we tested a wide variety of concomitant antibiotic usage and found that piperacillin-tazobactam use, as an independent variable, was associated with a signi icant increase in the risk of VA-AKI. The effect of piperacillin-tazobactam in increasing the risk of AKI remained signi icant even after adjustment for serum vancomycin levels > 20 mg/L. These results provide an impetus to obtain further insight into the true mechanisms at a molecular level to explain this nephrotoxic effect of vancomycin and piperacillin-tazobactam combination. Although simultaneous use of aminoglycosides with vancomycin is a well-known risk factor for AKI, the association was not observed in our study [15,24]. This could perhaps be due to the limited number of patients (46 out of 598 patients) who received this combination.
Our study adds to the literature by looking at a more comprehensive range of variables including administration of a variety of concomitant antibiotics to identify potential risk factors for VA-AKI. Furthermore, the adjustment models for confounders were built to narrow down the risk factors which play a key contributing role in VA-AKI. Identi ication of such key risk factors helps to build risk prediction models to prescribe and use vancomycin safely, such as in a study conducted by Miyai and colleagues [28].
Although our study contributes to better understanding of the incidence and risk factors of VA-AKI, it does have some limitations. It is a retrospective study and, therefore, any conclusions should be interpreted as exploratory in nature. Furthermore, we were limited by the data that was already collected in the database. As such, there may still be residual confounding effects that we could not ascertain. For example, the signi icance of concomitant metronidazole use in patients was not persistent once adjusted for other variables, and the signi icance of this is not clear to us. In order to minimize system errors, however, a detailed chart review of patient records who developed AKI was carried out by a nephrologist once the data was collected. Furthermore, the lack of con irmed trough serum vancomycin levels is a limitation of our study, however, this is more re lective of real-life practices when using vancomycin. Exclusion of patients with eGFR less than 30 is also a limitation of our study.

Conclusion
The incid ence of VA-AKI in hospitalized patients with eGFR greater than 30 ml/min was 11.7%, and cause of the AKI in most cases was considered as multifactorial. Higher vancomycin levels of more than 20 mg/L signi icantly increased the risk of AKI. The presence of hypotension and the administration of iodinated contrast were among other risk factors. Patients with eGFR < 60 ml/min are at a higher risk of developing VA-AKI. Piperacillin-tazobactam, when used with vancomycin, was noted to be an independent predictor of AKI, regardless of serum vancomycin trough levels, prompting a reevaluation of the safety of the widespread practice of this combination as an empiric therapy. Close monitoring of kidney function, avoiding high serum vancomycin levels, maintaining hemodynamic stability, and avoiding unnecessary use of iodinated contrast seem to be essential for the diagnosis and prevention of VA-AKI.