A Quasi-Experimental Evaluation of Single Trough-Based Area Under the Curve Guided Dosing on the Incidence of Vancomycin Associated Nephrotoxicity in Veteran Patients

Abstract

Background: Two common dosing strategies for vancomycin are trough-based and area under the curve (AUC)-based dosing. Objective: To compare the incidence of nephrotoxicity in trough-based dosing group with the single trough-based AUC dosing at the Salem VA Medical Center. Methods: This retrospective study included patients who received trough-based dosing of vancomycin between January 1, 2017, and January 1, 2019 (preimplementation group) and AUC-based dosing (postimplementation) between October 1, 2019, and October 1, 2021, at the Salem VA Medical Center. The primary outcome was nephrotoxicity at 96 hours, 7 days, and entire hospital length of stay (LOS). Secondary outcomes included 30-day readmission and all-cause mortality rates, cumulative doses at 24, 48, and 72 hours, and percentage of patients considered at goal (AUC 400-600 or trough between 10 and 20 mg/L). Propensity score (PS) matching was utilized to adjust for confounding. Results: After PS matching 100 patients were included in preimplementation and 95 patients in the postimplementation group. The average study patient was a 68-year-old white male. There was significant reduction in the risk of nephrotoxicity in postimplementation cohort at 96 hours (adjusted (a)HR: 0.28, 95% CI (0.12-0.66); 7 days (aHR: 0.39, 95% CI (0.18-0.85); and entire hospital LOS (aHR: 0.46, 95% CI (0.22-0.95). Secondary outcomes showed no difference between the groups except significantly higher proportion of patients were considered at therapeutic goal in the postimplementation cohort compared with pre-implementation cohort. Conclusion: This hypothesis generating study shows that AUC-based dosing calculated using single trough concentration may result in reduced rate of nephrotoxicity than trough-based dosing.

Background

Vancomycin, a glycopeptide antibiotic, is widely utilized in hospitals as the treatment of choice for serious Gram-positive infections, especially methicillin-resistant Staphylococcus aureus (MRSA). However, its use is associated with significant adverse effects including nephrotoxicity. ¹ Vancomycin-associated nephrotoxicity is linked to increased duration of hospitalization, medical costs, and risk of mortality. ¹ The rate of nephrotoxicity varies in the literature ranging from 0% to over 40% in the absence or presence of concurrent nephrotoxins respectively. ² Potential risk factors include those directly related to vancomycin exposure such as total daily dose, area under the curve vs. time (AUC), trough level, and duration of therapy, as well as patient-related factors including obesity, concurrent nephrotoxins, and chronic kidney disease. ²

Guideline authors suggest that therapeutic drug concentrations (TDM) using AUC is a better predictor of vancomycin activity against MRSA. ³ While traditionally trough concentrations of 15 to 20 mg/L have been used for serious infections such as bacteremia, endocarditis, osteomyelitis, and pneumonia they are usually unnecessary to achieve AUC target of ≥400 mg h/L. ³ Contrary to this, the infective endocarditis scientific statement recommends targeting trough concentrations of 10 to 20 mg/L to reduce the vancomycin exposure. ⁴ Vancomycin trough concentrations of 15 to 20 mg/L correlate with AUC > 600 mg.h/L which have been associated with increased incidence of nephrotoxicity.^2,5-8 More recent data suggest that >50% patients can meet this daily AUC target with trough concentrations of <15 mg/L which may ultimately result in reduced nephrotoxicity.^9,10 In a retrospective quasi-experimental study of 1280 hospitalized patients, vancomycin AUC-guided dosing was associated with lower total daily vancomycin doses, AUC values, and trough concentration, ultimately resulting in reduced nephrotoxicity. ¹⁰ The most recent guidelines for the therapeutic monitoring of vancomycin for serious MRSA infections recommends targeting an AUC of 400 to 600 mg h/L in most infections with assumed minimum inhibitory concentration (MIC) of 1 mg/L to reduce vancomycin exposure based on the national vancomycin susceptibility surveillance and has moved away from endorsing trough monitoring. ¹¹ While an increasing body of evidence supports AUC use, the majority of the data are based on using two-point vancomycin concentrations and use of Bayesian statistics, often not feasible at smaller institutions with limited resources. Evidence indicates that veteran patients have higher comorbidity burden specifically when serving a rural underserved area such as our institution limiting the generalizability of previous studies conducted in the non-veteran population. ¹²

Objective

We aim to investigate the incidence of nephrotoxicity in veterans’ pre-and post-implementation of the AUC-guided dosing based on single-trough concentration at Salem Veterans Affair Health Care System.

Methods

Study Setting

This quasi-experimental study included patients who received at least 2 doses of vancomycin. Patients were further screened to include only if they had at least one trough obtained near steady state within 96 hours of therapy. Patients who received less than 3 doses of vancomycin and had pre-existing need for renal replacement therapy were excluded from the study. The AUC-dosing strategy was implemented in September 2019. The pre-implementation cohort consisted of patients in whom vancomycin was initiated using a trough-based dosing strategy between January 1, 2017, and January 1, 2019. The post-implementation group included patients in whom vancomycin was initiated under an AUC-based dosing strategy between October 1, 2019, and October 1, 2021. Patients with COVID-19 infection were further excluded from this study. Propensity score (PS) was utilized to match the two cohort prior to conducting manual chart review. Minimum inhibitory concentration (MIC) numbers were available for all MRSA isolates but not reported in medical records due to interface issues between 2020 through 2022. The Clinical Laboratory Standards Institute (CLSI) guidelines were applied for interpretation of sensitivity, only interpretations (sensitive, intermediate, and resistant) were entered manually into the patients’ chart. MIC < 2 were considered susceptible and all our isolates were considered sensitive with majority MIC = 1.

Data Parameters

Baseline demographics collected include age, gender, race, inpatient location (medical vs intensive care), length of stay, height, weight, serum creatinine (SCr), estimated glomerular filtration rate (eGFR), indication for vancomycin use, average total daily dose, concurrent nephrotoxic medication (ie, diuretics, IV contrast dye, RAAS inhibitors, nonsteroidal anti-inflammatory drugs (NSAIDs), IV piperacillin/tazobactam), duration of vancomycin therapy, steady state trough concentration, calculated AUC and past medical history such as obesity defined as BMI > 30 kg/m², cardiac arrythmias, valvular disease, neurological disease, and additional comorbidities as defined by Charlson Comorbidity Index. ¹³ 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). ¹⁴ Baseline SCr was defined as the SCr value immediately preceding the first dose of vancomycin. 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. ¹⁵

Vancomycin Dosing at Salem VA Medical Center

Prior to September 2019, vancomycin was dosed to target serum trough concentrations of 10 to 20 mg/L based on published pharmacokinetic (PK) equations and indications.^16,17 Clinical Pharmacy Practitioners at our facility are considered advanced practitioner and use their scope of practice to manage all patients admitted to the hospital who are initiated on vancomycin.

Appropriately timed trough concentration near steady state provides a good way of estimating AUC. This can be accomplished by either Bayesian statistics or PK equations. Implementation of Bayesian methodology for AUC-guided vancomycin dosing requires specialized commercial software tools that are expensive and not available at our institution. In addition, trying to get two labs on either side of the dose was logistically challenging and therefore a significant limitation in being able to implement two level peak- and trough-based AUC method at our facility. Solving complex PK equation manually is not practical, and we utilized the open accessGlobalRPh.com calculator based on single trough concentration to estimate AUC for vancomycin therapeutic drug monitoring (TDM) with an assumed MIC of 1.^18,19 This has been a consistent practice at our facility for the past 5 years. GlobalRPh.com requires users to specify a desired peak and trough, which are then used to calculate empiric dose and dosing interval (tau) based on the trapezoidal method to estimate initial AUC. ¹⁸ It uses Ambrose–Winter’s single trough method to estimate AUC based on appropriately sampled trough concentration at steady state. A population volume of distribution (Vd) is used to estimate the vancomycin peak concentration (Peak=measured trough + Dose/Vd). This calculated peak and measured trough are then used to calculate an elimination rate constant (k_el = ln (Peak/Trough)/Tau). Infusion equations are then used to calculate a new tau and dose. ¹⁹ In a study comparing open access vancomycin dosing calculators, GlobalRPh.com essentially calculated the same dosage regimen for patients with normal body weight. ²⁰ Most of the general adult Vd models have limited sampling of patients with extremes of body weight.^21,22

Sample Measurements

Vancomycin concentrations are measured by the Siemens Dimension Vista 1500 (Siemens Healthineers, Inc., Tarrytown, NY). The methodology for vancomycin concentration is based on a homogeneous particle enhanced turbidimetric inhibition immunoassay (PETINIA) technique. Measurement range is 0.8 to 50 ug/mL.

Outcomes

Primary Outcome

The primary outcome of this study was comparing the rate of nephrotoxicity at 96 hours, 7 days and entire hospital length of stay (LOS) from the first dose of vancomycin between pre-implementation (trough-based) and post-implementation (AUC-guided dosing) cohorts. Vancomycin-associated nephrotoxicity was defined as a minimum of two consecutive documented increases in SCr concentrations defined as an increase of 0.5 mg/dL or a ≥50% increase from baseline, whichever is greater, according to the 2009 vancomycin consensus guideline. ²³

Secondary Outcomes

The secondary outcomes of this study were comparing the rate of acute kidney injury as defined by Rifle (Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease) injury/worse classification published by the Acute Dialysis Quality Initiative (ADQI) and Acute Kidney Injury Network (AKIN) staging 1 or worse.^24,25 In addition, the proportion of patients deemed to be at treatment goal, cumulative doses at 24, 48, and 72 hours, 30-day mortality, and 30-day readmission rates were also compared between the pre-and post-implementation cohort. Since early nephrotoxicity may be attributed to pre-renal acute kidney injury (AKI) in hospitalized patients when first admitted, we conducted a sensitivity analysis to exclude nephrotoxicity events that occurred within the first 24 hours for the primary outcome.

PS Matching

After data extraction using Structured Query Language from corporate database warehouse and applying initial exclusion criteria, patients in the post-implementation group were matched with those in the pre-implementation group in a 1:1 ratio using PS derived by using logistic regression based on predefined variables reported in Table 1. Conventional greedy algorithm with nearest neighbor matching without replacement and maximum caliper width of no more than 0.2 of standard deviation of the logit of score was utilized. Additional manual chart review was conducted to confirm the number of vancomycin doses patient received, trough values, AUC, and use of dialysis during hospitalization.

Table 1.

Baseline Demographics.

	Pre-implementation Cohort N = 100 (%)	Post-implementation Cohort N = 95 (%)	P value
Age in years (mean ± SD)	69.08 ± 12.10	67.12 ± 11.69	0.254
Male	95 (95.00%)	91 (95.79%)	0.937
Race
White	86 (86.00%)	79 (83.16%)	0.107
Black or African American	12 (12.00%)	13 (13.68)	0.891
% Native Hawaiian/Pacific Islander	1 (1.00%)	0 (0.00%)	—
Declined to answer	1 (1.00%)	3 (3.16%)	—
Vancomycin doses (mean ± SD)	4.12 ± 2.08	3.75 ± 3.08	0.321
Baseline serum creatinine (mean ± SD)	1.20 ± 0.56	1.20 ± 0.65	0.968
Baseline eGFR (mean ± SD)	74.03 ± 30.63	79.08 ± 37.29	0.301
Height (inches, mean ± SD)	69.72 ± 3.22	69.23 ± 3.24	0.298
Weight (lbs, mean ± SD)	209.57 ±52.91	213.72 ± 66.79	0.629
BMI (mean ± SD)	30.05 ± 7.41	31.33 ± 9.49	0.379
Charlson comorbidity index (CCI) median [IQR]	6.35 [3-9]	6.6 [4-9]	0.496
Length of stay (median [IQR])	10.11 [5.62-12.26]	10.87 [3.78-14.79]	0.618
Trough concentration (mean ± SD)	12.75 ± 5.39	12.78 ± 4.17	0.949
AUC/MIC (mean ± SD)	—	502.45 ± 68.43	—
Vancomycin indication
Pneumonia	35 (35.00%)	29 (30.53%)	0.608
SSTI	35 (35.00%)	35 (36.84%)	0.906
Bacteremia	9 (9.00%)	9 (9.47%)	0.909
Osteomyelitis	8 (8.00%)	6 (6.32%)	0.859
Urinary tract infection	6 (6.00%)	5 (5.26%)	0.999
Central nervous system infection	2 (2.00%)	0 (0.00%)	—
Febrile neutropenia	2 (2.00%)	3 (3.16%)	—
Abdominal/pelvic infection	1 (1.00%)	7 (7.37%)	—
Endocarditis	1 (1.00%)	1 (1.05%)	—
Other/undifferentiated	1 (1.00%)	0 (0.00%)	—
Vancomycin duration ≥ 7 days	21 (21.00%)	13 (13.68%)	0.178
Vancomycin duration	4 (5-7)	3 (2-8)	0.172
Admitted to ICU	32 (32.00%)	30 (31.58%)	0.950
Concomitant IV contrast	45 (45.00%)	44 (44.21%)	0.912
Concomitant disease states
Chronic pulmonary disease	60 (60.00%)	53 (55.79%)	0.552
Depression	58 (58.00%)	64 (67.37%)	0.177
Diabetes without chronic complications	59 (59.00%)	61 (64.21%)	0.455
Diabetes with chronic complications	46 (46.00%)	46 (48.42%)	0.735
Peripheral vascular disease	41 (41.00%)	36 (37.89%)	0.657
Renal disease	31 (31.00%)	31 (32.63%)	0.807
Cerebrovascular disease	37 (37.00%)	35 (36.84%)	0.982
Cardiac arrythmias	37 (37.00%)	40 (42.11%)	0.466
Any malignancy	29 (29.00%)	27 (28.42%)	0.929
Deficiency anemia	24 (24.00%)	29 (30.53%)	0.306
Pulmonary circulatory disorders	23 (23.00%)	22 (23.16%)	0.979
Hypotension	22 (22.00%)	22 (23.16%)	0.847
Mild liver disease	19 (19.00%)	18 (18.95%)	0.993
Drug abuse	19 (19.00%)	20 (21.05%)	0.720
Myocardial infarction	16 (16.00%)	20 (21.05%)	0.363
Valvular heart disease	15 (15.00%)	22 (23.16%)	0.146
Coagulopathy	15 (15.00%)	13 (13.68%)	0.793
Neurological disease	14 (14.00%)	22 (23.16%)	0.099
Dementia	13 (13.00%)	13 (13.68%)	0.888
Hypothyroidism	13 (13.00%)	13 (13.68%)	0.888
Peptic ulcer disease	10 (10.00%)	6 (6.32%)	0.349
Psychosis	9 (9.00%)	9 (9.47%)	0.909
Alcohol abuse	9 (9.00%)	9 (9.47%)	0.909
Metastatic solid tumor	8 (8.00%)	5 (5.26%)	0.444
Rheumatoid arthritis	8 (8.00%)	10 (10.53%)	0.542
Blood loss anemia	8 (8.00%)	14 (14.74%)
Hemiplegia/paraplegia	5 (5.00%)	4 (4.21%)	0.793
Lymphoma	5 (5.00%)	2 (2.11%)	0.277
Moderate/severe liver disease	3 (3.00%)	3 (3.16%)	0.949
Congestive heart failure	1 (1.00%)	2 (2.11%)	0.531
Concomitant nephrotoxins inpatient
Piperacillin-tazobactam	28 (28.00%)	29 (30.53%)	0.698
Loop diuretics	6 (6.00%)	3 (3.16%)	0.344
Vasopressor	4 (4.00%)	2 (2.11%)	0.444
Cefepime	2 (2.00%)	4 (4.21%)	0.372
NSAIDs	2 (2.00%)	1 (1.05%)	>0.999
Aminoglycosides	0 (0.00%)	1 (1.05%)	0.487
Concomitant nephrotoxins outpatient
ACE inhibitors	22 (22.00%)	25 (26.32%)	0.481
Loop Diuretics	19 (19.00%)	18 (18.95%)	0.993
NSAIDs	17 (17.00%)	30 (31.58%)	0.017
ARBs	7 (7.00%)	7 (7.37%)	0.921
ARNI	3 (3.00%)	1 (1.05%)	0.338

Abbreviations: ACE, angiotensin converting enzyme; ARB, angiotensin receptor blockers’; ARNI, angiotensin receptor and neprilysin inhibitor; AUC/MIC, Area under the curve/minimum inhibitory concentration; BMI, body mass index; eGFR, estimated glomerular filtration rate; ICU, intensive care unit; IQR, interquartile range; IV, intravenous; NSAIDs, non-steroidal anti-inflammatory drugs; SSTI, skin and soft tissue infection.

Statistics

Baseline characteristics were described using descriptive statistics as count and percentage for categorical variables and either mean with standard deviation or median with interquartile range based on the skewness and kurtosis of the data variables. Dichotomous data was compared between the groups using chi-square or Fisher exact test. Continuous data was analyzed using independent t-test or Mann–Whitney U tests. Based on a previously published study in the veteran population, the following factors were analyzed for association with incidence of nephrotoxicity using bivariate analysis to include in the primary outcome analysis by cox proportional model: baseline CrCl <60 mL/min, vancomycin therapy ≥7 days, ICU stay, co-administration of piperacillin-tazobactam, use of IV contrast, and CCI adjusted for age. ²⁶ Variables with P value <0.2 in bivariate analysis were then included in the multivariable logistic regression model and backward elimination methods was used to identify variables associated with nephrotoxicity to assess for inclusion in the cox proportional hazard model to adjust hazard ratios (HRs). Model fit was assessed with the Hosmer–Lemeshow goodness-of-fit test and a nonsignificant result was considered adequate. Primary outcomes were analyzed using cox proportional and both unadjusted and adjusted hazard ratios (HR) with corresponding 95% confidence interval (CI) and P-value were reported. Proportional-hazards assumptions were assessed with Schoenfeld residuals. Kaplan–Meier failure graphs were reported, and log-rank test was used to analyze the differences. All statistical tests were two-sided, and P values of ≤0.05 were considered statistically significant. All analyses were performed using SAS software, version 9.4 (SAS institute, Inc., Cary, NC, USA) and Stata version 17, College Station, TX:StataCorp LLC.

Results

A total of 593 patients in pre-implementation cohort and 401 patients in post-implementation cohort were initially screened. After initial patient exclusions, 1:1 PS matching resulted in 131 patients in each cohort. Additional manual chart review resulted in further exclusions (Figure 1). Finally, 100 patients were included in the pre-implementation group and 95 patients in the post-implementation group. The average patient in this study was a 68-year-old white male. The baseline characteristics between each group were similar (Table 1) except for outpatient NSAID utilization which was significantly higher in the post-implementation cohort as compared with the pre-implementation cohort (31.58% vs. 17.0%; P value = 0.017).

Figure 1.

Patient selection flowchart. Please note that patients were initially pulled using Structured Query language (SQL) and excluded based on the initial exclusions such as number of vancomycin doses, if they received dialysis during the hospital stay, if no trough level was available, or if patients had duplicate entry due to change in their wards, that is, step down from ICU to medical floors. After which 139 eligible patients from the post-implementation cohort were matched 1:1 using PS as described in the methods section of the manuscript resulting in 131 matched patients in each cohort Due to the limitation of SQL and Electronic medical records, manual review of charts resulted in additional exclusion criteria resulting in total of 95 patients in post-implementation cohort and 100 patients in pre-implementation cohort.

Abbreviation: PS, Propensity Score.

Primary Outcome

A total of 21 patients had nephrotoxicity in the pre-implementation cohort, all of which occurred within 96 hours of the first vancomycin dose. In the post-implementation cohort, nephrotoxicity occurred in 12 patients for the duration of hospital stay: in 10 patients within 7 days and in 7 patients within 96 hours. There was a significant reduction in the risk of nephrotoxicity in the post-implementation cohort at 96 hours (adjusted (1) HR: 0.28, 95% CI (0.12-0.66); P value = 0.004); 7 days (aHR: 0.39, 95% CI (0.18-0.85); P value = 0.017); and entire hospital LOS (aHR: 0.46, 95% CI (0.22-0.95); P value = 0.036). While the unadjusted HR with 95% CI did not reach statistical significance for the outcome of nephrotoxicity for entire hospital LOS, aHR with 95% CI was statistically significant (Table 2). Multivariable logistic regression identified baseline CrCl <60 mL/min, vancomycin therapy ≥7 days, ICU stay, co-administration of piperacillin-tazobactam in the final model as being significantly associated with the odds of nephrotoxicity and hazed ratios reported were adjusted for these variables (Table 3). However, sensitivity analysis of primary outcome after excluding events that occurred in the first 24 hours of hospitalization showed no difference in the rate of nephrotoxicity between pre-and post-implementation groups (18 (18%) vs 12 (12.63%), aHR: 0.64, 95% CI (0.31-1.35); P-value = 0.246). Kaplan–Meier failure graphs for primary outcomes are reported in Figure 2a-2c. Average time to nephrotoxicity was 4.76 days in trough dosing group versus 1.32 days in the AUC dosing group. In addition, eGFR < 60, ICU stay and inpatient use of piperacillin-tazobactam were significantly associated with the primary outcome in the cox proportional regression model.

Table 2.

Primary and Secondary Outcomes.

Primary outcomes	Pre-implementation Cohort N = 100	Post-implementation Cohort N = 95	Unadjusted HR (95% CI)	P value	Adjusted HR (95% CI)	P value
Nephrotoxicity at 96 hours	21 (21%)	7 (7.37%)	0.32 (0.14-0.76)	0.010	0.28 (0.12-0.66)	0.004
Nephrotoxicity at 7 days	21 (21%)	10 (10.53%)	0.46 (0.22-0.98)	0.045	0.39 (0.18-0.85)	0.017
Total Nephrotoxicity	21 (21%)	12 (12.63%)	0.53 (0.26-1.08)	0.080	0.46 (0.22-0.95)	0.036
Secondary outcomes		Pre-implementation Cohort N = 100 (%)		Post-implementation Cohort N = 95 (%)
AKIN stage 1 or worse		14 (14%)		10 (10.53%)
AKIN Stage 2 or worse		4 (4%)		2 (2.10%)
RIFLE risk/worse		25 (25%)		19 (20.00%)
RIFLE injury/worse		6 (6%)		4 (4.21%)
All-cause mortality 30 days		14 (14.00%)		15 (15.79%)
Re-admission rate at 30 days		15 (15.00%)		8 (8.42%)
Cumulative vancomycin doses		Mean ± SD		Mean ± SD
24 hours		2092.50 ± 853.29		2252.63 ± 1073.43
48 hours		4277.50 ± 1648.73		4336.84 ± 2018.46
72 hours		6043.50 ± 2621.08		5615.79 ± 2705.84

Post implementation group.

Abbreviations: AKIN, acute kidney injury classification; CI, confidence interval; HR, hazard ratios; RIFLE, risk, injury, failure, loss of kidney function and end-stage kidney disease.

Table 3.

Multiple Variable Logistic Regression: Predictors of Nephrotoxicity.

Variables	Odds ratio (95% CI)	P value
Vancomycin duration >7 days	2.73 (1.04-7.19)	0.041
eGFR < 60 mL/min	4.24 (1.82-9.86)	0.001
ICU stay	2.13 (0.93-4.85)	0.073
Concomitant piperacillin-tazobactam use	1.28 (1.09-1.89)	0.030

Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate; ICU, Intensive care unit.

Figure 2.

(a) Time to nephrotoxicity at 96 hours. (b) Time to nephrotoxicity at 7 days. (c) Time to nephrotoxicity for hospital LOS.

Abbreviation: LOS, length of stay.

Secondary Outcomes

No significant differences in the secondary outcomes were noted between the two groups (Table 2). More patients were deemed to be at goal in the AUC-based dosing cohort than in the trough-based dosing cohort (90.53% vs 71.0%: P value = 0.001).

Conclusion

This single center quasi-experimental study showed that AUC-guided dosing reduced nephrotoxicity at 96 hours, 7 days, and hospital LOS. More patients were deemed at goal in the post-implementation cohort compared to pre-implementation cohort (90.53% vs. 71%).

In a retrospective quasi-experimental study of 1280 patients by Finch NA et al., AUC-guided dosing strategy based on Bayesian methodology was associated with significantly lower risk of AKI (OR: 0.52; 95% CI: 0.34-0.80) compared with the trough guided strategy. This study excluded patients with skin and soft tissues infections without concomitant bacteremia and patients were much younger with higher proportion of female population. Concomitant use of piperacillin-tazobactam, and baseline use of other nephrotoxins were also not reported. Like our study, it did not show significant difference in the AKIN stage 1 or worse and RIFLE category of risk or worse between the two dosing strategies. Of note when evaluated by the seriousness of the kidney injury indicated as AKIN stage 2 or worse and RIFLE category of Injury or worse, it was apparent that majority patient experienced less severe form of kidney injury in our study. ¹⁰ One possible explanation is a rapid recovery of renal function during the treatment duration may reduce the severity of the AKI experienced by the patient.

A metanalysis that included 57 studies by Lim AS et al., showed that the risk of nephrotoxicity was lower in AUC-guided TDM compared with the trough-guided approach (OR: 53; 95% CI: 0.32-0.89). A majority of the included studies utilized trough of 15 to 20 mg/L as comparator and so reduced dose in the AUC group may account for the decreased nephrotoxicity observed. In addition, AUC thresholds correlated with nephrotoxicity risk only for the first 96 hours suggesting that factors beyond vancomycin use likely associated with longer LOS may impact the incidence of nephrotoxicity. ²⁷ Our study observed similar trends. Tsutsuura et al. performed a meta-analysis of 16 studies to explore the relationship between trough concentrations and AKI incidence and showed significant risk of AKI with trough concentrations >20 mg/mL. High AUC/MIC (>600 also significantly increased the risk of AKI (OR: 2.10; 95% CI: 1.13-3.89) in this metanalysis. ²⁸

In a 3-year prospective trial of 252 patients, vancomycin AUC-guided dosing was associated with decreased nephrotoxicity and shorter length of therapy in 252 patients. Overall, this study showed significantly higher proportion of patients as being therapeutic based on AUC compared with the trough-guided strategy (70% vs 19%). About 25% of trough-based cohort had duration of therapy greater than 14 days, with median duration of vancomycin therapy 7.8 days in year 1, and 5.4 and 4.7 days in years 2 and 3 when AUC-guided dosing was used. ⁹ This along with the results of our study demonstrate that the duration to be an important risk factor for vancomycin associated nephrotoxicity. Our study showed no difference in the cumulative dosing of vancomycin at 24, 48, and 72 hours. While the mean dose of vancomycin was lower in the post-implementation cohort at 72 hours by 428 mg, this difference did not reach significance. Previous studies have indicated higher total daily dose in patients with higher BMI. However, we could not stratify based on BMI because of low sample size. ²⁹

Generally, it is preferable to use a two-point vancomycin concentration to estimate AUC but for purpose of our study only single trough concentrations were available. Recently, a retrospective cohort study in 161 patients at a 614 bed Japanese hospital showed that AUC-guided vancomycin dosing based on single trough concentration was better predictor of nephrotoxicity than trough guided monitoring in patients receiving vancomycin. ³⁰ Another study by Ueda et al examined the effectiveness of AUC estimation in 260 adults with MRSA bacteremia and showed that AUC estimate using one point trough concentration was a better predictor of vancomycin-related nephrotoxicity in critically ill patients. ³¹ The results of our study would be relevant to many institutions as getting two serum samples and ensuring accurate time of both administration and sampling may not be always feasible. In a mixed-methods study of 163 patients, a collection of multiple samples within the same dosing interval only occurred in 7% of the occasions, indicating poor concordance of dosing and TDM with institutional guidelines. ³² In the clinical environment, only 20% to 59% of all TDM samples are true trough concentration representing widely rampant poorly timed sample collection practices.^33-35

Single center retrospective study design with low sample size and primarily white male population limits the study generalizability. Our sensitivity analysis failed to preserve the benefits of AUC-dosing when the events in first 24 hours were excluded from the analysis. Numerically, there were lower incidence of nephrotoxicity events in the post-implementation group; however, low event rate resulting in inadequate power precludes the ability to detect difference. While broth microdilution for MIC determination is recommended, it is not available for use at all facilities and can produce substantially different results depending on the methods utilized. We also did not evaluate the impact of change in renal function as time varying coefficient, so the analysis does not account for change in the patient’s status biasing the observed results toward null. The study results do not apply for the indication bacteremia due to limited patients included. Although the groups were matched using PSs there may be unaccounted confounding variables that can affect the observed results. For example, attrition problem is possible as vancomycin is renally eliminated, it is difficult to assess if elevated levels are results of or cause of the reduced renal function for reasons other than vancomycin toxicity. In addition, exclusion after PS matching may have introduced unequal covariate balance reducing the strength of the observed results. We also did not collect information on APACHE-II or SOFA score or sepsis, however, we report the use of vasopressors, nephrotoxins, and ICU stay. We only evaluated the impact of this dosing strategy on nephrotoxicity and the study was not designed to evaluate impact on clinical effectiveness. We cannot rule out the possibility where MIC may have been 0.5; however, its impact is likely to be insignificant based on our study findings where AUC-guided dosing was associated with less nephrotoxicity.

In conclusion, this study showed that AUC-guided dosing calculated using single trough concentration may result in reduced rate of nephrotoxicity than trough-based dosing. In the future, larger, prospective studies could help identify if there are any clinically significant differences in cumulative doses, 30-day all-cause mortality, and readmission rates.