Acute kidney injury in patients treated with IV beta-lactam/beta-lactamase inhibitor combinations

Abstract

Study Objective

Increased acute kidney injury (AKI) incidence has been reported in patients receiving piperacillin-tazobactam (PTZ) therapy compared to other beta-lactams. This study sought to determine if the addition of beta-lactamase inhibitors impact AKI incidence by comparing patients treated with PTZ or ampicillin-sulbactam (SAM).

Design

Retrospective cohort study

Setting

Large academic tertiary care hospital

Patients

Overall, 2,448 patients received PTZ (n=1,836) or SAM (n=612) for at least 48 hours between September 1, 2007 and September 30, 2015. Patients were excluded for pregnancy, cystic fibrosis, chronic kidney disease, and initial creatinine clearance (CrCl) < 30 mL/min. Patients were matched on Charlson Comorbidity Index (CCI), initial CrCl, hypotension exposure, various nephrotoxic drug exposures, history of diabetes, heart failure, and hypertension.

Measurements and Main results

AKI occurred in 265 patients at similar rates for both groups (PTZ 11.4% vs SAM 9.2%; p=0.14). After stratifying by vancomycin exposure and controlling for confounders, there was no difference in the risk of AKI for SAM or PTZ (adjusted OR 0.87, 95% CI 0.59–1.25). The addition of vancomycin to PTZ increased the likelihood of AKI compared to PTZ alone (adjusted OR 1.77, 95% CI 1.26–2.46). Concomitant SAM and VAN therapy was not associated with a significant increase in AKI compared to SAM monotherapy (adjusted OR 1.01, 95% CI 0.48–1.97).

Conclusion

Rates of AKI were similar for PTZ and SAM in a matched cohort. The addition of a beta-lactamase inhibitor is not likely the mechanism in the observed increased rates of AKI in patients treated with vancomycin and PTZ.

Background

The use of piperacillin-tazobactam in combination with vancomycin has been associated with elevated rates of acute kidney injury (AKI) in many small retrospective studies. Incidence rates of AKI range from 9.5% to 34.8%, depending on the target population, with critically ill populations having higher incidence of AKI.^1–8 The most common comparator agent in these studies was cefepime^4–6, because piperacillin-tazobactam and cefepime share a similar niche in antipseudomonal therapy. In these comparisons, piperacillin-tazobactam (PTZ) was associated with higher rates of AKI vs cefepime. The mechanism for the increased incidence of AKI is unknown. However, PTZ is unique, when compared to cefepime, in that it contains a beta-lactam as well as a beta-lactamase inhibitor that is structurally similar to other beta-lactams.

Ampicillin-sulbactam (SAM) is another intravenous beta-lactam/beta-lactamase inhibitor combination that is commonly used in practice. No studies examining the difference in AKI incidence between these 2 agents are available. The primary objective of this study was to determine if there is a significant difference in the rate of AKI in patients treated with PTZ or SAM. We hypothesize that these groups have similar rates of AKI and therefore the observed increase in AKI with PTZ is not related to co-administration of a beta-lactam with a beta-lactamase inhibitor.

Methods

This was an institutional review board-approved retrospective cohort study conducted at the University of Kentucky HealthCare between September 1, 2007 and September 30, 2015. Adult patients receiving PTZor SAM for a least 48 hours were included. Exclusion criteria were current treatment with other beta-lactam agents, pregnancy, cystic fibrosis or chronic kidney disease, and an initial creatinine clearance (CrCl) less than 30 mL/min. To account for the effect of vancomycin on AKI, patients were stratified by vancomycin exposure.

The primary outcome was incidence of AKI as defined by the Risk Injury Failure Loss and End-stage (RIFLE) criteria, an established method of identifying AKI.⁹ Creatinine clearance was used as a marker for glomerular filtration rate (GFR) and was estimated by the adjusted Cockcroft-Gault equation.¹⁰ Exposure to other nephrotoxic agents was defined as receipt of at least one dose within the 24 hours prior to treatment initiation through treatment discontinuation. Nephrotoxic agents included in this analysis were aminoglycosides, amphotericin B, angiotensin-converting enzyme inhibitors, angiotensin II receptor antagonists, intravenous radiocontrast dye, loop diuretics, non-steroidal antiinflamatory drugs, calcineurin inhibitors, vancomycin, and vasopressors. Hypotension was defined as a composite of mean arterial pressure less than 65 mmHg, systolic blood pressure less than 90 mmHg, or vasopressor exposure during treatment. The Charlson comorbidity index (CCI) was used to approximate underlying severity of chronic illness.¹¹

Data collected from the University of Kentucky Center for Clinical and Translational Science Enterprise Data Trust (EDT) included demographic data, drug dosing and administration data, laboratory data, and comorbidity data. The EDT contains electronic medical record data from University of Kentucky HealthCare from 2006 to present. Data stored in the EDT includes patient demographics, financial classification (Medicare, Medicaid, private insurance), provider-level details, diagnoses, procedures, lab tests and results, medication administration, visit details (age, length of stay, etc.), and vital signs. Patients were followed from admission to discharge.

Results are reported using descriptive statistics. The Student’s t test or the Wilcoxon’s rank-sum test were used to compare continuous variables. Categorical variables were compared with chi-squared or Fisher’s exact tests as appropriate. Patients in the PTZ and SAM groups were propensity score matched¹⁴ in a 3 to 1 fashion between PTZ and SAM groups based on CCI, initial CrCl (30–59, 60–89, and >90 mL/min), hypotension exposure, exposure to aminoglycosides, amphotericin B, ACE inhibitors, loop diuretics, calcineurin inhibitors, or vancomycin, history of diabetes, heart failure, or hypertension. These variables were selected based on unadjusted associations with AKI in bivariate analysis and were equally weighted in the final propensity score model. Variables were analyzed in bivariate logistic regression with AKI as the response variable. All variables that were significant in simple regressions were included in the initial multivariate logistic regression of AKI. Variables were removed from the multivariate model in a step-wise fashion to minimize the Akaike Information Criterion to maximize model fit for the final multivariate model. Goodness-of-fit was tested with the Hosmer-Lemeshow test and the area under the receiver-operator-characteristic curve (ROC) or c-statistic. All tests were two-tailed, with an alpha of 0.05 considered significant. All data analyses were performed using R v3.1.3 (R Foundation for Statistical Computing, Vienna, Austria) and RStudio v0.98 (RStudio, Inc., Boston, MA).^12,13

Results

Following matching, 2,448 patients were analyzed for the primary outcome, with 1,836 patients receiving PTZ and 612 receiving SAM. Baseline covariates were evenly distributed between groups (Table 1), with the exceptions of CCI (5 [IQR 2–9] vs. 4 [IQR 1–9] in SAM and PTZ groups, respectively; p=0.002) and IV radiocontrast dye exposure (3.3% vs. 5.8% in SAM and PTZ groups, respectively; p=0.02). Overall AKI incidence was similar between SAM (9.2%) and PTZ (11.4%, p=0.15). However, PTZ was associated with higher rates of GFR reductions compared with SAM (>50%: 1.5% vs. 1.1%, p=0.02; and >75%: 1.3% vs. 0.2%, p=0.0001) (Table 2).

Table 1.

Patient demographics among matched cohort

Variable	SAM (N=612)	PTZ (N=1,836)	p
Age (median [IQR])	52 (42–62)	53 (40–63)	0.7
Male Gender	338 (55.2%)	954 (52.0%)	0.2
Caucasian Race	554 (90.5%)	1,648 (89.8%)	0.6
Weight, kg (mean[SD])	79.9 (22.4)	80.5 (23.8)	0.8
BMI (mean[SD])	27.7 (7.0)	27.8 (8.8)	0.9
CCI (median [IQR])	5 (2–9)	4 (1–9)	0.002
Initial CrCl (median [IQR])	100 (77.8–127.5)	103.9 (76.8–130.9)	0.4
Initial CrCl (mL/min)			0.8
30–59	64 (10.5%)	181 (9.9%)
60–89	165 (27.0%)	482 (26.3%)
≥90	383 (62.6%)	1,173 (63.9%)
Hypotension	42 (6.9%)	144 (7.8%)	0.5
Comorbidities
Diabetes	119 (19.4%)	336 (18.3%)	0.6
Heart Failure	28 (4.6%)	56 (3.1%)	0.1
Hypertension	267 (43.6%)	779 (42.4%)	0.6
Concomitant nephrotoxic agents
Aminoglycoside	17 (2.8%)	36 (2.0%)	0.3
Amphotericin B	1 (0.2%)	3 (0.2%)	1
ACE inhibitor	88 (14.4%)	217 (11.8%)	0.1
ARB	25 (4.1%)	59 (3.2%)	0.4
IV Contrast	20 (3.3%)	107 (5.8%)	0.02
Loop diuretic	92 (15.0%)	242 (13.2%)	0.3
NSAID	106 (17.2%)	294 (16.0%)	0.5
Calcineurin inhibitors	21 (3.4%)	57 (3.1%)	0.8
Vancomycin	128 (20.9%)	397 (21.6%)	0.7
Vasopressors	4 (0.7%)	12 (0.7%)	1
Beta-lactam duration of therapy (days; median [IQR])	4 (2–5)	4 (3–6)	<0.00001
Vancomycin duration of therapy (days; median [IQR])	(N=128) 7 (4–11)	(N=397) 7 (4–13)	0.6

Data are N(%) unless specified otherwise. ACE – angiotensin converting enzyme; ARB – Angiotensin II receptor blocker; IQR – interquartile range; NSAID – nonsteroidal anti-inflammatory drug; PTZ – piperacillin-tazobactam; SAM – Ampicillin-sulbactam; SD – standard deviation

Table 2.

Primary outcome

Outcome	SAM	PTZ	p-value
AKI	56 (9.15%)	209 (11.38%)	0.14
Risk	48 (7.84%)	159 (8.66%)
Injury	7 (1.14%)	27 (1.47%)
Failure	1 (0.16%)	23 (1.25%)

AKI – acute kidney injury; PTZ – piperacillin-tazobactam; SAM – ampicillin-sulbactam; Risk: ≥ 25% decrease in CrCl; Injury: ≥ 50% decrease in CrCl; Failure: ≥ 75% decrease in CrCl

Vancomycin exposure was significantly associated with AKI independent of treatment group on bivariate logistic regression (OR 1.07; 95% CI 1.04–1.1) (Table 3). The incidence of AKI was lowest in the SAM monotherapy group (8.9%) and highest in the PTZ plus vancomycin group (18.1%). The rates of AKI for PTZ monotherapy (9.5%) and SAM plus vancomycin (10.2%) were somewhat higher than SAM monotherapy. Following multivariate regression (Table 4), there were no difference in the odds of AKI between SAM and PTZ (aOR 0.87, 95% CI 0.59–1.25), however, the addition of vancomycin to PTZ significantly increased the odds of AKI compared to PTZ monotherapy (aOR 1.77, 95% CI 1.26–2.46). The addition of VAN to SAM therapy did not increase the odds of AKI (aOR 1.01; 95% CI 0.48–1.97).

Table 3.

Differences in AKI stratified by vancomycin exposure for patients treated with piperacillin-tazobactam or ampicillin-sulbactam

Outcome	SAM monotherapy (N=484)	PTZ monotherapy (N=1,439)	p-value	SAM+VAN (N=128)	PTZ+VAN (N=397)	p-value
AKI	43 (8.9%)	137 (9.5%)	0.74	13 (10.2%)	72 (18.1%)	0.038
Risk	38 (7.8%)	112 (7.8%)		10 (7.8%)	47 (11.8%)
Injury	4 (0.8%)	15 (1.0%)		3 (2.3%)	12 (3.0%)
Failure	1 (0.2%)	10 (0.7%)		0 (0%)	13 (3.3%)

AKI – acute kidney injury; PTZ – piperacillin-tazobactam; SAM – ampicillin-sulbactam; VAN – vancomycin; Risk: ≥ 25% decrease in CrCl; Injury: ≥ 50% decrease in CrCl; Failure: ≥ 75% decrease in CrCl

Table 4.

Bivariate and Multivariate AKI associations

	Bivariate			Multivariate
Variable	OR	95% CI	p	aOR	95% CI	p
Treatment characteristics
Treatment group
PTZ	(reference)			(reference)
SAM	0.93	0.64 – 1.32	0.678	0.87	0.59 – 1.25	0.465
SAM+VAN	1.07	0.56 – 1.89	0.815	0.88	0.44 – 1.61	0.691
PTZ+VAN	2.11	1.54 – 2.86	<0.001	1.77	1.26 – 2.46	0.001
Duration of beta-lactam therapy (per day increase)	1.12	1.08 – 1.16	<0.001	1.08	1.04 – 1.12	<0.001
Patient characteristics
Age (per year increase)	1	0.99 – 1.01	0.875
CCI (per point increase)	1.01	0.98 – 1.04	0.348
Male gender	1.19	0.92 – 1.54	0.187
Caucasian race	1.14	0.75 – 1.81	0.57
Baseline creatinine clearance (mL/min)
30–59	(reference)			(reference)
60–59	0.59	0.36 – 0.98	0.039	0.64	0.38 – 1.10	0.1
≥90	1.15	0.76 – 1.80	0.521	1.55	1.00 – 2.50	0.062
Heart failure	2.52	1.45 – 4.18	0.001	2.06	1.11 – 3.66	0.016
Diabetes	1.39	1.02 – 1.88	0.034	1.32	0.94 – 1.83	0.104
Hypertension	1.12	0.87 – 1.45	0.374
Hypotension†	2.05	1.37 – 3.00	<0.001
Concomitant nephrotoxins
Aminoglycoside	2.76	1.40 – 5.10	0.002	1.89	0.88 – 3.78	0.086
Amphotericin B	2.75	0.14 – 21.59	0.381
ACE inhibitor	1.24	0.85 – 1.77	0.239
ARB	0.51	0.18 – 1.15	0.151
IV Radiocontrast Dye	0.94	0.50 – 1.63	0.826
Loop diuretic	3.59	2.68 – 4.79	<0.001	3.06	2.22 – 4.19	<0.001
Calcineurin inhibitor	3.44	2.02 – 5.65	<0.001	4.28	2.42 – 7.35	<0.001
NSAID	0.79	0.54 – 1.12	0.2
Vasopressor	2.77	0.77 – 8.02	0.079

AKI – acute kidney injury; ACE – angiotensin converting enzyme; aOR – adjusted odds ratio; ARB – angiotensin II receptor antagonist; CCI – Charlson Comorbidity Index; CI – confidence interval; NSAID – nonsteroidal anti-inflammatory drug; OR – odds ratio; PTZ – piperacillin-tazobactam; SAM – ampicillin-sulbactam; VAN – vancomycin;

^†Hypotension removed during AIC minimization step of multivariate analysis

Additional factors that independently predicted AKI were duration of beta-lactam therapy, history of heart failure, loop diuretic exposure, and calcineurin inhibitor exposure (Table 4). There was no evidence of overfitting in the model with a Hosmer-Lemeshow p value of 0.37, and the model had an AUC under the ROC of 0.71.

Discussion

In this large retrospective cohort study, the incidence of AKI between two common beta-lactam/beta-lactamase inhibitor combinations was similar after controlling for confounding factors associated with AKI. To our knowledge, this is the first study to examine this relationship.

The literature in this area describes a high variability in the rates of AKI associated with PTZ, mainly as combination therapy with vancomycin. Kim and colleagues reported the rate of PTZ monotherapy-associated AKI to be 15.4%, which was not significantly different from the combination therapy arm (18.8%).⁸ AKI rates when PTZ is combined with VAN range from 9.5 to 34.8%.^1–8 Few studies have compared AKI incidence among different treatment regimens, but the most common comparator is cefepime. Gomes and colleagues found PTZ combined with VAN had significantly higher rates of AKI compared to the combination of cefepime and VAN (34.8% vs 12.5%).⁴ In a study of patients with diabetic osteomyelitis, PTZ-VAN was associated with a non-statistically significant higher rate of AKI (29.3%) compared with cefepime-VAN (13.3%).⁶ Additionally, in critically ill patients, no difference in AKI incidence was noted between PTZ or CFP when combined with VAN.⁵ These studies used varying definitions of AKI and examined relatively small sample sizes. Additionally, although cefepime and PTZ are commonly interchanged clinically, PTZ is distinct because of the addition of a beta-lactamase inhibitor to the beta-lactam. The unadjusted AKI rate in our study for patients in the PTZ arm was 22.2%; however, following matching the AKI rate was only 11.4% in the PTZ group (p=0.15). After stratification by vancomycin use, PTZ-VAN had higher rates of AKI compared to PTZ alone (18.1% vs. 9.5%; aOR 1.77; 95%CI 1.26–2.46), which is consistent with prior literature.

Ampicillin-sulbactam is another beta-lactam/beta-lactamase inhibitor combination used intravenously for the treatment of a variety of infections. Nephrotoxicity data for SAM are limited; however, one small study of high-dose SAM for multidrug resistant Acinetobacter baumanii pneumonia found AKI rates of approximately 15.3%.¹⁵ Another study, examining SAM use in multidrug resistant A. baumanii infections found AKI renal failure occurred in 26% of patients.¹⁶ These findings are limited by sample size and selection of critically ill patients, who have higher rates of nephrotoxicity. In contrast, we found that AKI occurred in 9.2% of patients receiving SAM. Distinct data for patients receiving SAM in combination with vancomycin is not readily available from earlier SAM studies. When stratified by vancomycin exposure, we found a numerical, but statistically insignificant, increase in AKI (10.2% SAM-VAN vs 8.9% SAM alone; aOR 1.01, 95% CI 0.48–1.97).

Despite the marked interest in the increase in nephrotoxicity noted with combination PTZ and VAN therapy, there have been no hypothesized pathophysiological mechanisms for this finding. We considered the addition of tazobactam to piperacillin as a possible contributing factor to the increase in AKI due to the administration of two beta-lactam-like agents. This is specifically important when comparing PTZ-VAN with other beta-lactam combinations that contain only a single beta-lactam agent, such as cefepime or meropenem. Nephrotoxicity data for beta-lactamase inhibitors administered alone are lacking. Ampicillin-sulbactam is the only beta-lactam/beta-lactamase inhibitor agent commonly used as an alternative to PTZ at our institution. Our findings demonstrate that rates of AKI are similar among beta-lactam/beta-lactamase inhibitor combinations at our institution, and that the combination of vancomycin and piperacillin-tazobactam is a major factor in AKI.

This study is not without limitations. While we employed a robust analysis via matching patients on several possible confounders, there is still the possibility of unmeasured confounders in our sample. However, we did control for many nephrotoxic exposures, such as hypotension and other nephrotoxic drug administration, which should explain the majority of confounding in this study. Additionally, we attempted to control for the temporal relation of nephrotoxic exposure to the treatment window of the study agents. For other nephrotoxic agents, dose-response relationships were not assessed and all exposures were defined as receipt of at least one dose within 24 hours prior to initiation of study agents. This may overestimate the impact of those exposures on AKI, which in turn would bias our results towards the null hypothesis. Between-group differences in chronic illness, as assessed by the CCI, could bias results suggesting that SAM is more nephrotoxic than PTZ. However, our results show the opposite. Critical illness is not well captured by the CCI, and there is a chance that there was a higher proportion of critically ill patients in the PTZ arm. To counter this, we matched on presence of hypotension during the treatment period and baseline severity of illness. Finally, it is unclear if the nephrotoxic potentials of the beta-lactam agents are similar. Due to the timeframe of this study, no patients received piperacillin monotherapy, which precludes any inference regarding the additional nephrotoxic potential of tazobactam. Further prospective studies of combination antimicrobial chemotherapy are warranted, as are animal and human studies of the mechanism for increased nephrotoxicity.

Conclusion

The rates of AKI for piperacillin-tazobactam and ampicillin-sulbactam were similar in our large matched cohort study. Additionally, concomitant vancomycin exposure was associated with significant increases in AKI incidence. The magnitude of increase was significantly different for piperacillin-tazobactam compared to ampicillin-sulbactam.