Systematic Review and Metaanalysis of Acute Kidney Injury Associated With Concomitant Vancomycin and Piperacillin/Tazobactam

Abstract

Concomitant vancomycin and piperacillin/tazobactam (PT) may be associated with increased acute kidney injury (AKI) compared to vancomycin without PT. Medline, Cochrane, and Scopus were searched through October 2016 using “vancomycin,” “piperacillin,” “tazobactam,” and “AKI,” “acute renal failure,” or “nephrotoxicity.” A registered meta-analysis (PROSPERO: CRD42016041646) with relevant scenarios was performed. Fourteen observational studies totaling 3549 patients were analyzed. Concomitant vancomycin and PT was associated with AKI in unadjusted odds ratio (OR, 3.12; 95% confidence interval [CI], 2.04–4.78) and in adjusted OR (aOR, 3.11; 95% CI, 1.77–5.47) analyses. Similar findings were seen assessing studies in adults (aOR, 3.15; 95% CI, 1.72–5.76), children (OR, 4.55; 95% CI, 2.71–10.21), and when <50% of patients received care in an intensive care unit (aOR, 3.04; 95% CI, 1.49–6.22) but not ≥50% (aOR, 2.83; 95% CI, 0.74–10.85). Increased AKI with concomitant vancomycin and PT should be considered when determining beta-lactam therapy.

acute kidney injury, nephrotoxicity, vancomycin, piperacillin-tazobactam.

Acute kidney injury (AKI) may occur in up to 23% of hospitalized patients with an associated mortality of 11%. In the intensive care unit (ICU), AKI rates are increased, with documented incidence of up to 66% and a comparable increase in ICU mortality [1]. Many medications, including antimicrobials such as vancomycin, contribute to the development of AKI [2]. Rates of vancomycin-associated nephrotoxicity may be as high as 34.6% in patients who receive daily doses of >4 g [3]. In addition to higher vancomycin doses, risk factors for vancomycin-associated nephrotoxicity are multifaceted and include increased body weight, extended duration of therapy, history of kidney disease, higher severity of illness, hypotension, elevated trough concentrations, and receipt of concomitant nephrotoxic medications including other antimicrobials [4–7].

Historically, beta-lactam antibiotics have been independently associated with nephrotoxicity, often characterized by an acute interstitial nephritis [2]. Until recently, the literature on a potential synergistic nephrotoxicity associated with concomitant vancomycin and antipseudomonal beta-lactam therapy (eg, piperacillin/tazobactam [PT]) was limited [8–23]. Given the commonality of empirical combination therapy with vancomycin and beta-lactams, the possible potentiation of vancomycin nephrotoxic effects is a clinically valuable question and may represent a modifiable risk factor for vancomycin-associated nephrotoxicity. The purpose of this systematic review and metaanalysis was to determine if concomitant vancomycin and PT is associated with a greater incidence of AKI compared to vancomycin with any other beta-lactam other than PT or as monotherapy.

MATERIALS AND METHODS

Search Strategy and Selection Criteria

Observational and controlled studies from January 1995 to October 2016 were retrieved from the PubMed, Cochrane Controlled Trial Registry, and Scopus databases. The search performed included the following terms: “piperacillin” and “tazobactam” and “vancomycin” as well as “acute kidney injury” or “acute renal failure” or “nephrotoxicity.” References from the bibliographies of the studies retrieved from the literature search were reviewed to identify additional studies. After studies were selected, their references were reviewed for potential inclusion. Studies written in languages other than English and Chinese and those presented solely as abstracts at scientific conferences were not considered for inclusion in this analysis.

Study Selection

Two authors (D. A. H., M. N. S.) reviewed all study abstracts. A study was considered eligible for inclusion if the odds ratio (OR) for AKI with PT added to vancomycin therapy could be extracted or determined from available data. When data relevant to this review were unavailable, further details were requested from study authors whenever possible.

Data Extraction, Outcomes, and Data Analyses

Data extracted from the identified studies included clinical setting, AKI definition, inclusion and exclusion criteria, number of study participants, comparator group, age, severity of illness, ICU residence, receipt of concomitant nephrotoxins, length of vancomycin and PT therapies, vancomycin trough concentrations, and patient outcomes, if available. Two authors (D. A. H., M. N. S.), using quality assessment tools from the National Heart, Lung, and Blood Institute, independently assessed study quality as good, fair, or poor [24]. Any discrepancies were discussed between all investigators.

Outcomes

The primary outcome was incidence of AKI among patients who received concomitant vancomycin and PT relative to those who received vancomycin without PT. This outcome was evaluated for all studies for which an OR and an adjusted OR (aOR) could be calculated. Since the definition of AKI could differ across the studies, the definition of AKI used in the primary outcome analysis from each reviewed study was used. No secondary outcomes were examined.

Data Synthesis

Effect size was calculated as an OR and 95% confidence interval (CI) for AKI, comparing patients who received concomitant vancomycin and PT to those who received vancomycin without PT. Pooled estimates of both crude ORs and aORs were calculated. Cochran Q test was used to determine homogeneity of study results. The degree of inconsistency between the studies was assessed using I², which describes the percentage of total variation across studies that is due to heterogeneity rather than sampling error. Because considerable statistical heterogeneity was present among the selected studies (I² >75%), the treatment effect was determined by using a random-effects model that used the DerSimonian and Laird method with inverse variance weighting. Both publication and sample size biases were graphically assessed using funnel plots, and statistically significant biases were determined using Egger’s test. To assess the potential sources of heterogeneity, subgroup analyses stratified by the following were determined a priori: patients’ age: adults aged ≥18 years] vs children aged <18 years; comparator groups: vancomycin alone vs vancomycin with another beta-lactam; study quality: good vs fair or poor; inclusion of critically ill patients: ≥50% of patients with ICU admittance vs <50%; concomitant nephrotoxin use; and declared conflict of interest. The subgroup for concomitant nephrotoxin use was not assessed because concomitant nephrotoxin use was inconsistently reported. The subgroup for declared conflict of interest was not assessed because no authors of the included studies had current declared conflicts of interest with pharmaceutical companies. Data analyses were performed using StataSE 14. For all analyses, P ≤ .05 (2-sided) was considered statistically significant. The systematic review and meta-analysis was registered (PROSPERO: CRD42016041646).

RESULTS

Our literature search identified 307 studies, of which 16 were reviewed and 14 were included in the metaanalysis (Table 1) [8–21]. Two studies were excluded because a comparator group was not used (Figure 1) [22, 23]. For 11 studies included in the metaanalysis, an aOR was provided or able to be calculated [8–15, 17–19].

Figure 1.

Flow chart for studies included and excluded from the systematic review and metaanalysis.

Table 1.

Details for Studies Evaluating Acute Kidney Injury With Concomitant Vancomycin and Piperacillin/Tazobactam

Study Characteristic				Patient Characteristics			Antibiotic Characteristics
Study Authors, Type, and Quality	Study Population	Exclusions	AKI Definition (n, % Patients With AKI)	Age (y)	ICU Residence, %	Receipt of Other Nephrotoxins. %	PT Dosing	V Duration	Mean V Trough (mg/dL)
Burgess et al. (2014) [8]; retrospective, unmatched cohort (n = 191); good	Patients ≥18 y; V + PT or V ± another; BL ≥48 h	Patients already experiencing AKI or with an admission diagnosis of CKD	RIFLE: V-PT (15/92, 16.3%) vs V (8/99, 8.08%), P = .041 PT was an independent predictor of AKI (OR, 2.22; 95% CI, 0.89–5.51)	58.4	17.3	75.9	Not provided	Not provided	17
Davies et al. (2016) [9]; retrospective, unmatched cohort (n = 530); good	Patients ≥18 y; V + PT or V alone; treatment duration unknown	Patients receiving RRT prior to first V dose	RIFLE: V-PT (76/372, 20.4%) PT was not an independent predictor of AKI (OR, 0.92; 95% CI, 0.63–1.33)	55	34.4	Not provided	Traditional interval	8.8 d	16.1
Fodero et al. (2016) [10]; retrospective, unmatched cohort (n = 453); good	Patients ≥18 y; V + PT or V ± another; BL ≥48 h	Patients receiving RRT prior to first V dose, having an admission diagnosis of CKD, or receiving another nephrotoxic agent	RIFLE: PT was an independent predictor of AKI (OR, 3.21; 95% CI, 1.43–7.96; P = .03)	68.2	21.4	Exclusion criteria	Traditional and extended interval	7.2 d	13.6
Gomes et al. (2014) [11]; retrospective, matched cohort (n = 224); good	Patients ≥18 y; V + PT or V + FEP ≥48 h	Patients already experiencing AKI, receiving RRT prior to first V dose, not on scheduled V dosing, with CKD diagnosis prior to admission or with febrile neutropenia or meningitis	AKIN: V-PT (20/55, 36.4%) vs V-FEP (6/55, 10.9%) PT was an independent predictor of AKI (OR, 5.67; 95% CI, 1.66–19.33; P = .006)	51.8	52.7	Yes, percentage unknown	Extended interval	6.9 d	13.6
Hammond et al. (2016) [12]; retrospective, matched cohort (n = 122); good	Patients ≥18 y; V + PT or V + FEP ≥48 h	Patients already experiencing AKI, receiving RRT prior to first V dose, not on scheduled V dosing, with CKD diagnosis prior to admission or with febrile neutropenia or myeloma	AKIN: V-PT (16/49, 32.7%) vs V-FEP (21/73, 28.8%) PT was not an independent predictor of AKI (OR, 1.20; 95% CI, 0.55–2.63; P = .06)	54.2	100	1.04 concomitant nephrotoxins	Not provided	Not provided	16.2
Kim et al. (2015) [13]; retrospective, matched cohort (n = 228); good	Patients ≥18 y; V + PT or V alone ≥48 h	Patients receiving RRT prior to first V dose or those admitted or transferred to an ICU or intermediate care unit	RIFLE: V-PT (19/101, 18.8%) vs V (4/101, 4%) PT was an independent predictor of AKI (OR, 5.62; 95% CI, 1.84–17.18; P = .006)	48.7 ± 15.3	0	36.8	Traditional interval	Not provided	Not provided
Navalkele et al. (2016) [14]; retrospective, matched cohort (n = 558); good	Patients ≥18 y; V + PT or V + FEP ≥48 h	Patients with baseline SCr >1.2 mg/dL or receiving RRT at time of initiation of combination therapy	RIFLE: (81/279, 29.0%); AKIN: (89/279, 31.9%); VAN: (67/279, 24.0%) PT was an independent predictor of AKI (HR, 4.3; 95% CI 2.7–6.7; P < .0001)	55.9	21.7	Yes, 1 (IQR, 0–2)	Not provided	Not provided	17.5
Sutton et al. (2015) [15]; retrospective, matched cohort (n = 292); good	Patients ≥18 y; V + PT or V ± another BL ≥48 h	Patients already experiencing AKI, receiving RRT prior to first V dose, or not on scheduled V dosing	RIFLE: (32/292, 11.0%); AKIN: (44/292, 15.1%); VAN: (29/292, 9.9%) PT was an independent predictor of AKI (OR, 3.97; 95% CI, 1.66–9.50; P = .002)	60.3 ± 18.1	Not provided	68.4	Not provided	96 h (72–156)	14.6 (11.3–18.1)
Gao et al. (2015) [16]; retrospective, unmatched cohort (n = 250); fair	Patients ≤18 y; V + PT or V ± another BL ≥48 h	Premature neonates or patients receiving another nephrotoxic agent	RIFLE: V-PT (61/125, 48.8%) vs V (15/125, 12%), P = .001 PT was an independent predictor of AKI (OR, 6.99; 95% CI, 3.67–13.30)	4.5 ± 3.9	100	Exclusion criteria	Not provided	Not provided	Not provided
Knoderer et al. (2015) [17]; retrospective, unmatched cohort (n = 167); fair	Patients 30 d to 17 y; V + PT or V ± another BL ≥8 d	Patients already experiencing AKI or developing AKI prior to day 8	pRIFLE: V-PT (13/69, 18.8%) PT was an independent predictor of AKI (OR, 2.61; 95% CI, 1.02–6.70)	1.87	40.7	71.9	Not provided	13.3	7.5
Meaney et al. (2014) [18]; retrospective, unmatched cohort (n = 125); fair	Patients ≥18 y; V + PT or V ± another BL ≥72 h	Patients already experiencing AKI or with an admission diagnosis of CKD	RIFLE: V-PT (13/58, 22.4%) vs V ± another BL (2/31, 6.5%) PT was an independent predictor of AKI (OR, 5.36; 95% CI, 1.41–20.5; P = .014)	50.9 ± 15.5	0	Yes, percentage unknown	Not provided	137 h (93–204)	13.5 (9.85–16.3)
Moenster et al. (2014) [19]; retrospective, unmatched cohort (n = 139); fair	Patients ≥18 y; V + PT or V + FEP ≥48 h	Patients with CKD diagnosis prior to admission or with febrile neutropenia	RIFLE: V-PT (32/109, 29.3%) vs V-FEP (4/30, 13.3%), P = .09 PT was not an independent predictor of AKI (OR, 3.45; 95% CI, 0.96–12.4; P = .06)	61.8	Not provided	Yes, percentage unknown	Not provided	14 d	15.5
McQueen et al. (2016) [20]; retrospective, cohort (n = 185); fair	Patients ≤19 y; V + PT or V alone ≥48 h	Patients without baseline laboratory values or repeated BUN or SCr values	SCr increase ≥100% or >0.5 mg/dL: V-PT (25/106, 23.6%) vs V (3/79, 3.8%)	5.7	57.8	51.9	Not provided	6.2 d	8.9
Peyko et al. (2016) [21]; prospective, observational cohort (n = 85); poor	Patients ≥18 y; V + PT or V + FEP/ MER ≥72 h	Patients receiving RRT prior to first V dose, having an admission diagnosis of CKD	AKIN: V-PT (22/59, 37.3%) vs V-FEP/MER (2/26, 7.7%) PT was an independent predictor of AKI (OR, 7.13; 95% CI, 1.54–33.15)	74.6	Not provided	35.3	Traditional interval	Not provided	17.1

Abbreviations: AKI, acute kidney injury; AKIN, acute kidney injury network; BL, beta-lactam antibiotic; BUN, blood urea nitrogen; CI, confidence interval; CKD, chronic kidney disease; FEP, cefepime; HR, hazard ratio; ICU, intensive care unit; IQR, interquartile range; MER, meropenem; pRIFLE, pediatric risk, injury, failure, loss, and end-stage renal disease; OR, odds ratio; PT, piperacillin/tazobactam; RRT, renal replacement therapy; RIFLE, risk, injury, failure, loss, and end-stage renal disease; SCr, serum creatinine; V, vancomycin; VAN, 2009 consensus vancomycin guideline.

Eleven of the studies included only adult patients [8–15, 17, 18] and 3 included only children [16, 19, 20]. The comparator groups were vancomycin with another beta-lactam in 6 studies [8, 10, 15–18], vancomycin with cefepime in 4 studies [11, 12, 14, 19], vancomycin alone in 3 studies [9, 13, 20], and vancomycin with cefepime or meropenem in 1 study [21]. When evaluated according to quality of study, 8 were good [8–15], 5 were fair [16–20], and 1 was poor [21]. There were 4 studies in which >50% of patients received care in an ICU [11, 12, 16, 20], 7 studies in which this was not the case [8–10, 13, 14, 17, 18], and 3 studies in which the location of care was not provided [15, 19, 21].

Primary Unadjusted and Adjusted Analyses

The incidence of AKI with concomitant vancomycin and PT ranged from 11% to 48.8% [8–21]. Ten studies [8–10, 13–19] used RIFLE (risk, injury, failure, loss of kidney function, end-stage renal disease) or pediatric RIFLE classification of AKI as the primary definition [25], 3 studies [11, 12, 21] used the AKI Network classification [26], and 1 used a doubling of the serum creatinine or an increase of 0.5 mg/dL [20].

In the primary analysis of all included studies, concomitant vancomycin and PT was associated with increased AKI (OR 3.12; 95% CI, 2.04–4.78; P < .001) relative to vancomycin regardless of concomitant beta-lactam therapy. The results did not change considerably when the analysis was restricted to studies for which an aOR was provided or calculated (aOR 3.11; 95% CI, 1.77–5.47; P < .001; Figure 2). Results from unadjusted and adjusted analyses are provided in Table 2.

Figure 2.

Forest plot using adjusted odds ratios and a random-effects model approach. Heterogeneity χ² = 71.23 (d.f. = 10); z = 3.93; P < .001; I² (variation in ES attributable to heterogeneity) = 86.0%; estimate of between-study variance τ² = 0.665.

Abbreviations: AKI, acute kidney injury; CI, confidence interval; d.f., degrees of freedom; ES, estimated; OR, odds ratio; PT, piperacillin/tazobactam; V, vancomycin.

Table 2.

Acute Kidney Injury With Concomitant Vancomycin and Piperacillin/Tazobactam in Unadjusted and Adjusted Primary and Subgroup Analyses

Population	Unadjusted Analysis	Adjusted Analysis
All studies [12–25]	OR, 3.12; 95% CI, 2.04–4.78; P < .001	aOR, 3.11; 95% CI, 1.77–5.47; P < .001
Adults [12–19, 21, 22]	OR, 2.71; 95% CI, 1.72–4.27; P < .001	aOR, 3.15; 95% CI, 1.72–5.76; P < .001
Children [20, 23, 24]	OR, 5.26; 95% CI, 2.71–10.21; P < .001	Deferred
Vancomycin and another beta-lactam [12, 14, 19–22]	OR, 3.60; 95% CI, 2.28–5.68; P < .001	aOR, 3.31; 95% CI, 2.13–5.12; P < .001
Vancomycin and cefepime [15, 16, 18, 23]	OR, 2.63; 95% CI, 1.62–4.28; P < .001	aOR, 3.78; 95% CI, 2.48–5.78; P < .001
Vancomycin and cefepime or meropenem [25]	OR, 7.14; 95% CI, 1.54–33.15; P = .012	Deferred
Vancomycin alone [13, 17, 24]	OR, 3.16; 95% CI, 0.67–14.91; P = .146	aOR, 2.50; 95% CI, 0.41–15.44; P = .323
Good quality [12–19]	OR, 2.44; 95% CI, 1.45–4.09; P = .001	aOR, 2.97; 95% CI, 1.53–5.76; P = .001
Fair or poor quality [20–25]	OR, 4.98; 95% CI, 3.32–7.46; P < .001	aOR, 3.53; 95% CI ,1.76–7.06; P < .001
Critically ill [15, 16, 20, 24]	OR, 3.83; 95% CI, 1.67–8.78; P = .002	aOR, 2.83; 95% CI, 0.74–10.85; P = .128
Noncritically ill [12–14, 17, 18, 21, 22]	OR, 2.44; 95% CI, 1.40–4.27; P = .002	aOR, 3.04; 95% CI, 1.49–6.22; P = .002

Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; OR, odds ratio.

Unadjusted and Adjusted Subgroup Analyses

When only adults were considered, results were similar for unadjusted (OR, 2.71; 95% CI, 1.72–4.27; P < .001) and adjusted (aOR, 3.15; 95% CI, 1.72–5.76; P < .001) analyses. When only children were considered, concomitant vancomycin and PT was associated with increased AKI in the unadjusted analysis (OR, 5.26; 95% CI, 2.71–10.21; P < .001). The adjusted analysis was deferred because only 1 study met evaluation criteria.

Concomitant vancomycin and PT was associated with increased AKI in unadjusted and adjusted analyses when the comparator group was vancomycin with another beta-lactam (OR, 3.60; 95% CI, 2.28–5.68; P < .001 and aOR, 3.31; 95% CI, 2.13–5.12; P < .001) and vancomycin with cefepime (OR, 2.63; 95% CI, 1.62–4.28; P < .001 and aOR, 3.78; 95% CI, 2.48–5.78; P < .001). Concomitant vancomycin and PT was associated with increased AKI in an unadjusted analysis when the comparator group was vancomycin with cefepime or meropenem (OR, 7.14; 95% CI, 1.54–33.15; P = .012); an unadjusted analysis could not be performed. However, concomitant vancomycin and PT did not differ in AKI incidence compared to vancomycin alone in unadjusted or adjusted analyses (OR, 3.16; 95% CI, 0.67–14.91; P = .146 and aOR, 2.50; 95% CI, 0.41–15.44; P = .323).

When AKI incidence was evaluated in studies with a quality rating of good, concomitant vancomycin and PT was associated with increased AKI in unadjusted and adjusted analyses (OR, 2.44; 95% CI, 1.45–4.09; P = .001 and aOR, 2.97; 95% CI, 1.53–5.76; P = .001). When only studies with a quality rating of fair or poor were considered, concomitant vancomycin and PT was still associated with increased AKI in unadjusted and adjusted analyses (OR, 4.98; 95% CI, 3.32–7.46; P < .001 and aOR, 3.53; 95% CI, 1.76–7.06; P < .001).

Concomitant vancomycin and PT was associated with increased AKI in unadjusted but not adjusted analyses when evaluated in studies in which at least 50% of patients received care in an ICU (OR, 3.83; 95% CI, 1.67–8.78; P = .002 and aOR, 2.83; 95% CI, 0.74–10.85; P = .128) and in unadjusted and adjusted analyses in studies in which fewer than 50% of patients received care in an ICU (OR, 2.44; 95% CI, 1.40–4.27; P = .002 and aOR, 3.04; 95% CI, 1.49–6.22; P = .002). The adjusted effect estimates for each study were plotted against the standard error for each effect size, as depicted in Figure 3. Significant heterogeneity was observed. Seven of the 11 studies fell outside of the pseudo 95% CI. When the study that used sampling with repeated measures was removed, all 10 studies fell within the pseudo 95% CI (Figure 4).

Figure 3.

Funnel plot using 11 studies with adjusted odds ratios. Abbreviations: aOR, adjusted odds ratio; OR, odds ratio.

Figure 4.

Funnel plot after removal of study with repeated measures. Abbreviations: aOR, adjusted odds ratio; OR, odds ratio.

DISCUSSION

Vancomycin is prescribed frequently for inappropriate indications, especially community-acquired infections [27]. Similarly, antipseudomonal beta-lactam therapy, often PT, is overused in healthcare-associated infections because of the lack of use or understanding of identified risk factors for multidrug-resistant gram-negative microorganisms [28]. The risks and benefits must be assessed when selecting appropriate empirical antibiotic therapy and minimizing unnecessary or extended durations of broad-spectrum antibiotics. When adverse effects from individual or combinations of medications are unknown or underappreciated, it is challenging to assess the true risks and benefits. This unknown or underappreciated potential risk is especially true with nephrotoxicity and concomitant vancomycin and PT.

The findings from this systematic review and metaanalysis suggest that concomitant vancomycin and PT is associated with a greater incidence of AKI compared to vancomycin without PT. This relationship existed in unadjusted and adjusted analyses before subgroups were evaluated and persisted when the analyses were restricted to all available comparator groups other than studies in which at least 50% of patients received care in an ICU in the adjusted analysis and studies in which the comparator group was vancomycin alone in the unadjusted and adjusted analyses.

The mechanism(s) by which vancomycin may cause renal injury is not well understood. Accumulation of vancomycin in the proximal renal tubule may lead to acute tubular necrosis and glomerular destruction [29, 30]. Because semisynthetic penicillins, such as piperacillin, demonstrate high concentrations throughout the nephron, their likely mechanism of nephrotoxicity is acute interstitial nephritis [31]. Antipseudomonal beta-lactams, including cefepime and carbapenems, may cause nephrotoxicity through similar mechanisms as with piperacillin [32, 33]. It is believed that there is an additive effect from acute interstitial nephritis and direct cellular necrosis when a beta-lactam and vancomycin are used concomitantly [4], though the effect may not depend on serum vancomycin concentrations [14].

The incidence of AKI with concomitant vancomycin and PT was greatly dependent on the population evaluated. The highest incidences of AKI were observed in studies in which at least 50% of patients received care in an ICU, which is unsurprising because this population has an increased baseline risk of AKI, independent of vancomycin and PT exposure [34]. While this suggests that concomitant vancomycin and PT may not be responsible for all of the observed AKI, concomitant therapy remained independently associated with higher odds of AKI in unadjusted results and may augment the risk of AKI in this population. Because the adjusted analysis did not find a difference in AKI in this subgroup, further evaluation of AKI should be performed and incorporate the use of other nephrotoxins to improve the sensitivity of these analyses.

Concomitant vancomycin and PT was associated with AKI in the pediatric population in an unadjusted analysis. Pediatric patients may be less susceptible to developing AKI from vancomycin because target trough concentrations generally are lower in pediatric patients than in adults and pediatric patients frequently have fewer disease states that predispose them to developing nephrotoxicity [35]. Higher vancomycin trough concentrations did not impact AKI when PT was used concomitantly in adults [14]; however, this phenomenon has not been evaluated in pediatric patients. Because only 2 studies were assessed and an adjusted analysis was deferred, further evaluation in this patient population is warranted.

Surprisingly, when concomitant vancomycin and PT was compared with vancomycin alone, there was no difference in AKI in the pooled adjusted findings. Even though a random effect model was performed to account for heterogeneity within the studies, there may be inherent differences between patient populations whose infection was managed with vancomycin alone compared with those who received vancomycin and PT that were not assessed in the 3 studies that informed this analysis [9, 13, 20]. Given the small number of patients in the pooled analysis, very wide CI, and incongruence with results when a non-PT beta-lactam was used concomitantly with vancomycin, this finding may be spurious and the result of a type 2 error.

There are several limitations that should be considered when interpreting these results. Most importantly, all of the studies were performed retrospectively and in an observational manner, which removed our ability to establish a causal relationship between concomitant vancomycin and PT and AKI. At this time, our findings suggest that concomitant vancomycin and PT is associated with AKI compared with vancomycin with another beta-lactam antibiotic in select populations.

Important variables that would have allowed for a more complete assessment of AKI were not provided in many of the studies. The duration of vancomycin exposure was unavailable for 6 studies [8, 12–14, 16, 21]. This relationship is well documented [14, 25, 35, 36], and evaluating this variable would have improved confidence in the results. Additionally, the duration and concomitant use of specific nephrotoxic medications were not provided for 5 studies [9–11, 18, 19], which are important considerations. Reducing duration of vancomycin and PT and avoiding concomitant nephrotoxins require further investigation and may reduce the risk of AKI. Six studies included hypotension as a risk factor for AKI [10, 12, 14, 18, 19, 21], and 10 studies included intravenous contrast dye [10–15, 17–19, 21]. These may have affected the rate of AKI in these studies as well. Finally, although the use of extended-duration or traditional-duration infusions of PT were not described in every study, this may not be a variable that affects AKI [37].

There was significant heterogeneity between studies, as evidenced by the I² value of 78.1% in the overall adjusted analysis. A random effects model was used to reduce the impact of this variability, which was most pronounced in the inclusion and exclusion criteria, concomitant antibiotics and nephrotoxins, and definitions for AKI. Because not all studies used a conservative definition of AKI or the vancomycin-associated nephrotoxicity definition from the vancomycin therapeutic guidelines [38], the true incidence of AKI may have been underreported for some studies. Additionally, some studies did not describe which, if any, concomitant antibiotics were given in conjunction with vancomycin [8, 10, 15–18], particularly aminoglycosides. The effects of these unreported concomitant antibiotics were not quantifiable.

Finally, multiple desired secondary outcomes, such as mortality, length of stay, and the use of renal replacement therapy, could not be evaluated because they were not reported in most studies. AKI development is a meaningful outcome, which has been associated with a more than 4-fold increase in mortality for hospitalized patients [1]. Outcomes from AKI caused by concomitant vancomycin and PT, including progression to chronic kidney disease and need for renal replacement therapy, are desirable outcomes for evaluation in future studies.

In conclusion, concomitant vancomycin and PT appears to be associated with a greater incidence of AKI compared to vancomycin without PT. This relationship did not exist when an adjusted analysis was restricted to studies in which at least 50% of patients received care in an ICU. These results may allow for a better informed evaluation of the risks associated with concomitant vancomycin and PT and serve as the impetus for further evaluation into the true mechanisms behind this additive nephrotoxic effect and its potential implications on mortality. Currently, clinicians should recognize and monitor patients for the increased risk of nephrotoxicity and incorporate that consideration into selection of beta-lactam therapy in conjunction with other patient characteristics.

Notes

Potential conflicts of interest. K. T. L. and P. B. B. have received funding from Cubist Pharmaceuticals, now a subsidiary of Merck and Co. P. B. B. receives research support and consultation fees from Allergan, PLC. C. L. is a paid consultant for eMAX Health Systems for unrelated studies. All other authors: No reported conflicts. The authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Author notes

Correspondence: D. A. Hammond, 4301 W. Markam St., Slot 522 Little Rock, AR 72205 ([email protected]).