Vancomycin Plus Piperacillin-Tazobactam and Acute Kidney Injury in Adults: A Systematic Review and Meta-Analysis

Objectives: The objective of this systematic review and meta-analysis was to assess acute kidney injury with combination therapy of vancomycin plus piperacillin-tazobactam, in general, adult patients and in critically ill adults. Rates of acute kidney injury, time to acute kidney injury, and odds of acute kidney injury were compared with vancomycin monotherapy, vancomycin plus cefepime or carbapenem, or piperacillin-tazobactam monotherapy. Data Sources: Studies were identified by searching Pubmed, Embase, Web of Science, and Cochrane from inception to April 2017. Abstracts from selected conference proceedings were manually searched. Study Selection: Articles not in English, pediatric studies, and case reports were excluded. Data Extraction: Two authors independently extracted data on study methods, rates of acute kidney injury, and time to acute kidney injury. Effect estimates and 95% CIs were calculated using the random effects model in RevMan 5.3. Data Synthesis: Literature search identified 15 published studies and 17 conference abstracts with at least 24,799 patients. The overall occurrence rate of acute kidney injury was 16.7%, with 22.2% for vancomycin plus piperacillin-tazobactam and 12.9% for comparators. This yielded an overall number needed to harm of 11. Time to acute kidney injury was faster for vancomycin plus piperacillin-tazobactam than vancomycin plus cefepime or carbapenem, but not significantly (mean difference, –1.30; 95% CI, –3.00 to 0.41 d). The odds of acute kidney injury with vancomycin plus piperacillin-tazobactam were increased versus vancomycin monotherapy (odds ratio, 3.40; 95% CI, 2.57–4.50), versus vancomycin plus cefepime or carbapenem (odds ratio, 2.68; 95% CI, 1.83–3.91), and versus piperacillin-tazobactam monotherapy (odds ratio, 2.70; 95% CI, 1.97–3.69). In a small subanalysis of 968 critically ill patients, the odds of acute kidney injury were increased versus vancomycin monotherapy (odds ratio, 9.62; 95% CI, 4.48–20.68), but not significantly different for vancomycin plus cefepime or carbapenem (odds ratio, 1.43; 95% CI, 0.83–2.47) or piperacillin-tazobactam monotherapy (odds ratio, 1.35; 95% CI, 0.86–2.11). Conclusions: The combination of vancomycin plus piperacillin-tazobactam increased the odds of acute kidney injury over vancomycin monotherapy, vancomycin plus cefepime or carbapenem, and piperacillin-tazobactam monotherapy. Limited data in critically ill patients suggest the odds of acute kidney injury are increased versus vancomycin monotherapy, and mitigated versus the other comparators. Further research in the critically ill population is needed.

Acute kidney injury (AKI) is associated with increased morbidity and mortality (1)(2)(3)(4). 60 There are several definitions, but recent consensus documents focus on three: 1) Risk, Injury, 61 Failure, Loss, End-stage (RIFLE) (5), 2) Acute Kidney Injury Network (AKIN) (6), and 3) Kidney 62 Disease: Improving Global Outcomes (KDIGO) (7), which is a combination of the other two 63 definitions. When defined as an increase in serum creatinine of 0.5mg/dL or greater, one 64 component of the RIFLE definition, AKI increases length of hospital stay by approximately 3.5 65 days and costs by ~$7500 (1). Mortality is also increased approximately 6.5-fold, and increases 66 even more with larger increases in serum creatinine (1). The poor outcomes associated with 67 AKI have also been demonstrated in several studies of critically ill patients (2,4,8,9), with rates 68 of AKI in intensive care unit (ICU) populations ranging from 28-67% (8, 10-12). 69 Historically, risk of AKI during vancomycin treatment has been widely known, and 70 ranged from 5 to 7% (13, 14). Increases in doses and target trough concentrations may be 71 responsible for the recent observed increases in rates of vancomycin-associated AKI, Although there is some controversy over whether vancomycin monotherapy can 73 cause nephrotoxicity or AKI in an otherwise healthy person, it is generally agreed that 74 concomitant nephrotoxic agents, as well as many comorbid conditions and drug exposure 75 factors, such as dosing, trough concentrations, and duration of therapy, increase this risk (15, 76 20-22). Risk factors, including vasoactive medications, hypotension, and increased disease 77 severity, are often associated with the critically ill population, where vancomycin is prevalent 78 (23,24). 79 Since 2011, there have been multiple studies demonstrating an increase in AKI with 80 combination therapy of vancomycin plus piperacillin-tazobactam (25-56). Vancomycin plus 81 piperacillin-tazobactam is one of the most commonly used combinations of antimicrobials with 82 widespread use in hospitals (57). Rates of AKI in these initial studies ranged from 18 to 49% 83 with the combination (32,33,49,52 Titles and abstracts of potentially relevant studies were reviewed. Randomized or 105 observational reports were eligible to be included in the meta-analysis if they: 1) enrolled adult 106 patients (≥18 years old), 2) included patients on concomitant vancomycin and piperacillin-107 tazobactam and either vancomycin alone, vancomycin plus another beta-lactam, or piperacillin-108 tazobactam alone, and 3) nephrotoxicity/ acute kidney injury rates or odds ratios could be 109 extracted for each group. All definitions of AKI that referenced specific changes in serum 110 creatinine (e.g. 1.5-fold or 0.5mg/dL increase), urine output, or need for dialysis/ renal 111 replacement therapy were included. Studies that used a definition referring to an upper limit of 112 normal serum creatinine were excluded. Pediatric studies, case reports/ series, and articles not 113 in English were excluded. Abstracts from conference proceedings were included. In addition to 114 conference abstracts included in the database search, we manually searched abstract 115 meeting for full text abstracts using the keywords vancomycin, piperacillin, or zosyn. Data from 119 final posters were used when available online. Authors were not contacted for missing data. 120

Study Quality 121
The quality of included studies was assessed using the Newcastle-Ottawa quality 122 assessment score (58). Each study was scored from 0 to 9, based on eight criteria covering 123 selection of cohort, comparability of groups, and outcome. Discrepancies between the two 124 authors were resolved by consensus. 125

Data extraction 126
Data collected from each study included author, publication year, study design, location 127 and dates of enrollment, inclusion and exclusion criteria, definition of acute kidney injury used, 128 medications included, and measures of outcomes (e.g. acute kidney injury rates). 129

Outcomes 130
The primary outcome for the meta-analysis was acute kidney injury, as defined by the 131 individual study. Most studies used AKIN, RIFLE, KDIGO, or vancomycin consensus guidelines 132 to define acute kidney injury or nephrotoxicity (5)(6)(7)20). The percentage of patients developing 133 AKI with each drug regimen were calculated, and used to calculate an overall number needed to 134 harm. Time to AKI was extracted from studies when provided for groups of interest. Median and 135 interquartile range were converted to mean and standard deviation using methods from Wan et 136 al.(59) A secondary analysis was performed for critically ill patients, defined as being in an 137 intensive care unit, to determine whether the impact of these medications on AKI was mitigated 138 or enhanced in ICUs. 139 Statistical Analysis 140 AKI rate differences, and corresponding p-values, as well as the number needed to harm 141 were calculated from OpenEpi.(60) Meta-analysis was performed in Review Manager 5.3 142 (RevMan; Cochrane Library, UK) (61). Pooled odds ratios (OR) and 95% confidence intervals 143 (CI) were calculated using the generic inverse variance random effects model for each 144 comparator (vancomycin monotherapy, piperacillin-tazobactam monotherapy, or vancomycin 145 plus cefepime or a carbapenem). Crude odds ratios were calculated from the raw AKI rates in 146 each study. Adjusted odds ratios were used over the crude odds ratio when provided for the 147 groups of interest. Mean difference in time to AKI was calculated using a random effects model. 148 Publication bias was assessed using funnel plots. Heterogeneity was assessed by I 2 statistic 149 and Cochran's Q. A p-value <0.10 was considered statistically significant since Cochran's Q has 150 low power. Sensitivity analyses were performed (1) by removing each study individually in order 151 to determine whether an individual report has higher contribution to the heterogeneity or overall 152 effect estimate, (2) analyzing published studies separately from abstracts, (3) including only 153 high quality reports (Newcastle Ottawa score ≥7), and (4) including only reports that used 154 methods to control for confounding. Reporting for this meta-analysis is in accordance with the 155

Results 158
A flow diagram of the literature search is shown in Figure 1. The search identified 15 159 published studies meeting inclusion and exclusion criteria for the meta-analysis (Supplemental 160 Table 1). Six studies compared vancomycin plus piperacillin-tazobactam to vancomycin 161 monotherapy (26,27,30,32,37,39), while eight studies compared to vancomycin plus 162 cefepime or carbapenem (25,28,29,33,34,36,38,39) and four compared to piperacillin-163 tazobactam monotherapy (30,31,35,37). Three studies had multiple comparisons (30,37,39). 164 One study was excluded from the vancomycin plus cefepime analysis because the data 165 overlapped another study (34,39). We also identified 17 abstracts from conference proceedings 166 (Supplemental Table 2 Tables 1 and 2 is greater than this number. 171 There were significant differences in study populations evaluated. Mean age ranged 172 from 48 to 74, and severity of illness differed between studies (29, 30, 36, 52). There were also 173 differences among the exclusion criteria for studies which included varying serum creatinine 174 values of >1.2mg/dL(34), >1.5mg/dL(26, 40), >2mg/dL (27), and >2.5mg/dL (36), or creatinine 175 clearance values of <30(26), <40(33), and <60mL/min(28, 63) (See Supplemental Tables 1 and  176 2) . Administration of antibiotics was continuous or extended in some patients, but not 177 others.(41) Comorbidities, such as diabetes, infection type, and other concomitant medications 178 frequently play a role in AKI, but were not uniform across studies (27,33). Some studies 179 controlled for confounding factors in their analyses, by matching patients on other risk factors for 180 AKI or using logistic regression (27,32,34,44,53). Not all studies, however, adjusted for the 181 same variables. 182 In all reports evaluated, the rate of acute kidney injury ranged from 5 to 65% for 183 vancomycin plus piperacillin-tazobactam ( Figure 2) (27,29,33,46,51 Time to AKI, in days, was analyzed ( Figure 3). Only studies comparing vancomycin plus 195 piperacillin-tazobactam to vancomycin plus cefepime reported time to AKI separately for each 196 group. Among five studies reporting this outcome, time to AKI was shorter with vancomycin plus 197 piperacillin-tazobactam, but not significantly (mean difference -1.30, 95%CI -3.00-0.41, p=0.14). 198 Among other studies reporting an average time to AKI for all patients, average AKI onset 199 occurred by 8 days (27, 28, 32-34, 37, 38, 45, 47, 50). Unfortunately, some studies only 200 identified AKI within the first 7 days of therapy, or excluded patients with AKI within 48-72h 201 depending on the study's inclusion criteria for minimum antibiotic duration (26,50,51). 202 Vancomycin plus piperacillin-tazobactam increased the odds of AKI versus each 203 comparator. The odds of AKI increased with vancomycin plus piperacillin-tazobactam versus 204 vancomycin monotherapy (OR 3.40, 95%CI 2.57-4.50; Figure 4A). Compared to vancomycin 205 plus cefepime or carbapenem, the OR for AKI with vancomycin plus piperacillin-tazobactam was 206 2.68 (95%CI 1.83-3.91; Figure 4B), and compared to piperacillin-tazobactam monotherapy, the 207 OR was 2.70 (95% CI 1.97-3.69; Figure 4C). Heterogeneity was significant for each of these 208 analyses (I 2 ≥ 53%, p≤0.01). In an analysis separating studies with vancomycin plus cefepime 209 and vancomycin plus carbapenem, no significant differences in the OR for AKI were found (2.39 210 vs 3.46, respectively, p=0.33, see Supplemental material). 211 Among critically ill populations, the odds of AKI vary depending on the comparator 212 antibiotic. One recent study of the critically ill found no significant increase in AKI with the 213 combination of vancomycin and piperacillin-tazobactam compared to vancomycin plus cefepime 214 (29). Another study found an almost 10-fold increase compared to vancomycin monotherapy in 215 patients from a surgical ICU (52). Two studies in patients in burn units also found 7 to 10-fold 216 increases in AKI over vancomycin monotherapy (46,54). The meta-analysis of critically ill 217 patients included three studies comparing to vancomycin alone, three studies comparing to 218 vancomycin plus cefepime or carbapenem, and one study comparing to piperacillin-tazobactam 219 alone, for a total of 968 patients. In the subset of critically ill patients, the odds of AKI compared 220 to vancomycin were increased (OR 9.62; 95%CI 4.48-20.68; Figure 5). The odds of AKI 221 compared to vancomycin plus cefepime or carbapenem, or piperacillin-tazobactam alone were 222 decreased and no longer significantly different. 223 Multiple sensitivity analyses were conducted, which resulted in overall similar odds 224 ratios. In a sensitivity analysis evaluating the removal of individual studies, only Rutter et al. 225 comparing vancomycin plus piperacillin-tazobactam to vancomycin alone resulted in significant 226 changes in the heterogeneity, which accounted for over 2/3rds of patients in this analysis, with a 227 relatively small confidence interval.(37) In a sensitivity analyses looking at published studies 228 versus abstracts, the ORs for published manuscripts were similar to the overall analysis 229 (published and abstracts) for vancomycin monotherapy and vancomycin plus cefepime or 230 carbapenem, but the heterogeneity was lower for published studies (p>0.10; see Supplemental 231 material). The point estimate for published manuscripts was slightly lower for piperacillin-232 tazobactam (1.89 vs 2.70), but heterogeneity was still significant (I 2 =59%, p=0.06). In the quality 233 assessment, the range of NOS scores was between 3 and 9 (maximum of 9; see supplemental 234 Among the critically ill adult population, there was wide variability in the odds of AKI, 258 depending on the comparator medication ( Figure 5). Within each comparator group, however, 259 there was no heterogeneity observed. The meta-analysis subgroup of critically ill patients is 260 relatively small, since not all studies included data specifically on ICU patients, but was able to 261 demonstrate statistically significant results for vancomycin plus piperacillin-tazobactam versus 262 vancomycin monotherapy (OR 9.62;). Only seven studies included critically 263 ill data, with a total of 968 patients. None of these studies included adjusted odds ratios for 264 these patients, so it is possible that risk factors for kidney injury, such as severity or type of 265 illness, contrast media, hypotension, or other factors are responsible or playing a role in these 266 cases of AKI. Randomized controlled trials comparing monotherapy and combination therapy 267 are unlikely, but by comparing vancomycin plus piperacillin-tazobactam to vancomycin plus 268 cefepime or carbapenem, some of the concerns about confounding can be limited. These 269 patients would theoretically have similar risks of sepsis or ICU admission, however this may not 270 eliminate potential confounding entirely. The critically ill subset of this meta-analysis with 271 vancomycin plus cefepime or carbapenem did not demonstrate significant differences in AKI 272 from vancomycin plus piperacillin-tazobactam (OR 1.43; 95% CI 0.83-2.47) which may indicate 273 that these patients are similar, or have more similar risks for AKI. Only one study in the literature 274 search included data on AKI in critically ill patients on piperacillin-tazobactam monotherapy, in 275 patients with intraabdominal infections (35). This study was included as a comparator in the 276 critically ill sub-analysis, but should be considered carefully due to the limited size and lack of 277 similar studies in the meta-analysis. The analysis demonstrates possible differential effects in 278 ICU patients, which should be investigated in future studies. In addition, prospective randomized 279 controlled trials investigating vancomycin plus piperacillin-tazobactam versus vancomycin plus 280 cefepime would be helpful in determining the true effect size. 281 It may be of clinical interest to compare the vancomycin plus cefepime and vancomycin 282 plus carbapenem subgroups. In these analyses, there was no significant difference in the odds 283 of AKI versus vancomycin plus piperacillin-tazobactam. One study included both cefepime and 284 carbapenem, with a wide confidence interval (1.54-33.15) (36), but the chi-square test remained 285 non-significant when removed, indicating no difference between the cefepime and carbapenem 286 subgroups. There were, however, only three studies, and a limited number of patients, that used 287 vancomycin plus carbapenem (see Supplemental material). Consideration may be given to 288 clinical scenarios or select patients in which vancomycin plus cefepime or a carbapenem may 289 be preferable for antibiotic coverage to limit the risk of AKI. 290 Given the number needed to harm of 11, along with the widespread use of this 291 combination therapy, AKI with vancomycin plus piperacillin-tazobactam likely has a large impact 292 on patient outcomes with the increased length of stay, costs, and mortality associated with AKI 293 (1). Although AKI with vancomycin is typically reversible, even transient AKI in critically ill 294 patients has been associated with increased mortality (8). Daily ICU costs can also be much 295 higher, which would increase the costs above the $7500 previously quoted for hospital-wide 296 patients (1). Reducing the use and duration of vancomycin and piperacillin-tazobactam could 297 reduce AKI incidence (31,34). Strategies to aid in this aim include utilizing antimicrobial 298 stewardship policies; such as protocols for using alternative antibiotics when appropriate, 299 institution-specific guidance on when combination therapy is necessary, and institutional 300 antibiograms for susceptibility. Other stewardship programs have utilized antibiotic restriction 301 and time-outs to decrease use of vancomycin and/or piperacillin-tazobactam (31, 65). 302 Stewardship programs have thus demonstrated reduced rates of AKI (27). Most studies in the 303 meta-analysis required at least 48-72h of antibiotic therapy to be included, and the analysis of 304 time to AKI noted an onset within 8 days. Rapid diagnostic test implementation in hospital 305 settings may help to de-escalate from vancomycin plus piperacillin-tazobactam therapy sooner, 306 potentially avoiding AKI.(66) Unfortunately, time to AKI was not available for studies in ICU 307 populations, so further research is needed. 308 The mechanism of increased AKI with vancomycin and piperacillin-tazobactam is not 309 known. Though there are reports of acute kidney injury or acute interstitial nephritis with 310 cefepime, carbapenems, and other non-piperacillin-tazobactam beta-lactams, these reports are 311 rare compared to the studies of piperacillin-tazobactam (67-69). One study also noted that 312 piperacillin-tazobactam had the lowest renal recovery rate (measured by change in creatinine 313 clearance) among beta-lactams tested, indicating possible kidney hazard with piperacillin-314 tazobactam (70). Piperacillin-tazobactam has not traditionally been considered a nephrotoxic 315 medication, however several studies, and the pooled percentages of AKI we calculated, 316 demonstrated increased odds of AKI with piperacillin-tazobactam monotherapy over 317 vancomycin monotherapy (30,41,42,44,49,55). Since both vancomycin and piperacillin-318 tazobactam have been associated with interstitial nephritis, and vancomycin has also been 319 associated with acute tubular necrosis, it is possible that this combination has augmented 320 effects on nephrotoxicity rates ( piperacillin-tazobactam) that may include cefepime or carbapenems. The results, however, of 329 the meta-analyses for vancomycin alone and vancomycin plus cefepime or carbapenem were 330 similar, and any changes from misclassification would likely be small. We cannot rule out 331 publication bias among the included reports (supplemental material). Larger studies and studies 332 indicating a higher risk of AKI with combinations of vancomycin and piperacillin-tazobactam may 333 be more likely to be published than those not demonstrating a significant difference. In the 334 funnel plots in supplemental material, this is indicated by the lack of studies with low ORs and 335 higher standard errors. We chose to present the results of conference abstracts, to see the 336 impact on the overall odds ratio and compare the results. Of course, conference abstracts have 337 limitations, including that they may have been edited before final presentation, they are not 338 always peer-reviewed, and some information may be missing. Additionally, not all abstracts 339 from the included conferences could be accessed in the collections searched. Our results, 340 Abbreviations: VAN+PT, vancomycin plus piperacillin-tazobactam; VAN,vancomycin 390 monotherapy; FEP/CAR, cefepime or a carbapenem; PT, piperacillin-tazobactam monotherapy; 391 SE, standard error; IV, inverse variance; CI, confidence interval; AKI, acute kidney injury 392 393 394 395