Risk of acute kidney injury associated with anti-pseudomonal and anti-MRSA antibiotic strategies in critically ill patients.

BACKGROUND: An increased risk of acute kidney injury (AKI) with the widely prescribed piperacillin-tazobactam(PTZ)-vancomycin combination in hospitalized patients has recently been reported, but evidence in ICU patients remain uncertain. This study evaluates the association between the exposure of various broad-spectrum antibiotic regimens with Pseudomonas and/or methicillin-resistance Staphylococcus aureus (MRSA) coverage and the risk of AKI in critically ill patients. METHODS AND
FINDINGS: A retrospective cohort study based on the publicly available MIMIC-III database reporting hospitalization data from ICU patients from a large academic medical center between 2001 and 2012. Adult patients receiving an anti-pseudomonal or an anti-MRSA agent in the ICU for more than 24-hours were included. Non-PTZ anti-pseudomonal agents were compared to PTZ; non-vancomycin agents covering MRSA were compared to vancomycin; and their combinations were compared to the PTZ-vancomycin combination. The primary outcome was defined as new or worsening AKI within 7 days of the antibiotic exposure using an adjusted binomial generalized estimating equation. Overall, 18 510 admissions from 15 673 individual patients, cumulating 169 966 days of antibiotherapy were included. When compared to PTZ, exposure to another anti-pseudomonal agent was associated with lower AKI risk (OR, 0.85; 95% CI, 0.80-0.91; p < .001). When compared to vancomycin, exposure to another anti-MRSA was also associated with lower AKI risk (OR, 0.71; 95% CI, 0.64-0.80; p < .001). Finally, when compared to the PTZ-vancomycin combination, exposure to another regimen with a similar coverage was associated with an even lower risk (OR, 0.63; 95% CI; 0.54-0.73; p < .001). A sensitivity analysis of patients with high illness severity showed similar results.
CONCLUSIONS: These results suggest that the risk of AKI in ICU patients requiring antibiotherapy may be partially mitigated by the choice of antibiotics administered. Further clinical trials are required to confirm these findings.

Introduction

Antibiotics are widely used in intensive care units (ICUs) and help to save millions of lives. In a large international study, approximately 71% of ICU patients received antibiotics during their stay [1], representing a colossal number of patients worldwide exposed to antibiotics each year. Early empirical administration of broad-spectrum antibiotics is critical for the treatment of patients with severe infection. However, multidrug-resistant organisms are increasingly common and associated with a longer length-of-stay and higher mortality [2]. On the other side, inappropriate use of antibiotics is associated with the development of multidrug-resistant organism. The choice of the empirical antibiotic regimen should be individualised according to various factors such as local resistance rate, previous patient’s infections as well as the suspected site of infection [3]. As infections caused by Pseudomonas aeruginosa and methicillin-resistant Staphylococcus aureus (MRSA) are becoming increasingly prevalent, the initial administration of broad-spectrum antibiotics regimen active against these organisms in high-risk patients has been endorsed by the Surviving Sepsis Campaign [4].

However, among the adverse effects of antimicrobial therapies, nephrotoxicity is a well-described complication for some of these antibiotic classes, such as aminoglycosides and vancomycin [5]. More recently, an increased risk of acute kidney injury (AKI) has been associated with the widely used piperacillin-tazobactam (PTZ) and vancomycin combination [6]. However, patients receiving this combination are likely to be severely ill–and consequently more at risk of AKI occurrence–than those who received other antibiotic classes. Moreover, recent groups have reported mixed results when analyzing the risk of AKI in critically ill patients receiving PTZ or PTZ-vancomycin when compared to carbapenems and carbapenem-vancomycin combination [6-8]. Furthermore, the severity of AKI is rarely reported. We have therefore hypothesized that this increased risk of AKI with the combination of vancomycin and PTZ might be limited to non-severe AKI and could be partially attributed to an inhibition of the tubular secretion of creatinine and might therefore represent pseudo-nephrotoxicity [6,8-10]. In this context, an increased risk of severe AKI should not be observed once adjusting for the severity of the illness. As a large proportion of ICU patients with a severe infection will receive an empirical anti-pseudomonal/anti-MRSA antibiotic regimen, confirming the safety of the PTZ-vancomycin is crucial.

This study, using the Medical Information Mart for Intensive Care (MIMIC)-III database, aimed to compare the risk of AKI in critically ill patients receiving PTZ and vancomycin, alone or in combination, to regimen with coverage for either Pseudomonas, MRSA, or both. We hypothesized a limited association with new of worsening AKI and no association with severe AKI requiring kidney replacement therapy (KRT) initiation.

Methods

Design and study population

This is a retrospective cohort study performed on the publicly available de-identified MIMIC-III database, derived from the Beth Israel Deaconess Medical Center’s medical records, Boston, MA. The database contains detailed information from 46,520 patients hospitalized between 2001 and 2012 [11]. The database was approved for research by the Massachusetts Institute of Technology (MIT) and the Beth Israel Deaconess Medical Center institutional review boards. Secondary use of this database does not require informed consent, which was waived from local ethic committee review.

Adult patients of 18 years and older receiving a broad-spectrum antibiotic of interest for ≥24 hours at the ICU were included. Patients with end-stage kidney disease (ESKD) at admission and patients who died within the first 24 hours of ICU admission were excluded. Multiple ICU admissions from the same patient were included if they all met eligibility criteria. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline [12].

The following antibiotics of interest were categorized as follows: anti-pseudomonal (PTZ, ciprofloxacin, aminoglycosides [gentamycin, tobramycin, amikacin], ceftazidime, cefepime, carbapenems [imipenem cilastatin, meropenem] and aztreonam), anti-MRSA (intravenous vancomycin, daptomycin, linezolid) or anti-pseudomonal/anti-MRSA (combination of both classes).

Observations and endpoints

The entire treatment duration for all antibiotics was converted into individual observation periods of 24-h (from 0:00 to 23:59). All periods where any antibiotics of interest were prescribed for at least 24-h (using pharmacy records data) were therefore included. For vancomycin, any additional observation period with a serum vancomycin level ≥10 mg/L were also included, even if no dose was administered at that day. The primary endpoint was new or worsening AKI within 7 days, defined as new-onset AKI or progression to a higher AKI stage for patients already having AKI criteria at the observation period, based on the serum creatinine KDIGO-AKI criteria (as depicted in S1 Fig, Suppl. Material) [13]. Secondary endpoints include new-onset KRT within 7 days and 30 days, a proxy for severe AKI. To explore the creatinine secretion inhibition attributed to PTZ-vancomycin, we compared the changes in serum creatinine and serum urea in the entire cohort, and patients with confirmed AKI (based on KDIGO-AKI staging).

Covariates

Demographic characteristics include age, sex and ethnicity. Comorbidities were determined from the International Classification of Disease (ICD)-9 coding. Positive pressure ventilation was defined as receiving invasive or non-invasive mechanical ventilation documented by the presence of a recorded PEEP value on any blood gas result at the observation period. Leukopenia was defined as a white blood cell count <1.0 x109/mL, and corticosteroid as receiving any equivalent dose of prednisone ≥10mg/day during the antibiotic treatment duration. Active bacteremia was defined as any positive blood culture within three days of the observation period. No imputation was performed for these variables, except for the Sequential Organ Failure Assessment (SOFA) Score [14]. Due to multiple missing values, integration of the SOFA score in the multivariate model was limited. Therefore, missing values from each component of the score were imputed using the median.

Statistical analysis

We used mean values (SD) for continuous variables and proportions for categorical variables for descriptive statistics. All statistical associations used PTZ as reference group for anti-pseudomonal agents, vancomycin for anti-MRSA agents, and PTZ-vancomycin for anti-pseudomonal/anti-MRSA combinations. For the primary analysis, associations between the antibiotic regimen (exposure) and all endpoints were assessed using a generalized estimating equation (GEE) with a logistic (binomial) link function, producing Odds ratios (OR) with confidence intervals. This analysis considers the repeated measure design, implying that 24-h observation periods are not independent of each other and were clustered at the patients’ level. An adjusted model that considered the most relevant factors, including non-traditional variables that showed a strong and clinically meaningful association with the risk of new or worsening AKI, was used for the multivariate model. Therefore, clinical endpoints were adjusted for age, sex, black ethnicity status, comorbidities (CKD, heart failure, liver disease and diabetes), SOFA score, hyperlactatemia, vasopressor requirement, positive pressure ventilation, bacteremia, corticosteroid therapy, leukopenia and antibiotic treatment duration.

Exploratory analyses included the following subgroup analyses: confirmed Pseudomonas spp. infection; confirmed MRSA infection; patients with CKD; antibiotics initiated within the first 48-hours of ICU admission; receiving the same antimicrobial agent for at least 72-hours; SOFA score >5 at ICU admission; excluding observations with known toxic vancomycin levels (>20 mg/L within 48h) and, finally, when excluding aminoglycoside exposure from the analysis. All primary analyses used a level of significance of 0.05. However, to compensate for multiple comparisons and potential type I error, results from all exploratory and subgroup analyses were interpreted as substantially significant only when p<0.01. Data source manipulation and variables selection were performed using the KNIME platform–The Konstanz Information Miner, version 4.3.0 (2021), while all statistical analyses were performed in R, version 4.0.3 (R Project for Statistical Computing) using the R-package “geepack” for general estimating equation [15] and SPSS 27.0 (Armonk, NY, IBM Corp).

Results

Patient selection and characteristics

Of the 46,520 patients reported in the entire MIMIC-III database, we identified 18,510 admissions from 15,673 eligible patients, totalizing 169,966 observation days (Fig 1). The mean (±SD) age was 65 (±16) year, and 43% were female (Table 1). Most patients (55%) were hospitalized in a cardiac or medical ICU, while vasopressors or positive pressure ventilation were required for at least one day during the antibiotic treatment duration in 31% and 34% of all admissions, respectively. An MRSA infection or colonization was confirmed in 1,945 (11%) admissions, while Pseudomonas spp. was identified in 1,263 (6.8%) of them. Most patients (74%) received broad-spectrum antibiotics within the first 48-hours of ICU admission. Regarding the choice of antibiotics, 6,136 (33%) of all admissions received PTZ for at least 24-hours, while 16,105 (87%) received at least one dose of intravenous vancomycin. The count of 24-h observations for each agent is further described in Table 2.

Fig 1

Flowchart of included patients.

Table 1

Baseline characteristics of included patients.

	All patients
Count, n
Patients	15,673
Hospital admission	18,510
Demographics (%)
Age, y (SD)	65 (±16)
Male Sex	10,468 (57)
Non-white ethnicity	4,861 (26)
Diabetes	5,780 (31)
Hypertension	10,281 (56)
Heart failure	6,175 (33)
Liver disease	1,795 (9.7)
CKD, eGFR<60 mL/min‡	3,867 (21)
eGFR<30 mL/min‡	980 (5.3)
eGFR<15 mL/min‡	218 (1.2)
Hospitalization and ICU data (%)
MICU or CCU	10,104 (55)
SICU or CSRU	8,403 (45)
Positive pressure ventilation ¶	6,253 (34)
Vasopressor ¶	5,717 (31)
Lactate >2.4 mmol/L ¶	6,094 (33)
SOFA Score at admission (SD)	6.3 (±3.9)
Hospital LOS, days (SD)	14 (±13)
In-hospital mortality	2,815 (15)
Infection-related factors (%)
Leukopenia	386 (2.1)
Corticosteroid §	4,465 (24)
Confirmed pseudomonas spp. ¥	1,263 (6.8)
Confirmed MRSA ¥	1,945 (11)
Active bacteremia	3,656 (20)
Antibiotics exposure (%)
Started ≤48h of admission	13,612 (74)
Piperacillin-tazobactam	6,136 (33)
Ciprofloxacin	5,633 (30)
Aminoglycoside	1,590 (8.6)
Ceftazidime	1,128 (6.1)
Cefepime	3,389 (18)
Carbapenem	2,394 (13)
Aztreonam	541 (2.9)
Vancomycin	16,105 (87)
Daptomycin	474 (2.6)
Linezolid	1,124 (6.1)

† Based on hospital admission count.

‡ Using the lowest creatinine value available within 3 months before admission (CKD-EPI).

¶ At any time during the entire antibiotic treatment duration.

§ At least 10 mg/day of prednisone (or equivalent) for at least one day of the antibiotic treatment duration.

¥ Represent any positive Pseudomonas spp. and MRSA microbiology event, including colonization and active infection, within the same hospitalization.

Table 2

Risk of new or worsening AKI and KRT associated with exposure to various anti-pseudomonas, anti-MRSA or their combination (multivariate).

	Observation days†		AKI within 7d, OR [95% CI]	New onset KRT within 7d, OR [95% CI]	New onset KRT within 30d, OR [95% CI]
	Investiga-ted drug	Compa-rison
Non-PTZ anti-pseudomonas (REF = PTZ)	73,544	32,648	0.85 [0.80–0.91]***	0.86 [0.69–1.07]NS	0.72 [0.57–0.90]**
Non-vanco anti-MRSA (REF = vancomycin)	10,474	112,938	0.71 [0.64–0.80]***	1.13 [0.80–1.60]NS	0.58 [0.40–0.85]**
Non-PTZ anti-pseudomonas + non-vanco anti-MRSA (REF = PTZ + vancomycin)	6,471	22,873	0.63 [0.54–0.73]***	1.05 [0.61–1.79]NS	0.51 [0.26–1.01]NS

NS: p-value≥.05, *: p-value < .05

**: p-value < .01

***: p-value < .001, AKI: Acute kidney injury, KRT: Kidney replacement therapy, REF: Reference group, PTZ: Piperacillin-tazobactam.

Results reported are Odds ratios with confidence intervals from a generalized estimating equation (binomial GEE) adjusted for: Age, sex, ethnicity, comorbidities (heart failure, liver disease and diabetes), SOFA score, hyperlactatemia, vasopressors, chronic kidney disease, antibiotic treatment duration, active bacteremia, positive ventilation, active corticosteroid therapy and leukopenia. Analyses for all anti-pseudomonal agents were also adjusted for the presence of a concomitant anti-MRSA agent, while analyses for anti-MRSA agents were adjusted for the presence of an anti-pseudomonal agent.

†Observations where both investigated, and comparator antibiotics were concomitantly received and where KRT was ongoing (ie. not at risk of progression) were excluded from the analysis.

Independent factors associated with the risk of AKI

Within the overall cohort, various factors such as heart failure, liver disease, and higher SOFA score at admission were associated with an increased risk of new or worsening AKI within 7 days (Fig 2). Similarly, active bacteremia, vasopressors, positive pressure ventilation requirement, and longer antibiotic duration were also associated with an increased risk of AKI progression at 7 days. In observations reporting exposure to vancomycin, higher serum levels were associated with a stepwise increase in the risk of AKI.

Fig 2

Forest plot of major independent factors associated with new or worsening AKI (primary outcome) within 7 days (univariate).

Results reported are Odds ratios with confidence intervals from a generalized estimating equation (binomial GEE) non-ajusted. NS: p-value≥.05, *: p-value < .05, **: p-value < .01, ***: p-value < .001.

Association between the antibiotic class and the risk of new or worsening AKI

The primary outcome of new or worsening AKI within 7 days occurred in 7,578 (57%), 9,672 (59%) and 6,646 (62%) of admitted patients when exposed to anti-pseudomonal, anti-MRSA or their combination regimens, respectively. Stage 3 AKI or KRT initiation within 7 days occurred in 2,835 (21%) of patients exposed to anti-pseudomonas, in 3,064 (19%) exposed to anti-MRSA and in 2,586 (24%) exposed to both coverage for at least 24 hours (S1 and S2 Tables, Suppl. Material). In the multivariate analysis, when compared to PTZ, exposure to another anti-pseudomonal agent was associated with a lower risk of AKI within 7 days (OR, 0.85; 95% CI, 0.80–0.91; p < .001) (Table 2). When compared to vancomycin, exposure to another anti-MRSA agent was also associated with a lower risk of AKI within 7 days (OR, 0.71; 95% CI, 0.64–0.80; p < .001). Even lower risk of AKI was associated with exposure to the combination of a non-PTZ anti-pseudomonal and non-vancomycin anti-MRSA regimen (OR, 0.63; 95% CI, 0.54–0.73; p < .001). The associations between the risk of the primary outcome and exposure to each antibiotic agent solely or in combination are depicted in Table 3. These findings were also consistent in the univariate analysis and when considering at least 72h of treatment duration (S3 and S4 Tables, Suppl. Material).

Table 3

Frequency of observations and risk of new or worsening AKI within 7 days associated with exposure to various anti-pseudomonas, anti-MRSA or their combination (multivariate).

Frequencies of 24-hour observations reporting exposure to anti-Pseudomonas, anti-MRSA or both for the entire cohort, n (%)
			Anti-MRSA
			Vancomycin	Daptomycin	Linezolid
			118,909 (70)	4,352 (2.6)	8,696 (5.1)
Anti-pseudomonas	Piperacillin tazobactam	41,265 (24)	29,494 (17)	599 (0.4)	1,584 (0.9)
	Ciprofloxacin	31,018 (18)	13,572 (8.0)	528 (0.3)	773 (0.5)
	Aminoglycoside	9,024 (5.3)	4,876 (2.9)	285 (0.2)	579 (0.3)
	Ceftazidime	7,319 (4.3)	4,990 (2.9)	88 (0.1)	363 (0.2)
	Cefepime	21,334 (13)	15,865 (9.3)	616 (0.4)	702 (0.4)
	Carbapenem	21,937 (13)	11,993 (7.1)	1,301 (0.8)	2,317 (1.4)
	Aztreonam	3,770 (2.2)	2,679 (1.6)	243 (0.1)	333 (0.2)
Risk of new or worsening AKI within 7 days associated with each individual antibiotic exposure, Odds ratio [95% CI]
			Anti-MRSA
			Vancomycin	Daptomycin	Linezolid
			REF	0.73 *** [0.61–0.87]	0.72 *** [0.63–0.82]
Anti-pseudomonas	Piperacillin tazobactam	REF	REF	-	-
	Ciprofloxacin	0.91 * [0.84–0.99]	-	0.57 * [0.36–0.92]	0.56 ** [0.38–0.84]
	Aminoglycoside	1.30 *** [1.16–1.46]	-	1.05 NS [0.57–1.91]	0.90 NS [0.59–1.39]
	Ceftazidime	0.83 ** [0.73–0.94]	-	0.33 NS [0.07–1.63]	0.74 NS [0.51–1.08]
	Cefepime	0.91 * [0.84–0.99]	-	0.62 * [0.40–0.97]	0.66 * [0.43–0.99]
	Carbapenem	0.67 *** [0.61–0.74]	-	0.60 ** [0.44–0.81]	0.73 * [0.57–0.93]
	Aztreonam	0.65 *** [0.55–0.78]	-	1.05 NS [0.59–1.88]	0.25 *** [0.16–0.41]
	Anti-pseudomonas		Anti-MRSA		Anti-pseudomonas + Anti-MRSA

NS: p-value≥.05

*: p-value < .05

**: p-value < .01

***: p-value < .001. REF: Reference group. MRSA: Methicillin-resistant staphylococcus aureus.

Results reported are Odds ratios with confidence intervals from a generalized estimating equation (GEE) adjusted for: Age, sex, ethnicity, comorbidities (heart failure, liver disease and diabetes), SOFA score, hyperlactatemia, vasopressors, chronic kidney disease, antibiotic treatment duration, active bacteremia, positive ventilation, active corticosteroid therapy, leukopenia. Analyses for all anti-pseudomonal agents were also adjusted for the presence of a concomitant anti-MRSA agent, while analyses for anti-MRSA agents were adjusted for the presence of an anti-pseudomonal agent.

Observations where both investigated, and reference antibiotics were concomitantly received and where KRT was ongoing (ie. not at risk of progression) were excluded from the analysis.

Importantly, new or worsening AKI occurred within 7 days in 27.2%, 33.3% and 34.9% of observations exposed to PTZ only, to vancomycin only and to the PTZ-vancomycin combination respectively. Therefore, in the multivariate model, when comparing the PTZ-vancomycin combination to PTZ only, there was an increased risk of new or worsening AKI within 7 days (OR, 1.47; 95% CI, 1.36–1.60; p < .001). However, when comparing that combination to vancomycin only, no increased risk was observed (OR, 1.02; 95% CI, 0.96–1.08; p = NS) (Table 4).

Table 4

Risk of new or worsening AKI and KRT associated with exposure to the PTZ-vancomycin combination compared to PTZ or vancomycin individually (multivariate).

	Observation days†		AKI within 7d, OR [95% CI]	New onset KRT within 7d, OR [95% CI]	New onset KRT within 30d, OR [95% CI]
	Investiga-ted drug	Compa-rison
PTZ + Vancomycin (REF = Only PTZ)	28,002	11,265	1.47 [1.36–1.60]***	1.17 [0.90–1.52]NS	1.28 [1.11–1.47]***
PTZ + Vancomycin (REF = Only Vanco)	28,002	86,390	1.02 [0.96–1.08]NS	1.30 [1.09–1.56]**	1.14 [0.94–1.39]NS

NS: p-value≥.05

*: p-value < .05

**: p-value < .01

***: p-value < .001, AKI: Acute kidney injury, KRT: Kidney replacement therapy, REF: Reference group, PTZ: Piperacillin-tazobactam

Results reported are Odds ratios with confidence intervals from a generalized estimating equation (binomial GEE) adjusted for: Age, sex, ethnicity, comorbidities (heart failure, liver disease and diabetes), SOFA score, hyperlactatemia, vasopressors, chronic kidney disease, antibiotic treatment duration, active bacteremia, positive ventilation, active corticosteroid therapy and leukopenia.

†Observations where KRT was ongoing (ie. not at risk of progression) were excluded from the analysis.

Association between the antibiotic class and new-onset of kidney replacement therapy

KRT occurred in 4% to 5% of all patients within 7 days and 30 days of antibiotic exposure for all three groups (S1 Table, Suppl. Material). As shown in Table 2, the association between the antibiotic regimen and the risk of KRT initiation within 7 days did not reach significance for any of the three groups. However, when compared to PTZ, exposure to another anti-pseudomonal agent was associated with a lower risk of requiring KRT initiation within 30 days (OR, 0.72; 95% CI, 0.57–0.90; p < .01). Similarly, exposure to a non-vancomycin anti-MRSA agent was associated with a lower risk of KRT initiation (OR, 0.58; 95% CI, 0.40–0.85, p < .01). Despite a trend toward lower risk of KRT within 30 days, exposure to a non-PTZ or non-vancomycin combination did not reach statistical significance (OR, 0.51; 95% CI, 0.26–1.01; p = NS).

Subgroup analyses

As shown in Fig 3 (and S5 Table, Suppl. Material), we found that associations between the risk of new or worsening AKI within 7 days and the choice of anti-pseudomonal agent, anti-MRSA agent or their combination were consistent among all subgroup analyses, except in patients with CKD or confirmed Pseudomonas infection.

Fig 3

Forest plot of the association between exposure to anti-pseudomonas, anti-MRSA and their combination with new or worsening AKI within 7 days for the entire cohort and all subgroup analyses (multivariate).

Results reported are Odds ratios with confidence intervals from a generalized estimating equation (binomial GEE) adjusted for: Age, sex, ethnicity, comorbidities (heart failure, liver disease and diabetes), SOFA score, hyperlactatemia, vasopressors, chronic kidney disease, antibiotic treatment duration, active bacteremia, positive ventilation, active corticosteroid therapy and leukopenia. Analyses for all anti-pseudomonal agents were also adjusted for the presence of a concomitant anti-MRSA agent, while analyses for anti-MRSA agents were adjusted for the presence of an anti-pseudomonal agent. Using PTZ as reference group for all other anti-pseudomonal agents, vancomycin for all other anti-MRSA agents and the PTZ-vancomycin combination for all other regiment with anti-pseudomonal and anti-MRSA coverage. NS: p-value≥.05, *: p-value < .05, **: p-value < .01, ***: p-value < .001, KRT: Kidney replacement therapy, REF: Reference group, PTZ: Piperacillin-tazobactam.

Exploration of the pseudo-nephrotoxicity

We compared the change in serum urea to the change in serum creatinine at 72-h of exposure to PTZ versus other anti-pseudomonal agents, with or without vancomycin (S6 Table, Suppl. Material). For the overall cohort, we found that exposure to PTZ was associated with an additional increase in serum creatinine by 3.2% (95% CI, 2.3–4.1%, p < .001), but not in serum urea (0.2%; 95% CI, _0.9–1.2%, p = NS) when compared to other anti-pseudomonal agents, with similar results when considering concomitant vancomycin administration. However, in patients who progressed to stage 2 or 3 AKI within 7 days, exposure to the PTZ-vancomycin combination was associated with a higher increase in serum creatinine than other anti-pseudomonas/anti-MRSA regimens (5.6% [95% CI, 3.1–8.2%, p<0.001]), with no significant difference in creatinine elevation in patients with limited stage 1 AKI. Instead, in that group with non-severe AKI, patients exposed to PTZ, with or without concomitant vancomycin, achieved the AKI creatinine elevation criteria despite a reduction in serum urea at 72-h.

Discussion

To our knowledge, this is the most comprehensive ICU study reporting the association of AKI and the use of various anti-pseudomonal agents, anti-MRSA agents, and their combinations. We observed that both PTZ and vancomycin were associated with a higher risk of new or worsened AKI compared to other anti-pseudomonal or anti-MRSA agents, respectively. Moreover, there was a potential synergistic nephrotoxic association with the PTZ-vancomycin combination, where we showed an increased risk of new or worsened AKI when compared to PTZ only, as well as when compared to other regiments with Pseudomonas and MRSA coverage. Furthermore, these associations were robust in sub-group analyses.

An increasing body of observational cohort studies reported an increased risk of AKI with PTZ-vancomycin in hospitalized patients when compared to alternative broad-spectrum beta-lactams, such as carbapenems [6]. In high-risk patients, including those admitted to the ICU, this association was also reported as quite substantial, with Odds ratio up to 2.16 (95% CI, 1.62–2.88) with the PTZ-vancomycin combination [16]. Another group using the WHO pharmacovigilance database reported a similar excess risk of AKI with the PTZ-vancomycin combination (ROR 2.1 [95% CI, 1.8–2.4]), that differed from other vancomycin-containing regimen [17]. In non-ICU studies, the increased risk of AKI attributed to the PTZ-vancomycin combination was mostly limited to non-severe AKI [18-20] and delayed AKI recovery [21]. Until now, most ICU studies were limited in size and none were powered to investigate an association with severe AKI or the receipt of KRT adequately.

Based on findings from these previous observational studies, our initial premise was an expected increased risk of AKI when exposed to PTZ or its combination with vancomycin, but limited to non-severe AKI [6]. Instead, we found a statistically significant increased risk of KRT initiation within 30 days associated with PTZ and vancomycin exposure, with a notable trend toward a similar risk with the PTZ-vancomycin combination. However, as with all observational studies, such findings should be interpreted cautiously as the reported association might be attributable to a residual confounding effect. We cannot extrapolate with certainty these findings resulting from secondary analyses, as many aspects of KRT prescription might have changed since 2001–2012, including the timing of initiation as recently investigated in the STARRT-AKI trial [22]. In addition, the proportion of patients who progressed to the primary endpoint of new or worsening AKI (57.0%) was slightly higher than usually reported in previous ICU cohorts [23], which might be due to the use of KDIGO criteria and selection of relatively sick patients with active infection requiring broad-spectrum antibiotics.

Our findings could not entirely corroborate the concept of pseudo-nephrotoxicity attributed to creatinine secretion inhibition with PTZ and/or vancomycin exposure. First, we showed an additional increase in serum creatinine but no change in serum urea at 72-h when considering the overall cohort as well as the subgroup of patients with severe AKI (S5 Table, Suppl. Material). We also showed, when exposed to PTZ as opposed to other regimens, that patients with non-severe AKI achieved the stage 1 AKI criteria (based on serum creatinine) despite a relative reduction in serum urea, which could be partially attributed to an inhibition of creatinine tubular secretion. However, a moderate elevation in serum creatinine solely due to its inhibition of tubular secretion is unlikely to translate into an increased risk of severe AKI and KRT use. Consequently, the contribution of this possible pseudo-nephrotoxicity is unlikely to explain the strong epidemiological association with severe AKI observed, as well as the presence of higher levels of kidney stress biomarkers previously reported [24,25].

Severe pseudomonas infections are associated with higher risk of therapeutic failure. Some guidelines have therefore recommended the use of double anti-pseudomonal coverage in high-risk patients, such as ventilator-associated pneumonia [3]. This study wasn’t designed to evaluate the additional risk of AKI in patients requiring dual anti-pseudomonal coverage. However, as shown in Fig 3 (and S5 Table, Suppl. Material), among patients with confirmed pseudomonas infection, totalising 16,444 observation days, there was no increased risk of new or worsening AKI within 7 days associated with PTZ exposure. Therefore, in high-risk patients with known pseudomonas infection, PTZ may still represent an appropriate choice a priori.

Various pathophysiological mechanisms have been proposed regarding the PTZ-vancomycin-associated nephrotoxicity [6]. First, as other beta-lactams, PTZ may be associated with allergic acute interstitial nephritis, which could be aggravated in context of repeated exposure in patients previously sensitized to PTZ [18,26]. Second, vancomycin was shown to induce production of reactive oxygen species and to increase mitochondrial and cellular stress [27], which might be exacerbated by an allergic interstitial nephritis (to beta-lactams) and other factors, including the infection itself. Vancomycin has also been associated with cast nephropathy is some reports [6]. However, no definitive mechanism to explain this synergistic toxicity has been confirmed, and histological correlation on kidney biopsies remains sparse [6].

Although we consider the question of PTZ-vancomycin associated nephrotoxicity to be a topical and clinically important issue, clinicians in ICUs have now several choices of antibiotics active against drug resistant gram-negative bacteria such as Pseudomonas spp., and gram-positive bacteria including MRSA. Adequate assessment of the overall risk of AKI attributed to the treatment received should be part of the equation when prescribing any therapy. The PTZ-vancomycin combination remains a pillar of broad spectrum antibiotherapy in critically ill patients. To our knowledge, no published randomized clinical trial has compared the most appropriate anti-pseudomonal/anti-MRSA regimen when considering efficacy and renal safety. However, patients still receiving the PTZ-vancomycin combination might benefit from additional kidney monitoring, including markers of kidney damage and careful dosing of vancomycin to achieve adequate but not excessive trough values.

Our study has several strengths. First, this is the largest cohort to investigate exposure to various anti-pseudomonal and anti-MRSA regimens and the risk of AKI in critically ill patients. In addition, to enabling extensive multivariate adjustment for comorbidities and severity of critical illness, we were also able to perform multiple subgroup analyzes to confirm the robustness of the observed associations. To leverage the large amount of data available, we used a longitudinal repeated measures approach, thereby avoiding a narrow focus on a specific time period during ICU stay and opting to include all periods of antibiotics exposure.

Limitations include potential ambiguity for comorbidities associated with ICD-9 codes. Also, data on medication administration were based on prescriptions rather than on administered drugs, and no dose-effect relationship on AKI risk could be evaluated with the PTZ exposure. Data come from a single center, and results may have been influenced by local practice in regard to antibiotic prescribing. The primary outcome measurement was based on serum creatinine variation only, using KDIGO-AKI criteria, which is sensitive to minor creatinine elevations and might have led to identification of AKI events with no clear clinical significance. Some have reported that the inclusion of urine output might improve AKI detection [28]. However, the pertinence of AKI defined solely by urine output criteria in the context of drug toxicity-associated AKI remains unknown. As with most retrospective studies, we cannot ascertain the clinician’s intent underlying the use of these classes of antibiotics. Due to limited data, we cannot report the site of infection at the time of antibiotic administration, as well as previous exposure to these antibiotics or other beta-lactams. The total fluid balance could not be included due to the statistical design with repeated measures. Any positive blood culture was considered an active bacteremia as the MIMIC-III database did not allow to identify contaminant from true bacteremia. Finally, associations reported with secondary analyses should be interpreted cautiously as with any observational study.

Conclusion

In critically ill patients who received anti-pseudomonal and anti-MRSA antibiotics, exposure to PTZ and vancomycin individually or in combination is associated with an increased risk of new or worsening AKI within one week. As this risk of AKI might be partially mitigated by the choice of antibiotics administered, clinicians should be aware of this added nephrotoxicity when prescribing such combinations. Further clinical trials are required to confirm these findings prospectively.

Example of the follow-up for the primary endpoint.

(PDF)

Click here for additional data file.

Absolute risk of major clinical endpoints for each antibiotic class received by admitted patients.

(PDF)

Click here for additional data file.

Absolute risk of major clinical endpoints for each antibiotic class received by observation periods.

(PDF)

Click here for additional data file.

Risk of new or worsening AKI and KRT associated with exposure to various anti-pseudomonas, anti-MRSA or their combination (univariate).

(PDF)

Click here for additional data file.

Risk of new or worsening AKI and KRT associated with exposure to various anti-pseudomonas, anti-MRSA or their combination (entire cohort and for at least 72h) (multivariate).

(PDF)

Click here for additional data file.

Risk of new or worsening AKI associated with exposure to various anti-pseudomonas, anti-MRSA or their combination (subgroup analyses, multivariate).

(PDF)

Click here for additional data file.

Change in serum creatinine and urea within 72-h associated with exposure to PTZ compared to another anti-pseudomonas with or without vancomycin.

(PDF)

Click here for additional data file.

Transfer Alert

This paper was transferred from another journal. As a result, its full editorial history (including decision letters, peer reviews and author responses) may not be present.

4 Jan 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Eili Y. Klein, PhD

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Additional Editor Comments (if provided):

Sorry for the delay in review. Given the ongoing pandemic I allowed extra time for the reviewers to complete their reviews. That extra time allowed for four reviews. There was a split on the outcome, but I think all agreed that it is an important and useful paper. While the attention to the statistical associations is appreciated, the question was raised by one of the reviewers why propensity scoring of some type was not included as part of the analysis given its wide-spread use in other AKI analyses. Given the uncertainty about the underlying reasons driving prescribing, on could argue that the groups in the paper may not be fully comparable (thus the reason at least one reviewer mentioned the need to discuss an RCT). Adding a propensity scoring algorithm of some flavor would significantly strengthen the paper and tie it to the broad literature about AKI already. Additionally, the question of why the PTZ-vancomycin combination was more nephrotoxic than PTZ alone but not any more nephrotoxic than vanc should be discussed further as mentioned by the reviewers.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Thank you for doing this study and asking me to review. This issue has been topical in recent years and there are several unanswered questions. This article contributes to our understanding of nephrotoxicity with vancomycin and piperacillin-tazobactam in critically ill patients, particularly due to the large number of cases included. I am sure the authors will agree with me that it a prospective RCT is needed to properly answer the question.

My general thoughts about the paper are:

Teasing out a single cause of nephrotoxicity in critically ill patients is hard with multiple confounders

This is demonstrated by Figure S1: 18 of 25 exposures assessed were associated with increased rate of kidney injury

Patients needed only to have 24 hours of antibiotics to be included in this study. I am skeptical that acute kidney injury could be blamed on these antibiotics for such a short timeframe. Admittedly, had the authors chosen a different timeframe their results would have shown higher levels of AKI in these antibiotic groups. I also acknowledge that the duration of Piptaz/Vanco/combination needed to cause AKI is not known

Progressive increase in rates of kidney injury with longer duration of antibiotic therapy is important. As with all elements of this study, the finding has many confounders but I still think it is an interesting and relevant finding.

It is unusual for patients who are on treatment for both P. aeruginosa and MRSA to not be prescribed either Piptaz or vancomycin. In my experience, if there is a concern re nephrotoxicity with combination vancomycin and piperacillin/tazobactam, patients are usually treated with cefepime or meropenem instead of Piptaz. I would be more interested in rates of AKI with cefepime/vancomycin or meropenem/vancomycin compared to Piptaz/vancomycin.

I like the included tables. However, I feel some of the most useful data is buried in the supplementary section, particularly Figure S1 and Tables S1 and S2.

The discussion makes a good point about changes in indications for dialysis in ICU over the study timeframe.

Completely agree that the findings of this study add yet more weight to the requirement for prospective RCTs on this issue rather than further retrospective data

Reviewer #2: Thank you for the opportunity to review, "Risk of Acute Kidney Injury associated with Anti-pseudomonal and Anti-MRSA Antibiotic Strategies in Critically ill Patients."

This is a retrospective cohort study of data from the MIMIC-III database including 18,510 encounters among 15,673 individual adult patients admitted to a single academic center’s ICUs over an 11-year period (2001-2012). The authors analyze the association between treatment with Vancomycin vs another anti-MRSA antibiotic, Piperacillin-Tazobactam (PTZ) vs another anti-Pseudomonal antibiotic, and combination therapy with Vancomycin-PTZ vs any other anti-MRSA-non-PTZ anti-Pseudomonal antibiotic combination with new or worsening acute kidney injury (AKI) within 7 days, according to creatinine-based KDIGO criteria. Secondarily, the authors analyze the association between the above antibiotics and new-onset renal replacement therapy within 7- and 30-days and changes in serum creatinine and serum urea among all study participants. The authors conclude that among ICU patients who received anti-Pseudomonal and anti-MRSA antibiotics, exposure to PTZ and vancomycin, individually or in combination, had an increased risk of new or worsening AKI within one week.

This is a well-performed research study that follows the STROBE criteria and adds to an existing body of evidence regarding the renal risks of exposure to various nephrotoxic medications in critically ill patients.

Major revision suggestion:

My only major criticism is the authors’ decision to not perform propensity score matching in their retrospective analysis of this observational dataset. Many recent well-performed large-scale retrospective cohort studies seeking to elucidate the association between various interventions and acute kidney injury have used this statistical tool to good effect and have been published in PLoS One (PMID 30142190) and other journals (PMID 31952871). The authors should consider whether this would strengthen their analysis by accounting for the numerous patient characteristics that may contribute to clinicians’ decision to choose one antibiotic agent over another in critically ill patients.

Minor revision suggestions include:

- Page 5, line 11: Missing closing parenthesis after “aztreonam”

- Page 14, line 1: Replace “but” with “except in” to make it clearer that these two sub-groups had different outcomes measured. As currently written, this is unclear.

- Page 14, Discussion section: The authors might consider discussing whether their sub-group analysis demonstrating that, among patients with confirmed Pseudomonal infection, administration of PTZ was not associated with worsening AKI might encourage clinicians to more carefully consider which patients are at higher risk for Pseudomonas to help determine when PTZ may indeed be the optimal choice of anti-Pseudomonal antibiotic if it does not have a significant association with new or worsening AKI for patients who end up having a Pseudomonal infection

- Page 17, line 17-19: Are the authors able to determine how often their studied antibiotics were prescribed but not administered? One might assume this would be a very small number and, if the authors are able to report how frequently this occurred, the reader would be better able to determine how much this inherent limitation of a database study should inform their conclusions.

Reviewer #3: This article examines the association of piperacillin-tazobactam with acute kidney injury (AKI) in a large single-center cohort of ICU patients. As stated by the authors, the concept is not new but the current study provides a more granular analysis of this association by controlling for common risk factors for AKI n ICU patients. Few comments and queries, if I may.

1. The finding that PTZ-vancomycin combination was more nephrotoxic than PTZ alone but not any more nephrotoxic than vanc alone is interesting and should be discussed further. How do the authors explain this finding? Does this suggest that when patients develop AKI while on this combination, the relative contribution of vancomycin is higher than PTZ? Does this finding suggest that for patients who--- for whatever reason--- are thought to require vancomycin as a preferred agent, the addition of PTZ should not expose them to higher risk of nephrotoxicity than vanc alone? Similarly, for those who require PTZ as the drug of choice (eg, when risk of cefepime-induced encephalopathy is considerable), the addition of vanc should be avoided unless it also is considered to be the drug of choice?

2. Is there data examining 24, 48 or 72 hrs of the combination vs longer duration as relates to risk of AKI? Please discuss how the results of this study may be used (or cannot be used) to determine if there is a “safe” duration of PTZ-vancomycin combination beyond which the risk of AKI increases significantly.

Others…

1. Abstract, line 17-18. Should be a complete sentence.

2. Page 5, line 16. Should be “was”

3. Page 6, line 13. Active bacteremia was defined as “any positive blood culture…” How were contaminants handled? Consider using the term “true bacteremia” with explanation of excluding contaminants based on clinical grounds.

4. Page 11, Table 2. Consider an alternative to “Intervention” column heading since this is a retrospective (not prospective) study. ? Investigated drugs

Thank you!

Reviewer #4: The manuscript examined the risk of AKI associated with anti-pseudomonal and anti-MRSA antibiotic strategies in critically ill patients. Here are some comments for consideration:

1. Page 3, line 6. May want to further expand on relationship between broad spectrum antibiotics and MDR organisms.

2. Page 3, line 11. Would want to update reference to the new Campaign recommendations for 2021.

3. Page 3, line 21. The study by Blevins A. AAC; 2019; 63: e02658.may be useful to include as that study examined the rates of AKI with different severity in the ICU. The study by Contejean A. JAC 2021; 76: 1311 would also be useful in the discussion on AKI with vancomycin/piperacillin/tazobactam combination.

4. Page 5, line 2. Is there a specific time frame for overlap of antibiotics?

5. Page 5, Observations and endpoints. Would be useful to define severe and non-severe AKI as this was discussed as a limitations of previous studies.

6. Page 5, line 21. Can further clarify day 7 observation period, is that after receipt of combination antibiotics or any antibiotics?

7. Page 6, covariates. Was patients in shock included in the study? If so, where they on vasopressors as this was considered a major risk factor for Aki development.

8. Page 7, line 10. Were comorbidities like shock, other nephrotoxins, and amount of fluids administered included in the covariate?

9. Page 7, line 17. What is defined as toxic vancomycin level?

10. Page 8, table 1. Is there a breakdown of what stage of CKD for the patient population?

11. Page 8, table 1. Is the mean time to treatment and mean time to AKI available?

12. Page 10. Is the data for AKI by severity available?

13. Page 14, line 6. This is referring to Table S5 rather than S6.

14. Page 14, line 11. Can expand on this statement a little more.

15. Page 14, discussion. Can discuss more on the overall incidence rate of AKI in this study compare to other studies. Can also discuss about the difference in outcome when observing the rates of AKI in patient receiving piperacillin vs those who received vancomycin.

16. Page 17, line 21. Can comment more about the limitation of using KDIGO-AKI criteria vs. the other definitions.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

Reviewer #4: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

18 Jan 2022

Editor Comments:

Q: Sorry for the delay in review. Given the ongoing pandemic I allowed extra time for the reviewers to complete their reviews. That extra time allowed for four reviews. There was a split on the outcome, but I think all agreed that it is an important and useful paper. While the attention to the statistical associations is appreciated, the question was raised by one of the reviewers why propensity scoring of some type was not included as part of the analysis given its wide-spread use in other AKI analyses. Given the uncertainty about the underlying reasons driving prescribing, on could argue that the groups in the paper may not be fully comparable (thus the reason at least one reviewer mentioned the need to discuss an RCT). Adding a propensity scoring algorithm of some flavor would significantly strengthen the paper and tie it to the broad literature about AKI already. Additionally, the question of why the PTZ-vancomycin combination was more nephrotoxic than PTZ alone but not any more nephrotoxic than vanc should be discussed further as mentioned by the reviewers.

R: First of all, thank you very much for accepting to review our work. We consider the antibiotics’ associated nephrotoxicity as a relevant clinic topic and data regarding ICU were lacking. We agree with you that Propensity Score (PS) Methods are increasingly used in large observational studies as a useful alternative to more conventional co-variate adjustment. Propensity score (especially methods using PS matching or PS stratification) is a reliable method to provide co-variate balance and is relatively easy to interpret. However, their superiority to standard co-variate adjustment in large dataset remains unclear1.

In our study, a generalized estimating equation (GEE) model has been performed, to allow the use of a repeated measures design clustered at the patient’ level. It means that a single patient, during his entire ICU stay, could have been exposed to multiple antibiotic combinations successively and been therefore classified with multiple exposure at different timepoints. For example, the same patient can be part of the ceftazidime-vancomycin “group” for some 3 observation days, part of the meropenem-vancomycin “group” for 4 other observation days, and finally only received meropenem for the last 2 days of observation. As result, including a patient-level co-variate(s) propensity score matching does not seem methodologically appropriate.

The only PS method that could be integrated into such GEE analysis would be computing a PS for all timepoints included (not per patient) and integrate it as a covariable into the final adjustment model. However, as shown par Elze et al.1, adding the PS as an additional covariate in large dataset (like our study) produces results very similar to standard co-variate adjustment, with similar estimates, SEs and no clear statistical benefits, especially when the high number of events does not limit the number of covariable than can be integrated into the models (as this is the case here). We therefore consider that our primary multivariate model using 15 different co-variates with clinical and strong statistical relevance (based on Table S1, Suppl Material)(e.g. age, sex, ethnicity, CKD, heart failure, liver disease, diabetes, SOFA score, lactatemia, vasopressor requirement, ventilation, bacteremia, corticosteroid therapy, leukopenia and antibiotic treatment duration) was enough to compensate for potential confounders.

1 Elze, M.C. et al. J Am Coll Cardiol. 2017;69(3):345-57

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1:

Q: Thank you for doing this study and asking me to review. This issue has been topical in recent years and there are several unanswered questions. This article contributes to our understanding of nephrotoxicity with vancomycin and piperacillin-tazobactam in critically ill patients, particularly due to the large number of cases included. I am sure the authors will agree with me that it a prospective RCT is needed to properly answer the question.

R: Thank you for reviewing our manuscript. Indeed, we certainly agree with you that prospective trials are lacking.

Q: My general thoughts about the paper are:

Teasing out a single cause of nephrotoxicity in critically ill patients is hard with multiple confounders

This is demonstrated by Figure S1: 18 of 25 exposures assessed were associated with increased rate of kidney injury

Patients needed only to have 24 hours of antibiotics to be included in this study. I am skeptical that acute kidney injury could be blamed on these antibiotics for such a short timeframe. Admittedly, had the authors chosen a different timeframe their results would have shown higher levels of AKI in these antibiotic groups. I also acknowledge that the duration of Piptaz/Vanco/combination needed to cause AKI is not known

Progressive increase in rates of kidney injury with longer duration of antibiotic therapy is important. As with all elements of this study, the finding has many confounders but I still think it is an interesting and relevant finding.

R: We agree that 24h is a short exposure to develop associated nephrotoxicity. However, it should be noted that a GEE (with a repeated measures design, using 24h observations) has been used to compare each antibiotic exposition. It means that, with this analysis, the effect of an exposure (ATB X) on the observed result (rate of AKI) is relatively proportional to the time of the exposure. For example, a patient exposed 7 days to piperacillin will add 7 observations to the “cohort”, while a patient exposed to 24h will add only 1 observation. Therefore, in this GEE the chance of measuring an effect of the exposure on the outcome is proportional to the duration of the treatment receive. Also, allowing short exposure (as 24h observation) has the advantage to identify non-consecutive exposure to the same antibiotic during the same ICU stay. For example, a patient is exposed to 24h of PTZ, then switch to ceftazidime for 6 days to complete his treatment. Two weeks later, he develops a new infection and is re-exposed to PTZ for a total of 5 days. Allowing to capture each day of exposure, even if short in duration, allow us to increase our statistical power to identify small effect size (as the relative increased risk of AKI attributed to that exposure). We also included the duration of treatment as an adjustment co-variate in the multivariate analysis.

In addition, as suggested, we performed a re-analysis of the dataset while excluding patients receiving (in total, during the entire ICU stay) less than 72h of PTZ, Vancomycin or combination of both. As shown below (highlighted in yellow), no significant difference in major outcome results could be seen. These interesting results were included in the Suppl. Material.

Q: It is unusual for patients who are on treatment for both P. aeruginosa and MRSA to not be prescribed either Piptaz or vancomycin. In my experience, if there is a concern re nephrotoxicity with combination vancomycin and piperacillin/tazobactam, patients are usually treated with cefepime or meropenem instead of Piptaz. I would be more interested in rates of AKI with cefepime/vancomycin or meropenem/vancomycin compared to Piptaz/vancomycin.

I like the included tables. However, I feel some of the most useful data is buried in the supplementary section, particularly Figure S1 and Tables S1 and S2.

R: We agree with you that, since the last 5-6 years, clinicians are more aware of the potential risk of nephrotoxicity with the PTZ-vancomycin combination. However, data from this study come from a cohort of patients hospitalised between 2001 and 2012. The potential risk of prescription bias (notoriety bias) could have been less important than if the study was performed on a more recent cohort. Nevertheless, if clinicians were aware of the potential risk of nephrotoxicity and modified their choice of antibiotics based on patients’ relative risk to develop AKI, it should lead toward the null hypothesis, as the overall risk of AKI should have been minimized as high-risk patients may receive non-PTZ-vancomycin combinations. Instead, we found a significant association between the risk of AKI and exposure to PTZ alone or in combination with vancomycin.

Regarding the usefulness of Tables and Figures integrated in the Suppl. Materials, we initially decided to minimize the amount of data showed in the main article to the essential. However, we agree with reviewer #1 that the Figure S1 is highly relevant, especially as it described the Odds associated with all co-variates incorporated into the adjusted analysis used for the primary outcome. As suggested, we decided to transfer that figure into the main text.

Q: The discussion makes a good point about changes in indications for dialysis in ICU over the study timeframe.

Completely agree that the findings of this study add yet more weight to the requirement for prospective RCTs on this issue rather than further retrospective data

R: Thank you very much.

Reviewer #2:

Q: Thank you for the opportunity to review, "Risk of Acute Kidney Injury associated with Anti-pseudomonal and Anti-MRSA Antibiotic Strategies in Critically ill Patients."

This is a retrospective cohort study of data from the MIMIC-III database including 18,510 encounters among 15,673 individual adult patients admitted to a single academic center’s ICUs over an 11-year period (2001-2012). The authors analyze the association between treatment with Vancomycin vs another anti-MRSA antibiotic, Piperacillin-Tazobactam (PTZ) vs another anti-Pseudomonal antibiotic, and combination therapy with Vancomycin-PTZ vs any other anti-MRSA-non-PTZ anti-Pseudomonal antibiotic combination with new or worsening acute kidney injury (AKI) within 7 days, according to creatinine-based KDIGO criteria. Secondarily, the authors analyze the association between the above antibiotics and new-onset renal replacement therapy within 7- and 30-days and changes in serum creatinine and serum urea among all study participants. The authors conclude that among ICU patients who received anti-Pseudomonal and anti-MRSA antibiotics, exposure to PTZ and vancomycin, individually or in combination, had an increased risk of new or worsening AKI within one week.

This is a well-performed research study that follows the STROBE criteria and adds to an existing body of evidence regarding the renal risks of exposure to various nephrotoxic medications in critically ill patients.

R: We would like to thank you for reviewing our work.

Major revision suggestion:

My only major criticism is the authors’ decision to not perform propensity score matching in their retrospective analysis of this observational dataset. Many recent well-performed large-scale retrospective cohort studies seeking to elucidate the association between various interventions and acute kidney injury have used this statistical tool to good effect and have been published in PLoS One (PMID 30142190) and other journals (PMID 31952871). The authors should consider whether this would strengthen their analysis by accounting for the numerous patient characteristics that may contribute to clinicians’ decision to choose one antibiotic agent over another in critically ill patients.

R: Thank you for sharing these two interesting studies. As mentioned above (see response to the Editor), Propensity score matching is a powerful tool to adjust for co-variates and potential confounders. However, the use of a GEE and its repeated measures design did not allow us to consider a PS matching as an option. In addition, the size of the overall cohort, with more than 169 000 observation days, and the number of outcome events makes the usefulness of incorporating a Propensity Score as a co-variate much less clear1, especially as 15 relevant variables have already been included into the multivariable model.

1 Elze, M.C. et al. J Am Coll Cardiol. 2017;69(3):345-57

Minor revision suggestions include:

Q: - Page 5, line 11: Missing closing parenthesis after “aztreonam”

R: Done

Q: - Page 14, line 1: Replace “but” with “except in” to make it clearer that these two sub-groups had different outcomes measured. As currently written, this is unclear.

R: Thank you for the suggestion. Done.

Q: - Page 14, Discussion section: The authors might consider discussing whether their sub-group analysis demonstrating that, among patients with confirmed Pseudomonal infection, administration of PTZ was not associated with worsening AKI might encourage clinicians to more carefully consider which patients are at higher risk for Pseudomonas to help determine when PTZ may indeed be the optimal choice of anti-Pseudomonal antibiotic if it does not have a significant association with new or worsening AKI for patients who end up having a Pseudomonal infection

R: This is a really interesting suggestion, especially as severe confirmed pseudomonal infections are at high risk of therapeutic failure, and some guidelines now recommend dual anti-pseudomonal coverage – where PTZ remains an agent of choice. We added the following lines to the main text (discussion):

“Severe pseudomonal infections are associated with higher risk of therapeutic failure. Some guidelines have therefore recommended the use of double anti-pseudomonal coverage in high-risk patients, such as ventilator-associated pneumonia(23). This study wasn’t designed to evaluate the additional risk of AKI in patients requiring dual anti-pseudomonal coverage. However, as shown in Figure 2 (and Table S5, Suppl. Material), among patients with confirmed pseudomonas infection, totalising 16,444 observation days, there was no increased risk of new or worsening AKI within 7 days associated with PTZ exposure. Therefore, in high-risk patients with known pseudomonas infection, PTZ may still represent an appropriate choice a priori.”

Q: - Page 17, line 17-19: Are the authors able to determine how often their studied antibiotics were prescribed but not administered? One might assume this would be a very small number and, if the authors are able to report how frequently this occurred, the reader would be better able to determine how much this inherent limitation of a database study should inform their conclusions.

R: This is a major limitation of observational studies using retrospective database. It is not possible to quantify the difference between prescribed drugs and administrated drugs in the MIMIC-III database.

We modified the discussion section to further report this limitation:

“[…] Also, data on medication administration were based on prescriptions rather than on administered drugs, which is a limitation of such retrospective studies.”

Reviewer #3:

This article examines the association of piperacillin-tazobactam with acute kidney injury (AKI) in a large single-center cohort of ICU patients. As stated by the authors, the concept is not new but the current study provides a more granular analysis of this association by controlling for common risk factors for AKI n ICU patients. Few comments and queries, if I may.

Q: 1. The finding that PTZ-vancomycin combination was more nephrotoxic than PTZ alone but not any more nephrotoxic than vanc alone is interesting and should be discussed further. How do the authors explain this finding? Does this suggest that when patients develop AKI while on this combination, the relative contribution of vancomycin is higher than PTZ? Does this finding suggest that for patients who--- for whatever reason--- are thought to require vancomycin as a preferred agent, the addition of PTZ should not expose them to higher risk of nephrotoxicity than vanc alone? Similarly, for those who require PTZ as the drug of choice (eg, when risk of cefepime-induced encephalopathy is considerable), the addition of vanc should be avoided unless it also is considered to be the drug of choice?

R: First, thank you for having reviewed our work. Regarding your question referring to the results presented in Table 4. Yes, we have the same interpretation as you: these results showed that the intrinsic nephrotoxicity of vancomycin is probably higher than the nephrotoxicity associated with PTZ exposure individually (which is compatible with previous data). In the univariate analysis (data not showed), the risk of developing the AKI outcome within 7 days occurred in 33.3% of observations with vancomycin only, in 34.9% of observations with the PTZ-Vancomycin combination and in 27.2% of observation with PTZ only. Therefore, once adjusted for the clinical model, no difference was observed between the vancomycin only and PTZ-vanco combination, but a significant difference was found when comparing PTZ-vanco to PTZ only. Here, we can consider that the major trigger of nephrotoxicity is probably the vancomycin itself.

However, when comparing PTZ to another anti-pseudomonal agent individual, as shown in Table 3, or pooled altogether (Table 2) there was an increased risk of AKI when exposed to PTZ compared to all other agents, except for aminoglycosides, where the risk of AKI was higher in the latter.

Thank for reporting this. We adapted the results section of the manuscript to improve that interpretation:

- Adding these event frequencies into the result section

Regarding the last question, results from an observational/retrospective study like this, despite the strong association observed, should be considered with caution. Until the elaboration of definitive prospective trials, no recommendation regarding the optimal MRSA/anti-pseudomonal combination can be elaborated. It is true that the risk of AKI should be part of equation when prescribing antibiotics, but other major aspects should also be integrated: cost, risk of antimicrobial resistance, local availability, etc.

Q: 2. Is there data examining 24, 48 or 72 hrs of the combination vs longer duration as relates to risk of AKI? Please discuss how the results of this study may be used (or cannot be used) to determine if there is a “safe” duration of PTZ-vancomycin combination beyond which the risk of AKI increases significantly.

R: This is a good question. The repeated measures design of the GEE analysis performed did not allow us to adequately evaluate the cumulative risk of AKI for each consecutive day of exposure, but instead for each additional observation day clustered at the patient’s level. Therefore, we cannot determine at which treatment duration, should we stop the PTZ-vancomycin combination to minimize the risk of AKI (time-to-AKI analysis).

Others…

Q: 1. Abstract, line 17-18. Should be a complete sentence.

R: Done

2. Page 5, line 16. Should be “was”

R: Thank you. Done.

3. Page 6, line 13. Active bacteremia was defined as “any positive blood culture…” How were contaminants handled? Consider using the term “true bacteremia” with explanation of excluding contaminants based on clinical grounds.

R: It is not possible to identify true bacteremia from contaminants, as the MIMIC-III database doesn’t mention the clinical relevance of such “positive blood culture”. However, we agree with you that this limitation should be mentioned. We therefore added the following line in the discussion section:

“Any positive blood culture was considered an active bacteremia as the MIMIC-III database did not allow to identify contaminant from true bacteremia”

4. Page 11, Table 2. Consider an alternative to “Intervention” column heading since this is a retrospective (not prospective) study. ? Investigated drugs

R: Thank you for this observation. We changed all columns heading accordingly.

Thank you!

Reviewer #4:

The manuscript examined the risk of AKI associated with anti-pseudomonal and anti-MRSA antibiotic strategies in critically ill patients. Here are some comments for consideration:

Q: 1. Page 3, line 6. May want to further expand on relationship between broad spectrum antibiotics and MDR organisms.

R: Thank you. We added the following lines to the introduction section:

“On the other side, inappropriate use of antibiotics is associated with the development of multidrug-resistant organism. The choice of the empirical antibiotic regimen should be individualised according to various factors such as local resistance rate, previous patient’s infections as well as the suspected site of infection(3).”

Q: 2. Page 3, line 11. Would want to update reference to the new Campaign recommendations for 2021.

R: Thank you for this observation. We updated the reference according to these new recommendations (were published after the initial submission).

Q: 3. Page 3, line 21. The study by Blevins A. AAC; 2019; 63: e02658.may be useful to include as that study examined the rates of AKI with different severity in the ICU. The study by Contejean A. JAC 2021; 76: 1311 would also be useful in the discussion on AKI with vancomycin/piperacillin/tazobactam combination.

R: The study by Blevins et Al. is already cited in this manuscript. Results from this study, including the odds ratio of developing AKI for patients admitted in ICU is reported in the Discussion section.

Regarding the second reference (Contejean et al.), it is a very interesting study using the WHO database of individual case safety reports that also confirmed the excess of AKI associated with the PTZ-vancomycin combination. We would like to thank you for this reference that we included in the manuscript.

Q: 4. Page 5, line 2. Is there a specific time frame for overlap of antibiotics?

R: Antibiotics need to be received for at least 24h to be included as an observation. For example, if a patient received PTZ alone for 3 days, then received vancomycin for 4 days. The patient is considered as receiving PTZ in monotherapy for 3 observation days, then the PTZ-vanco combination for 4 observation days (considering these observations as cluster).

Q: 5. Page 5, Observations and endpoints. Would be useful to define severe and non-severe AKI as this was discussed as a limitations of previous studies.

R: It is an interesting suggestion, as we consider one of the power of our work is to report that severity. As suggested;

- In the Method section, we mentioned that new-onset KRT within 7 and 30 days served as proxy for severe AKI.

- In the Results section, we added the following line: “Stage 3 AKI or KRT initiation within 7 days occurred in 2,835 (21%) of patients exposed to anti-pseudomonas, in 3,064 (19%) exposed to anti-MRSA and in 2,586 (24%) exposed to both coverage for at least 24 hours”

6. Page 5, line 21. Can further clarify day 7 observation period, is that after receipt of combination antibiotics or any antibiotics?

R: The 7 days observation period is the next 7 days following the receipt of one of the antibiotics of interest. Ex: If a patient is exposed to Ceftazidime for 3 consecutive days (D1 to D3). The observation period is the day D1 to D7 for the observation day #1, D2 to D8 for the observation day #2 and D3 to D9 for the observation day #3. To further illustrate this important concept, we added a new Figure in the Suppl. Material reporting the timeline of follow-up for a hypothetic patient.

7. Page 6, covariates. Was patients in shock included in the study? If so, where they on vasopressors as this was considered a major risk factor for Aki development.

R: Yes, vasopressors (at any dose) were administered in 31% of all patients included. This variable was associated with an increased risk of AKI in the univariate analysis (Figure 1) and was therefore included as an adjustment co-variate in the multivariate model.

8. Page 7, line 10. Were comorbidities like shock, other nephrotoxins, and amount of fluids administered included in the covariate?

R: Shock could be defined with the presence of vasopressors (co-variate) as well as using the SOFA Score (also a co-variate in the multivariate model). The amount of fluid administered was not included in the model, nor the fluid balance because the analysis used a repeated measures design (GEE), which means an exposure to a single antibiotic can occurs multiple times during an entire ICU stay (ex: Day 1 to 3, then D31 to 37, then D39 to 47). Fluid balance and volume of fluid administered cannot be easily included in this model, as the same patient is not followed continuously from admission.

We therefore add a new limitation to the discussion section: “The fluid balance could not be included due to the statistical design with repeated measures” As the concept of under-resuscitation and high positive fluid balance are both associate with increased risk of AKI.

Q: 9. Page 7, line 17. What is defined as toxic vancomycin level?

R: More than 20 mg/L. We adapted the text accordingly.

Q: 10. Page 8, table 1. Is there a breakdown of what stage of CKD for the patient population?

R: Yes. We adapted the Table 1 accordingly.

Q: 11. Page 8, table 1. Is the mean time to treatment and mean time to AKI available?

R: No. This variable has not been generated using the original MIMIC III dataset. However, as shown in Table 1, most patients received at least one of the antibiotics of interest within the first 48h (74%).

Q: 12. Page 10. Is the data for AKI by severity available?

R: The primary endpoint is new onset AKI or AKI progression within 7 days. As some patients had already AKI criteria at the time of antibiotics initiation, we considered this primary endpoint definition more clinically relevant, especially in a population at high risk of AKI (from baseline) as in the ICU. However, when considering patients individually, it is possible to identify the maximum AKI stage achieved during the 7days follow-up (Table S1). As suggested above, we included these non-adjusted results for severe AKI events into the Results Section: “Stage 3 AKI or KRT initiation within 7 days occurred in 2,835 (21%) of patients exposed to anti-pseudomonas, in 3,064 (19%) exposed to anti-MRSA and in 2,586 (24%) exposed to both coverage for at least 24 hours (Table S1 and Table S2 in the Suppl. Material)”.

Q: 13. Page 14, line 6. This is referring to Table S5 rather than S6.

R: Thank you for noticing this typo.

14. Page 14, line 11. Can expand on this statement a little more.

R: The text was adapted according to the suggestion in order to improve comprehension of this concept:

- In the Result section, we added: “However, in patients who progressed to stage 2 or 3 AKI within 7 days, exposure to the PTZ-vancomycin combination was associated with a higher increase in serum creatinine than other anti-pseudomonas/anti-MRSA regimens (5.6% [95% CI, 3.1-8.2%, p<0.001]), with no significant difference in creatinine elevation in patients with limited stage 1 AKI. Instead, in that group with non-severe AKI, patients exposed to PTZ, with or without concomitant vancomycin, achieved the AKI creatinine elevation criteria despite a reduction in serum urea within 72h.”

- In the Discussion section: “We also showed that, when exposed to PTZ as opposed to other regimens, patients with non-severe AKI achieved the stage 1 AKI criteria (based on serum creatinine) despite a relative reduction in serum urea, which could be partially attributed to a inhibition of creatinine tubular secretion. However, a moderate elevation in serum creatinine solely due to inhibition of tubular secretion is unlikely to translate into an increased risk of severe AKI and KRT use”

Q: 15. Page 14, discussion. Can discuss more on the overall incidence rate of AKI in this study compare to other studies. Can also discuss about the difference in outcome when observing the rates of AKI in patient receiving piperacillin vs those who received vancomycin.

R: Thank you for this suggestion. We added the following line to the Discussion section:

“In addition, the proportion of patients who progressed to the primary endpoint of new or worsening AKI (57.0%) was slightly higher than usually reported in previous ICU cohorts(23), which might be due to the use of KDIGO criteria and selection of relatively sick patients with active infection required broad-spectrum antibiotics.”

Q: 16. Page 17, line 21. Can comment more about the limitation of using KDIGO-AKI criteria vs. the other definitions.

R: Thank you very much for the suggestion. Modifications were made accordingly. We added: “(Using KDIGO-AKI criteria) which is sensitive to minor creatinine elevations and might have led to identification of AKI events with no clear clinical significance”.

Submitted filename: Response to Reiewer_PONE-D-21-34271.docx

Click here for additional data file.

8 Feb 2022

Risk of Acute Kidney Injury associated with Anti-pseudomonal and Anti-MRSA Antibiotic Strategies in Critically ill Patients

PONE-D-21-34271R1

Dear Dr. Cote,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Eili Y. Klein, PhD

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

1 Mar 2022

PONE-D-21-34271R1

Risk of Acute Kidney Injury associated with Anti-pseudomonal and Anti-MRSA Antibiotic Strategies in Critically ill Patients

Dear Dr. Côté:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Eili Y. Klein

Academic Editor

PLOS ONE