Estimated glomerular filtration rate as a tool for early identification of patients with insufficient exposure to beta-lactam antibiotics in intensive care units

Abstract Background Only about 50% of intensive care unit (ICU) patients reach a free trough concentration above MIC (100% fT > MIC) of beta-lactam antibiotics. Although dose adjustments based on therapeutic drug monitoring (TDM) could be beneficial, TDM is not widely available. We investigated serum creatinine-based estimated GFR (eGFR) as a rapid screening tool to identify ICU patients at risk of insufficient exposure. Method Ninety-three adult patients admitted to four ICUs in southeast Sweden treated with piperacillin/tazobactam, meropenem, or cefotaxime were included. Beta-lactam trough concentrations were measured. The concentration target was set to 100% fT > MICECOFF (2, 4, and 16 mg/L based on calculated free levels for meropenem, cefotaxime, and piperacillin, respectively). eGFR was primarily determined via Chronic Kidney Disease–Epidemiology Collaboration (CKD-EPI) and compared to three other eGFR equations. Data was analysed using logistic regression and receiver operative characteristic (ROC) curves. Results With intermittent standard dosing, insufficient exposure was common in patients with a relative eGFR ≥48mL/min/1.73m2 [85%, (45/53)], particularly when treated with cefotaxime [96%, (24/25)]. This eGFR cut-off had a sensitivity of 92% and specificity of 82% (AUC 0.871, p < 0.001) in identifying insufficient exposure. In contrast, patients with eGFR <48mL/min/1.73m2 had high target attainment [90%, (36/40)] with a wide variability in drug exposure. There was no difference between the four eGFR equations (AUC 0.866–0.872, cut-offs 44–51 ml/min/1.73m2). Conclusion Serum creatinine-based eGFR is a simple and widely available surrogate marker with potential for early identification of ICU patients at risk of insufficient exposure to piperacillin, meropenem, and cefotaxime.


Introduction
About 50% of intensive care unit (ICU) patients suffer from bacterial infections, associated with a 2-3 times increased mortality rate [1,2].Beta-lactam antibiotics are a cornerstone in empiric therapy of severe bacterial infections [2,3].Beta-lactam antibiotics have time-dependent efficacy, and the recommended pharmacokinetic/pharmacodynamic (PK/PD) target is to maintain free (unbound) concentrations above the minimum inhibitory concentration (MIC) of the pathogen during the entire dosing interval (100% fT > MIC) [4].Achieving this target is associated with improved clinical outcomes [5,6].
In many ICUs, patients are given standardised beta-lactam dosing regimens recommended by guidelines relying on data from non-critically ill patients [7-9].However, insufficient exposure has been reported in about 50% of ICU patients in Sweden and other parts of Europe [10][11][12].This is due to use of standardised dosing strategies and divergent pharmacokinetics in ICU patients, which results in very variable drug exposure [4, [11][12][13][14][15].
To ascertain target attainment, national guidelines and expert opinions suggest dose adjustments based on therapeutic drug monitoring (TDM), preferably in combination with model-informed precision dosing including Bayesian estimation [4,16].However, TDM for beta-lactam therapy is performed in approximately 10% of hospitals globally [2], and even if the technical turnaround time is less than an hour, it usually takes longer from sampling to results in clinical practice.Moreover, there is a discussion on the suitable timing for the first sample after treatment initiation where a recent position paper suggests at least 24 h [4].Thus, results may not be available during the initial phase of treatment when adequate exposure is essential [3,17,18].Therefore, tools are needed to identify patients who require dose adjustments at an early stage.Renal function estimation may have potential in these situations.
Glomerular filtration rate (GFR) is commonly estimated with serum creatinine in routine practice.Serum creatinine varies over time for critically ill patients [15], and estimation of GFR based on serum creatinine has been found less accurate in critically ill patients compared to more accurate markers such as iohexol clearance and 24-h creatinine clearance in urine [26].However, more accurate markers are not rapid, and therefore clinically less applicable in the early, critical stages of ICU care [27,28].The potential of serum creatinine-based estimated GFR (eGFR) to predict patients with insufficient beta-lactam exposure at admission to the ICU has not been well studied.Early identification of such patients could aid in selection of early optimal dosing regimens and selection of patients for TDM.
Therefore, the aim of this study was to investigate serum creatinine-based eGFR as a rapid and simple screening tool for identifying ICU patients at risk of insufficient exposure, defined as not reaching a target of 100% fT > MIC, when treated with piperacillin/tazobactam (piperacillin), meropenem, or cefotaxime.

Study design
The patient population has previously been described by Woksepp et al. [10].In summary, adult patients treated with intravenous (IV) beta-lactam antibiotics were included after admission to four ICUs (Kalmar, V€ axj€ o, Link€ oping, and J€ onk€ oping) in southeast Sweden from September 2014 to July 2015.Patients receiving renal replacement therapy during the first 24 h and those not treated with piperacillin/tazobactam (piperacillin), meropenem, or cefotaxime were excluded.

Data collection
Sample collection, drug concentration analyses and data collection have been described in detail previously [10].Serum creatinine was analysed at the inclusion day according to clinical routine at accredited laboratories in each hospital.Drug concentrations were analysed in blood collected just prior to the next dose of antibiotics after inclusion.The samples were centrifuged 2000 x g for 10 min within one hour of sampling, before being stored at −80 0 C and analysed using liquid chromatography-mass spectrometry.The measurement range was 0.2-100 mg/L for piperacillin (PIP) and 0.2-50 mg/L for meropenem (MER) and cefotaxime (CTX).Total beta-lactam therapy duration was recorded.Missing body weight data was imputed if individual weight was recorded the following day, by adjusting the weight using a gaussian-distributed mean weight difference in the population between day 1 and 2. Dosing regimens were recorded and compared to maximum daily dose information found in the product information and in the European Committee on Antimicrobial Susceptibility testing (EUCAST) rationale documents [8,9].Dosing adjustments for renally impaired patients were compared to national recommendations from the Swedish strategic program against antibiotic resistance (STRAMA) [7].

Target selection
The exposure target was set to 100% fT > MIC, a target recommended for ICU patients in a collaboration between multiple international expert groups specialised in intensive care, infectious disease, and therapeutic drug monitoring [4].Targets were based on epidemiological cut-offs (ECOFF), as determined by EUCAST, and were chosen as previously described from a 'worst-case' scenario (MIC ECOFF ); Pseudomonas aeruginosa piperacillin MIC 16 mg/L and meropenem MIC 2 mg/L, and Staphylococcus aureus cefotaxime MIC 4 mg/L [34].The free antibiotic concentration was calculated from measured total concentrations using published values for the free fraction (0.7 for piperacillin, 0.98 for meropenem, 0.6 for cefotaxime) [35][36][37].Insufficient exposure was defined as not reaching the target of 100% fT > MIC ECOFF .The percentages of individuals with calculated free trough concentration below and above 100% fT > MIC ECOFF were recorded.

Ethics
The original study was approved by the Regional Ethical Review Board, Link€ oping University, Sweden (DNR 2014/ 236-31) [10].Analyses in the current study were performed on data collected and pseudonymized in the original study.The objectives of the current study are part of the original ethical approval.

Statistical analysis
Continuous variables are expressed as mean ± standard deviation if normally distributed and as median and interquartile range (IQR) if not normally distributed.Categorical variables are expressed as numbers and percentages.
Differences between groups were analysed with ANOVA or Kruskal-Wallis test for continuous and v 2 test for categorical variables.A p value less than 0.05 was considered to indicate statistical significance.A post hoc analysis was performed for v 2 -test using adjusted standardised residuals and corrected for type I error with Bonferroni [38].Logistic univariate regression (logit model) and receiver operation characteristic (ROC) analysis were performed to investigate eGFR as a tool to detect insufficient exposure.Area under the curve (AUC) of ROC was analysed with DeLong testing and verified with Hanley & McNeil testing.Positive and negative predictive values were calculated, assuming that the study population correctly reflected the prevalence in the population.The v 2 , ANOVA, Kruskal-Wallis and logistic regression analyses were performed with Statistica (TIBCO Software, Version 13.5.0.17) and receiver operation characteristic (ROC) curve analysis was performed using MedCalc (MedCalc Software, Version 20.110).

Dosing according to renal function
Eighty-three of 93 individuals received dosing regimens according to national guidelines [7,9], and nine of the remaining ten received higher dosages than recommended (Table S1).Three of ten individuals with severe   renal impairment had reduced dosages as recommended by guidelines (Table S1) [7].

Discussion
Our results suggest that eGFR in clinical routine care might rapidly identify patients at risk for insufficient exposure to beta-lactam antibiotics.With a cut-off at 48 mL/min/1.73m 2 (relative eGFR) or 58 mL/min (absolute eGFR), insufficient exposure and target attainment were predicted with a sensitivity >90% and a specificity >80%.
The results show a high rate of insufficient exposure, for cefotaxime and piperacillin at eGFR as low as 48 mL/ min/1.73m 2 .Cut-offs for predicting augmented renal clearance have been estimated between 108 and 119mL/min/1.73m 2 [22,23], suggesting that augmented renal clearance alone is not the explanation.This implies that high initial dosing should be considered in patients with mild to moderate renal impairment.At eGFR <48mL/min/1.73m 2 , there was high target attainment, although with wide variability in trough concentrations (Figure 1).The ability of eGFR to identify insufficient exposure to beta-lactam antibiotics in ICU patients was the same, regardless of the eGFR estimation method used (Figure 3).Furthermore, the predictive ability was similar when using absolute eGFR and a higher cut-off, which was expected since the mean body surface area of 1.93 m 2 exceeded 1.73 m 2 (Table 1, Figure 2).
In one of the few studies investigating eGFR as a single predictor of insufficient exposure to beta-lactams in ICU patients [39], Imani et al. reported a lower ability of eGFR as a predictor (AUC 0.76, sensitivity 77%, specificity 65%, p < 0.001) with a higher cut-off of 72 mL/min/ 1.73m 2 than in the present study, using similar inclusion and exclusion criteria.Possible explanations for the differing results could be that Imani et al. used measured MIC in 37% of the cases, which increases the probability of target attainment.Furthermore, target attainment is affected by dosage as well as use of prolonged infusion, neither reported by Imani et al.Prolonged infusion increases the probability of target attainment eight-fold [40].In the present study, the daily dose never exceeded 1.5 times the maximum dose of the product information or EUCAST rationale documents (Table S1).Although information about the type of infusion was not recorded in our study, intermittent injections, or short infusions up to 30 min were clinical routine at the time.This underlines the importance to compare study settings with local clinical settings, such as target selection or dosing, and to validate results before implementation.
The differences in the rates of insufficient exposure between beta-lactams found in the present study have also been reported in another study from Sweden with a similar population and target selection [11], where a higher rate of insufficient exposure was found for cefotaxime (58%) compared to meropenem (30%).A difference between beta-lactam antibiotics is expected, since dosage, and risk factors for insufficient exposure vary between beta-lactam antibiotics with currently used dosing recommendations [8, 34,40].In the present study, individuals with an eGFR above 48 mL/min/1.73m 2 had the highest rate of insufficient exposure when treated with cefotaxime (24/25), with a majority receiving 1 g q8h (Table S1).Since the time of data collection, EUCAST changed the recommendations for cefotaxime to 2 g q8h as all susceptible S. aureus are now reported as 'I' (susceptible at increased exposure).Intermittent dosing of 2 g q8h may, however, still not be enough.A cefotaxime population PK/PD model predicted that only 60% of patients with eGFR >50mL/min treated with 2 g q8h against S. aureus infection reach 100% fT > MIC ECOFF [41].Using the cut-off of >48mL/min/1.73m 2 , in our study, none of the 4 patients with a dose of 2 g q8h reached target (Table S1).This finding must, however, be interpreted with caution due to the very small sample size.
Our finding that patients with eGFR <48mL/min/ 1.73m 2 reached the target is in accordance with previous studies [11,42], and similar findings were observed with other beta-lactam antibiotics [43].We also found that even if low exposures were rare in patients with eGFR <48mL/min/1.73m 2 , high trough concentrations were observed in some patients, suggesting that individual beta-lactam levels cannot be precisely predicted with eGFR alone.This is of concern, since observational studies report a possible association between high trough concentrations and adverse events such as acute kidney failure and encephalopathy [44][45][46][47].To avoid overexposure, many institutional guidelines recommend dosage reductions in patients with renal impairment [7,9].There is an ongoing discussion as to whether dosages should be reduced during the first 48 h of treatment [48].Camargo et al. found an association between dosage reduction and mortality in ICU patients, even after adjusting for co-factors such as disease state and antibiotic class [49].In our study, in only three of ten individuals dosage was reduced in accordance with national guidelines,(Table S1) indicating that physicians may avoid dose reductions in critically ill patients.
Our study has several limitations.Although steadystates were expected to be reached for beta-lactam antibiotics, supported by stable concentrations on three consecutive days [10], some individuals may not have reached steady state.Also, the predictive ability of eGFR in a pre-ICU setting was not investigated, since blood samples for beta-lactam concentrations and creatinine were drawn at the same time at study inclusion.
Serum creatinine was the only marker available for estimating eGFR, with the risk of underestimating GFR in ICU patients [26].This could partly explain that some patients had sufficient concentrations with eGFR above 48 ml/min/1.73m 2 .The predictive power may increase further if creatinine and cystatin-C combined estimation of eGFR is used, or if more accurate markers such as iohexol-or urine creatinine clearance are applied.
As with any real-world study on critically ill patients, a high heterogeneity among individuals should be expected, especially among individuals using different beta-lactams.Only piperacillin, cefotaxime, and meropenem were analysed in this study, and results cannot be directly transferred to other beta-lactam antibiotics.Additionally, this study focused on eGFR estimated from serum creatinine.Cut-offs or predictability may differ when other estimation methods are used, such as cystatin C-based eGFR, creatinine clearance using urine sampling, or iohexol clearance.The free fraction was based on calculations of measured total concentration and was used in the analysis to compare with MIC ECOFF .Protein binding capacity may vary between individuals, and analysing unbound concentrations could improve the accuracy in determining sufficient exposure.In a sensitivity analysis using a protein binding of 0 or 50%, there was a modest effect on the eGFR cut off for insufficient exposure (data not shown).
Although eGFR may have potential as a rapid screening tool to indicate patients at risk of inadequate betalactam exposures, it needs to be combined with TDM and model informed precision dosing to confirm target achievement and to individually adjust dosing regimens.

Conclusion
Serum creatinine-based eGFR has potential to simply and rapidly predict inadequate exposure to piperacillin, meropenem, and cefotaxime in ICU patients with the intermittent standard dosing as described by EUCAST.A creatinine-based eGFR of 48 mL/min/1.73m 2 or 58 mL/ min may be valid cut-offs to predict insufficient exposure in this patient population.

Table 1 .
Demographic and laboratory data.

Table 2 .
Predictive power of insufficient exposure based on eGFR with different cut-off values � .