FormalPara Key Points

Standard doses of meropenem allowed achieving pharmacokinetic/pharmacodynamic target attainments

Higher doses of piperacillin/tazobactam would be needed to cover microorganisms with MIC > 8 mg/l

CrCL was predictive of fCss in both therapies

1 Introduction

Morbidity and mortality due to severe infections are prevalent in intensive care units (ICUs). Antibiotic-resistant infections are expanding [1], and this situation demands several measures, such as (1) to use old antibiotics, (2) to develop new therapies and (3) to optimize existing therapies [2]. Therapeutic interventions and external artifacts may contribute to pharmacokinetic/pharmacodynamic antimicrobial alterations and variability [3, 4]. Antibiotic therapies in the ICU remain challenging since standard dosage guidelines might be unsuitable and fail to achieve pharmacokinetic/pharmacodynamic target attainment [2].

To achieve clinical cure and bacteriologic eradication, it is traditionally believed that it is sufficient to keep plasma concentrations of β-lactams above the minimum inhibitory concentration (MIC) during 40–70% of the time in mild/moderate infections [1, 5, 6]. Nevertheless, longer exposure times (e.g., 100%fT ≥ MIC) might be required for critically ill patients [3, 7, 8]. Besides, clinical data suggest that β-lactam concentration should be between four and eight times above MIC to maximize bacterial killing and to avoid resistances [1, 8,9,10,11].

Previous pharmacokinetic studies showed that continuous infusion of β-lactam provided several advantages compared to intermittent administration: (1) higher percentage of antibiotic concentration values greater than the MIC (100% vs. 22% and 75% vs. 36% for meropenem [MER] and piperacillin/tazobactam [PIP/TAZ], respectively) [12], even with lower daily doses of PIP/TAZ than the standard regimen [13]; (2) higher concentrations of meropenem in both plasma and subcutaneous tissue [14]; (3) similar or higher clinical cure rates [8, 15]. These data support the use of continuous β-lactam infusion in ICU patients and offer an encouraging administration alternative [16].

The primary aim was to explore whether standard total daily dose of MER and PIP/TAZ [17, 18] administered by continuous infusion achieved optimal pharmacokinetic/pharmacodynamic targets in the actual hospital environment. We also wanted to identify risk factors associated with subtherapeutic exposure and failure to attain pharmacokinetic/pharmacodynamics targets.

2 Patients and Methods

2.1 Ethical Issues

The study was approved by the local Ethics Committee (SFB-ATB-2014-01) and conducted following the Declaration of Helsinki. Written informed consent was requested of the patient or the closest relative before inclusion.

2.2 Study Setting

This pharmacokinetic prospective and observational study was carried out over a 3-year period (June 2015–September 2018) in a 34-bed mixed ICU at Hospital Universitari de Bellvitge (Barcelona), a 700-bed teaching hospital in the southern metropolitan area of Barcelona.

Inclusion criteria were: (1) patient ≥ 18 years old with sepsis according to the Survival Sepsis Campaign Guidelines [19]; (2) under MER or (PIP/TAZ) therapy and (3) creatinine clearance (CrCL) ≥ 60 ml/min/1.73 m2. Exclusion criteria were: (1) pregnancy or (2) impaired renal function (CrCL < 60 ml/min/1.73 m2 or renal replacement therapy).

Patients received a loading dose followed by the total daily dose in continuous infusion, i.e., 4/0.5 g followed by 12/1.5 g q24h of PIP/TAZ (80 mg/ml in 0.9% saline, stability of 24 h at 25 °C, 1 infusion/day) and 1 g followed by 3 g q24h of MER (22 mg/ml in 0.9% saline, stability of 17 h at 25 °C, 2 infusions/day) [20]. Patients who had started antibiotic therapy with intermittent infusion in the previous 24 h did not receive the loading dose, because it was considered they had already achieved the steady state.

2.3 Bioanalytical Assay

Total plasma concentrations were determined through previously validated methods of ultra-performance liquid chromatography-tandem coupled to mass spectrometry (UHPLC-MS/MS) [21]. The mobile phase consisted of a mixture of solution A (0.1% formic acid in water) and solution B (0.1% formic acid in acetonitrile) with an initial composition of 5% solution B. The mobile phase flow rate was maintained at 0.4 ml/min using a gradient mode elution. For chromatography, an Acquity® UPLC® BEHTM C18 reverse-phase column (100 × 2.1 mm id; 1.7 µm) was used. A simple procedure for protein precipitation was used to prepare the samples. Piperacillin-d5 and meropenem-d6 were used as internal standard for PIP and MER, respectively.

Inter-day lower limits of quantification (LLOQ) were 0.50 mg/l for MER (signal-to-noise [S/N] ratio of 5.5) and 0.54 mg/l for PIP (S/N ratio of 5.6). The calibration curve ranged from 0.50 to 175 mg/l for MER (a quadratic regression curve with a weighting scheme of 1/X2) and from 0.54 to 175 mg/l for PIP (a linear regression curve with a weighting scheme of 1/X). For MER, inter-day coefficients of variation (CV) obtained were 10.1%, 7.4% and 4.9% at 3.22, 30.9 and 126 mg/l, respectively; the relative biases (δr) were 7.3%, 3.0% and 4.6% at the same values. For PIP, the CVs were 8.9%, 6.7% and 3.5% at 3.13, 31.5 and 124 mg/l; the δrs were 4.3%, 5.0% and 3.3%.

Blood samples were obtained 24–48 h after the beginning of β-lactam continuous infusion (steady-state condition). Approximately 3 ml of blood was collected in lithium-heparin tubes (Vacuette, Kremsmünster, Austria) and immediately refrigerated at 2–8 °C for a maximum of 30 min. Samples were then centrifuged at 2000g for 10 min at (4 ± 1) °C, aliquoted and stored at (− 75 ± 3) °C until analysis [21].

We calculated free antibiotic concentrations (fCss) considering protein bindings (2% and 30% for MER and piperacillin [PIP], respectively) [22]. As upper limit of the therapeutic window, we adopted PIP Css of 157 mg/l [9, 23] and MER Css of 45 mg/l [14, 24].

2.4 Study Cohort Data

All data were collected from the electronic medical information, and we calculated CrCL from serum creatinine concentrations according to the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula. We established three groups according to the following CrCL cut-offs: 60–89; 90–119 (female) or 90–129 (male); and ≥ 120/130 (female/male) ml/min/1.73 m2. Finally, we defined augmented renal clearance (ARC) as a CrCL ≥ 120/130 (female/male) ml/min/1.73 m2 [25]. We considered neurocritical care patients as those with traumatic brain injury or subarachnoid haemorrhage.

2.5 Exposure and Pharmacokinetic Parameters

The achieved exposure was given by the free antibiotic concentrations (fCss) and the area under the curve of free concentrations at steady state (fAUCss). We calculated unbound plasma clearance (CLu) and fAUCss according Eqs. 1 and 2, respectively [26]:

$${\text{CL}}_{{\text{u}}} [L/h] = {\text{daily dose}} \left[ {{\text{mg}}} \right]/24{\text{h}} \cdot f{\text{C}}_{{{\text{ss}}}}^{ - 1} \left[ {{\text{mg}}/{\text{L}}} \right]$$
(1)
$$f{\text{AUC}}_{{{\text{ss}}}} \left[ {{\text{mg}}\cdot {\text h}/{\text{L}}} \right] = {\text{daily dose }}\left[ {{\text{mg}}} \right] /{\text{CL}}_{{\text{u}}} [{\text{L}}/{\text{h}}]$$
(2)

2.6 Pharmacokinetic/Pharmacodynamic Endpoints

The pharmacokinetic/pharmacodynamic target was to achieve fCss exceeding the pathogen MIC during 100% of the dosing interval (100%fT). We defined three pharmacokinetic/pharmacodynamic targets: (1) fCss during 100%fT ≥ 1xMIC (fCss/MIC ≥ 1); (2) fCss during 100%fT ≥ 4xMIC (fCss/MIC≥4) and (iii) fCss during 100%fT ≥ 8xMIC (fCss/MIC≥8).

We determined actual MIC values at isolated pathogens by the Etest® method. Otherwise, we inferred the highest MIC in the susceptible range from the European Committee on Antimicrobial Susceptibility Testing (EUCAST) [27]: Pseudomonas spp., 16 mg/l for PIP and 2 mg/l for MER; Enterobacteriaceae, 8 mg/l for PIP and 2 mg/l for MER.

2.7 Evaluation Endpoints

We analysed pharmacokinetic/pharmacodynamic target attainment either as binary (expressed as number or percentage of attainments) or as continuous dependent variables (expressed as fCss/MIC ratio). Similarly, we studied the influence of clinical, physiological and mechanical factors on the antibiotic exposure, given by both fCss and fAUCss.

2.8 Statistical Analysis

We summarized descriptive statistics of continuous variables as median [interquartile range (IQR) or range] or mean [standard deviation (SD)] and the categorical variables as numbers and percentages. We presented fCss, fAUCss and CLu values and fCss/MIC ratios as geometric means with 95% confidence interval (CI). Results of pharmacokinetic/pharmacodynamic target attainment by MIC and breakpoint at the sample level were presented as numbers and percentages.

To examine predictors of final outcomes, considered as continuous variables, we performed univariate and multivariable linear regression analyses. Due to sample size considerations, we only performed statistical evaluation from pharmacokinetic/pharmacodynamic values estimated from surrogate MICs. In univariate analyses, we made comparisons of mean fCss/MIC ratios, fCss, fAUCss and CLu between groups created from different levels within each risk factor. We used a two-way analysis of variance with variables included as fixed factors and patient considered as a random factor nested within these variables.

We included independent factors that showed, in univariate analysis, a significant effect on the outcome, in a multivariable regression model to investigate independent predictors of fCss/MIC ratios. In multivariable regression analysis, we used a mixed model with the patient as a cluster and the logarithmic transformation of the dependent variable.

In all multivariable analyses, we used stepwise procedures based on forward inclusion/backward elimination methods. We performed statistical analyses using R version 3.5.1. and set statistical significance to p < 0.05 in all the cases.

3 Results

3.1 Population and Samples

During the study period, 118 patients were included, and 149 samples were analysed [96 (64.4%) and 53 (35.6%) for PIP and MER, respectively]. Only 24 (20%) patients had more than one plasma sample. Baseline patient characteristics, clinical, microbiological data and concentrations achieved are shown in Table 1. Most patients (64.4%) had CrCL ≥ 90 ml/min/1.73 m2 and 60.2% had body mass index (BMI) ≥ 25 kg/m2. Median (IQR) MIC value in isolated pathogen was 0.05 (0.02–0.12) mg/l for MER and 3 (1–4) mg/l for PIP. Median fCss was 15.8 (IQR: 7.35–32.3) mg/l and 26.8 (IQR: 17.5–42.6) mg/l for MER and PIP, respectively. Eight meropenem Css values were > 45 mg/l but we did not find any adverse events.

Table 1 Demographics, clinical baseline and microbiological characteristics of the patients included in the study
Full size table

3.2 Pharmacokinetic/Pharmacodynamic Target Attainment

Table 2 shows the results for pharmacokinetic/pharmacodynamic target attainment considering two scenarios (actual or surrogate MIC values). Usually, we observed higher percentages of achievement with actual MIC values. Achievement of MER pharmacokinetic/pharmacodynamic targets using surrogate MIC was similar (fCss/MIC ≥ 1) (98.11% vs. 100%), 25% lower (71.70% vs. 96.15%) (fCss/MIC ≥ 4) and 49% lower (47.17% vs. 96.15%) (fCss/MIC ≥ 8) than those observed considering actual MICs. In the PIP cohort, we observed a similar trend with similar percentages (fCss/MIC ≥ 1) (95.83% vs. 95.56%) or reductions of 46% (44.79 % vs. 91.11%) (fCss/MIC ≥ 4) and 56% (6.25% vs. 62.22%) (fCss/MIC ≥ 8) when surrogate MIC 8 mg/l was evaluated and of 12% (83.33% vs. 95.56%) (fCss/MIC ≥ 1), 84% (6.25% vs. 91.11%) (fCss/MIC ≥ 4) and 61% (1.04% vs. 62.22%) (fCss/MIC ≥ 8) when surrogate MIC 16 mg/l was considered.

Table 2 Percentages of pharmacokinetic/pharmacodynamic target attainment by minimum inhibitory concentrations (MICs)
Full size table

3.3 Influence of Clinical Factors on Exposure, Pharmacokinetic Parameters and Pharmacokinetic/Pharmacodynamic Target Values

Results of the effect of the tested clinical factors on exposure (fCss, fCss/MIC, fAUCss) and pharmacokinetic parameters (CLU) are shown in Table 3. Univariate comparisons evidenced that, for MER, a trend to lower fCss values occurred in patients with CrCL ≥ 90 ml/min/1.73 m2 with respect to those with CrCL of 60–89 ml/min/1.73 m2 (fCss: 12.5 vs. 22.8 mg/l, p = 0.072). Similarly, overweight patients (BMI ≥ 25 kg/m2) presented almost half the exposure of those with BMI ≤ 24.9 kg/m2 (fCss: 11.6 vs. 20.0 mg/l, p = 0.118). Although these were the most influential covariates, no statistical significance was reached in any case. Patients under mechanical ventilation (fCss 12.6 vs. 17.8 mg/l, p = 0.34) and post-surgical drainage (fCss: 10 vs. 17.9 mg/l, p = 0.219) also tended to lower fCss values than the others, but statistical significance was not achieved. The trend shown in fCss values of patients treated with vasoactive drugs with respect to those that did not receive this treatment (fCss: 16.5 vs. 14.2 mg/l, p = 0.058) could be attributed to the high variability observed. One patient of the group that received the treatment showed much higher exposure (fCss = 96 mg/l) than the others, this contributing to these results.

Table 3 Effect of covariates on meropenem and piperacillin exposure and pharmacokinetic parameters
Full size table

In patients under PIP treatment, CrCL was the most influential covariate (p = 0.005) followed by neurocritical status (fCss 22.2 vs. 30.3 mg/l, p = 0.008) and mechanical ventilation (fCss 23 vs. 32.6 mg/L, p = 0.024). Patients with CrCL values ≥ 90 ml/min/1.73 m2 had lower exposures than the others (fCss 23 vs. 36.6 mg/l, p = 0.001) (Table 3, Fig. 1 and Supplementary file: Fig. 1S). Figure 2 displays the statistically significant correlation between fCss/MIC values and CrCL when surrogate MIC values (8 and 16 mg/l) were considered. An inversely proportional linear relationship is observed so that it is showed that 15.2% of the variation of the fCss/MIC value is due to the progressive increase of the CrCL.

Fig. 1
figure 1

Boxplot of piperacillin fCss/MIC ratio distributions, according to surrogate MIC values, sorted by each category within each variable. Footnote: The fCss/MIC distributions based on surrogate MIC values of 8 and 16 mg/l were the same, so only the boxplots for MIC 8 mg/l are represented. The bottom and top extremes of the box represent the first (Q1) and the third quartile (Q3) range of the data, respectively (Q3–Q1: interquartile range). The dark horizontal line in the box is the median and dots are the observed values. The bottom and top whiskers represent the Q1 – 1.5 times the IQR value and Q3 + 1.5 times the IQR values, respectively. fCss free antibiotic concentrations, MIC minimum inhibitory concentration. fCss/MIC ratio of fCss to surrogate MIC values (8 and 16 mg/l for piperacillin), BMI body mass index, CrCL creatinine clearance value (ml/min/1.73m2), calculated with the CKD-EPI equation

Full size image
Fig. 2
figure 2

Correlation between CrCL and fCss/MIC ratios for piperacillin concentrations. Footnote: MIC values of 8 mg/l (A) and 16 mg/l (B). Creatinine clearance (CrCL) value (ml/min/1.73 m2) was calculated with the CKD-EPI equation. fCss free antibiotic concentrations, MIC minimum inhibitory concentration, fCss/MIC ratio of fCss to surrogate MIC values (8 mg/l or 16 mg/l), R2 coefficient of determination

Full size image

The multivariable analysis showed the statistically significant effect of CrCL on MER exposure after adjusting by BMI. This finding was probably due to the reduction of variability associated with fCss values after inclusion of BMI in the multivariable analysis. However, its effect was not statistically significant, suggesting that BMI acts as a confounder due to its relationship with both the CrCL and the fCss. Thus, the final multivariable model included CrCL [β = − 0.01 (95% CI − 0.02 to − 0.0; p = 0.043)] as a significant factor that influenced MER exposure.

Regarding PIP, we could not find any statistically significant effect of mechanical ventilation on the fCss/MIC ratio when this covariate was entered on the multivariable model. Of note, in univariate analysis, it was the less influential covariate among those mentioned above. In PIP, the negative predictors of target achievement were CrCL [β = − 0.01 (95% CI − 0.02 to − 0.01, p < 0.001)] and neurocritical status [β=− 0.36 (95% CI − 0.61 to − 0.11, p = 0.005)] (see Supplementary file: Table 1S).

4 Discussion

Considering the actual MIC of isolated microorganisms (55, 47% of patients), our results suggest that standard doses of MER would reach 100%fT ≥ 1xCMI, 100%fT ≥ 4xCMI and 100%fT ≥ 8xCMI in > 96% of occasions. In the case of PIP/TAZ, also 100%fT ≥ 1xCMI and 100%fT ≥ 4xCMI were reached in more than 90% of occasions but higher PIP/TAZ doses would be needed to achieve the most ambitious target (100%fT ≥ 8xCMI).

Isolated pathogens had median MIC values much lower than the EUCAST cut-off [0.05 mg/l vs. 2 mg/l (MER) and 3 mg/l vs. 8 or 16 mg/l (PIP)]. Therefore, we found lower percentages of achievement when we considered the more conservative surrogate MIC values. We considered our results from the MER cohort (71.7% of 100%fT≥4xCMI target attainment) similar to those from Dhaese et al. [28] (75% of 100%fT ≥ 4xCMI target attainment). Compared to our study, they included patients with slightly higher renal function and more estimated CrCL (Cockroft-Gault equation) variability than ours [mean (SD): 117.8 (68.2) vs. 99.2 (22.5)]. However, fCss values from our MER cohort presented higher variability (range 0.7–96.5 mg/l) than their values (range 2–57.7 mg/l). Fifteen of 53 fCss (28.3%) were < 8 mg/l [vs. 10 of 48 (20.8%) in Dhaese’s study], and 7 fCss values were over the upper limit (vs. 2 in Dhaese’s study). In the PIP cohort, our 100%fT ≥ 1xCMI target attainment was in accordance with previous findings (83.3%) [29]. However, our study showed lower 100%fT ≥ 4xMIC target attainment (6.25%) than Dhaese et al. (37.1%) [28] and Richter et al. (55.6%) [30]. A combination of two situations could have contributed to these differences: (1) the administration of different daily doses of PIP (16 g in Dhaese et al. [28] and 12 g in Richter et al. [30]) with respect to our study (3 g); (2) the inclusion of patients with CrCL < 60 ml/min in Dhaese’s and Richter’s studies.

As reported earlier [29, 31,32,33,34,35], our findings suggest that high renal function is an important risk factor for non-target attainment. As expected for renal-excreted drugs, in the present study, drug concentrations were strongly associated with CrCL. For both antibiotics, CrCL was the most influential covariate in the multivariable analysis (MER, p = 0.043; PIP, p < 0.001) but the strongest relationship between concentration and CrCL was found for PIP. On the other hand, Carlier et al. [31] observed a higher impact of CrCL on pharmacokinetic/pharmacodynamic target attainment for MER than ours, i.e., 2.8% less probability to reach 100%fT≥CMI when CrCL increased [β − 0.028; 95% CI for Exp (β): 0.955–0.990; p < 0.002]. This estimation was obtained with multivariate logistic analysis. In our case, in multivariate linear analysis, we observed that for every unit increase in CrCL, fCss/MIC decreased by 1%. They [31] observed a larger range of variation in estimated CrCL, and this could justify the differences from our results. Thus, according to these results, drug monitoring of β-lactams and dose adjustment based on renal function could increase the pharmacokinetic/pharmacodynamic target attainment.

Curiously, in the multivariable analysis, the statistically significant influence of CrCL on MER fCss values (p = 0.043) could only be detected after inclusion in the BMI in the model. This suggested that BMI ≥ 25 kg/m2 could act as a confounding factor because of its association with both fCss and creatinine clearance.

In line with this, patients with BMI ≥ 25 kg/m2 showed lower MER fCss/MIC values (5.8 vs. 10, p = 0.118) compared to patients with BMI < 25 kg/m2. This could be explained by the effects of overweight on either drug clearance or volume of distribution (Vd), as previously reported [36, 37]. No data about Vd values were available in our study, but obese patients showed higher MER clearance than non-obese (10.8 vs. 6.2 l/h, p = 0.118) resulting in lower exposures. Increased kidney size and renal flow could be some of the physiological changes causing higher clearance [36]. Similar results were found by Hites et al. [38], with higher CL values in obese patients and 35% vs. 0% of non-target attainment in obese vs. non-obese patients (p = 0.02). Other authors [39] described a significant relationship between BMI and Vd without affecting pharmacokinetic/pharmacodynamic target attainment. Although post-surgical drainage has been postulated to produce antimicrobial loss because of augmented clearance [40, 41] and a false Vd increase [4, 42, 43], we did not find any significant influence of this covariate on the pharmacokinetic/pharmacodynamic target attainment.

To the best of our knowledge, this is the first study attempting to address the effect of MV on PIP pharmacokinetic/pharmacodynamic target attainment. We identified neurocritical status as an influential covariate for PIP fCss/MIC (p = 0.008). These results were in agreement with recent reports where brain-damaged patients failed to achieve pharmacokinetic/pharmacodynamic targets as they were at particular ARC risk [44,45,46,47]. Moreover, in univariant analysis, the influence of MV was statistically significant (p = 0.024), even though this effect was not retained in the final multivariable model. The lower target attainment in patients with MV could be associated to the effect of positive end-expiratory pressure (PEEP) on Vd [48,49,50,51]. Nevertheless, no data from our study could prove this hypothesis. Although vasoactive drugs could probably increase renal blood flow and thereby drug clearance, this effect could not be shown in the present study.

In the present study, the effect of several factors that had never been previously investigated [28, 30] such as diagnosis, MV, vasoactive drug use, neurocritical status and post-surgical drainage was analysed. We confirmed the lack of independent influence of body weight on target attainment as we analysed the effect of renal function estimated using CKDEPI formula that is independent of body weight. Rather than novelty one of the features of our study is that it was carried out by means of statistical analysis methods compliant with longitudinal data. Moreover, multivariable statistical analyses were performed on the basis of continuous variables rather than categorized or discrete data, this leading to a more powerful and robust analysis. The identification of the predictive capability of the investigated factors on target attainment is crucial for dose individualization during therapeutic drug monitoring in clinical practice.

Some limitations of our study are, first, the small sample size that led to the lack of statistical significance in the MER CL values between ARC and normal renal function. Second, direct urinary creatinine measurement, the most adequate method to assess CrCL, was not routinely available in our centre. Third, pathogens were only grown in 47% of patients and, to obtain robust results, we needed MIC assumptions. Moreover, the use of susceptibility breakpoints could inflate the frequency of sub-threshold levels when drug concentrations were instead adequate because the ‘true’ MIC was substantially lower. Fourth, tazobactam Css was not measured, since its analytical determination was not routinely available in our centre. Finally, this study includes a purely kinetic analysis; thus, we do not presume to draw any conclusions for the clinical outcome.

5 Conclusions

Standard total daily dose of MER (3g q24h) and PIP/TAZ (12/1.5g q24h) administered as a continuous infusion is usually adequate. However, in patients with CrCL ≥ 90 ml/min/1.73m2 (MER and PIP/TAZ), neurocritical status and infections caused by microorganisms with MIC > 8 mg/ (PIP/TAZ) caution is warranted to avoid underdosing. Therapeutic drug monitoring and dose adjustment are highly recommended in these specific situations.