Prospective monitoring of cefepime in intensive care unit adult patients

An empiric study in which the pharmacokinetics (PK) of imipenem were prospectively monitored in intensive care unit (ICU) children revealed wide inter-individual variations (up to four-fold at peak and >10-fold at through concentrations) that resulted in potentially too low dosages in 30% of cases [1]. Similar observations were also made with imipenem in adult patients [2, 3], suggesting that drug adjustment algorithms used at the bedside might not be always accurate in unstable ICU patients, and that drug monitoring should be used more often [1].

The present report describes a similar quality assessment study in which the PK of cefepime was monitored in ICU adult patients. As in the children's study alluded to above [1], PK results were not disclosed to the caretakers unless clinical problems were suspected to be associated with inappropriate drug dosages. This observation is timely in light of two meta-analyses that reported an increased mortality (risk ratio 1.26 (95% CI 1.08 to 1.49)) in patients treated for severe infection with cefepime, as compared to patients treated with other beta-lactams [4, 5]. Moreover, Bhat et al. [6] observed that bacteremia due to gram-negative pathogens with minimal inhibitory concentrations (MICs) of cefepime in the highest range of susceptibility (that is, 8 mg/l) or above [7] were associated with an increased mortality in patients treated with cefepime than in those treated with other antibacterials.

Alarmed by these reports, the Food and Drug Administration (FDA) completed a complementary meta-analysis of 88 comparative studies (including the 38 reported by Yahav et al) totalizing 9,467 cefepime-treated patients [8]. This analysis did not confirm a higher overall mortality related to cefepime. Nevertheless, in the absence of drug monitoring, the excess mortality or treatment failures reported in specific studies [4‐6] could be related to untoward overdosing or underdosing of cefepime in unstable patients.

Ideal dosing of cefepime should obey pharmacokinetic/pharmacodynamic (PK/PD) population kinetics that help adjust drug dosing to the most appropriate PK/PD profile against target organisms [9‐14]. This corresponds to a period of drug concentration above the MIC (T_>MIC) of >40% to 60% for beta-lactams in general [15‐20] and ≥50% for cefepime [19, 20]. However, whether these goals are reached in the empiric day-to-day clinical setting is uncertain, especially in unstable ICU patients. The present work examined these issues in 21 consecutive ICU adult patients treated with cefepime. Individual PKs were prospectively determined following a similar study design as for imipenem in children [1]. The results further strengthen the need for antibiotic monitoring in complicated clinical situations.

The present empirical study confirms the great inter-individual variability of cefepime PK in the clinical setting, as reported with cefepime and imipenem by others [1, 2, 27, 28]. Moreover, it underlines the difficulty of bedside prediction of cefepime PK, based on standard drug adjustment algorithms, including calculated CL_Cr. In the present series, this resulted in extreme cefepime concentrations in the plasma from rather low values (trough cefepime concentrations below 4 mg/l in ca 50% of the patients) (Figure 1) to unpredicted toxic values in two other patients with renal impairment.

A major parameter for cefepime drug adjustment is CL_Cr, which is often calculated by the classical Cockcroft-Gault equation [21]. However, calculated clearance may be subject to errors because it does not take into account features such as muscular mass and turnover, which may influence creatinine concentrations in the serum [29]. Therefore, biases in calculated CL_Cr could be one potential explanation for the inter-individual PK variability observed. Nevertheless, although the Cockcroft-Gault equation may suffer from inaccuracies, the calculated CL_Cr values correlated very well with cefepime clearance, as indicated in Figure 2. Additionally, we also tentatively calculated CL_Cr values using the MDRD (Modification of Diet in Renal Disease) method [30], but the results were quite concordant with the values presented herein (data not presented). Hence, some of the variations might be due to other factors.

For instance, some patients had increased CL_Cr as previously reported (>120 ml/minute, Figure 2) [31] and might have benefited from increasing drug dosages. Alternatively, additional more intricate parameters presented in Table 3 might also interfere. Among these, some relations were expected, such as the direct correlation between Cl_Cr and cefepime elimination, whereas others were less obvious, such as the direct correlation between the concentration of plasma albumin and K_β (Figure 2D). Depending on the circumstances, high plasma albumin may be associated either with dehydration, which could result in poor renal perfusion and decreased cefepime clearance, or with good cardiovascular performance and good cefepime clearance, which was likely to be the case herein.

Other parameters for initial dosing are weight and gender, which might call less attention by the caretakers in adult than in pediatric medicine. However extreme weights in our series varied by three times (Table 1) and were not likely to explain the up to 40-times difference in drug levels observed. Moreover, similar variations were observed in other PK studies [2, 3], and especially in children, where weight is a prime consideration in drug dosing decision [1]. Taken together, the extreme variations observed are likely to result from intricate interactions between multiple factors, which are by no way simple to integrate in the bedside decision process.

Most patients with a preserved renal function had stable individual PK profiles over time in spite of a wide range of CL_Cr values ranging from 160 to 53 ml/minute (Figure 2), and the fact that no drug adjustments were performed. In contrast, drug accumulation and toxicity was observed in two patients with renal impairment (CL_Cr < 50 ml/minute), in spite of drug adjustment. This is potentially important because caretakers did not attribute neuropsychological alterations, which may be multi-factorial in ICU conditions, to drug toxicity until the high concentrations of cefepime were disclosed to them and the symptoms resolved promptly after treatment arrest. Moreover, there is a lack of information in the literature regarding the threshold of cefepime plasma levels predicting neurotoxicity. Indeed, out of 35 patients with cefepime-induced neurological complications reported in 10 studies (excluding reviews and chronic dialysis patients) [27, 28, 32‐39], the concentrations of cefepime were determined in only one case (in the plasma and the CSF) and were quite high, that is, 284 mg/l and 18 mg/l, respectively [28]. Besides, only one recent study in neutropenic patients with mild renal failure indicated that trough plasma concentrations of cefepime above 22 mg/l were likely to be associated with encephalopathy [40]. The main constant over all the reported cases is the association of neurotoxicity with renal impairment. While renal impairment implies possible drug accumulation, it might also potentiate the effect of additional neurotoxic factors, including factors related to the patient, or maybe the C-3' substituent N-methylpyrrolidine metabolite of cefepime, which may accumulate in the case of renal failure [25, 41]. Thus, the threshold of toxicity might be patient-dependent. On the other hand, most studies examining the PK produced by 2 g of the drug administered intravenously or intramuscularly to healthy volunteers or patients without renal failure report trough cefepime concentrations in the plasma ≤10 mg/l in [9, 11, 42‐46], which was also the case herein. Therefore a safe assumption is that trough concentrations of >10 mg/l of cefepime should alert the clinician on the risk of neurotoxicity in susceptible patients, and concentrations of >20 mg/l should probably be avoided.

On the other extreme, too low dosages may result in treatment failures, at least as predicted by PK/PD studies [15‐19, 26]. Postulating that T_>MIC measured is pertinent to predict clinical outcome, then all of our patients had appropriate coverage of cefepime (T_>MIC ≥ 50%) as recently proposed [19, 20] for the presumed bacterial pathogens recovered herein (MIC ≤ 4 mg/l) (Table 4). On the other hand, if one postulates an MIC of 8 mg/l, which was associated with treatment failures in patients with bacteremia due to Gram negative pathogens [6], then close to 50% of the patients would have had an inappropriate coverage (T_>MIC > 50%). This is of particular concern when considering problematic pathogens such as those producing extended-spectrum beta-lactamases, or P. aeruginosa and Acinetobacter spp., which may have high cefepime MICs (≥8 mg/l) and pose major therapeutic challenges, and if one takes into account that up to 20% of the total drug is bound to serum proteins [47, 48]. Moreover, in addition to pure MIC concerns, a recent study identified P. aeruginosa infection, mechanical ventilation, and neutropenia as independent risk factors for cefepime treatment failure [49]. Higher cefepime doses were proposed to overcome some of these issues (for example, 2 g q 8 h) [9], but high doses may also increase the risk of neurological side effects. Hence, adjusting dosage on the basis of drug monitoring is reasonable in such cases.

Taken together, these results of drug monitoring independently validate the population kinetics of cefepime elaborated by others [9‐14]. Moreover, they show that empirical drug dosing following standard drug adjustment algorithms in the ICU is not accurate enough to prevent extreme PK deviations, which might be one or the possible explanations for the toxicity and treatment failure problems reported by Yahav et al. [4] and Bhat et al. [6]. Eventually, they indicate that 2 g of cefepime q 12 h is safe and effective for patients with CL_Cr ≥ 50 ml/minute and against pathogens with cefepime MICs ≤ 4 mg/l, but that drug monitoring should be considered in any conditions falling outside these limits.