Amikacin pharmacokinetic/pharmacodynamic in intensive care unit: a prospective database

Management of severe infections in intensive care unit (ICU) represent a major challenge for clinicians, and prompt initiation of effective antibiotic therapy is essential to improve patient’s survival [1, 2]. In patients with septic shock, guidelines recommend combination antimicrobial therapy with aminoglycosides, particularly when Gram-negative bacilli (GNBs) are suspected or documented [2].

Optimal anti-GNB activity of amikacin is achieved when peak serum concentration (C_max) reach eight to tenfold the minimal inhibitory concentration (MIC) [3]. With an amikacin MIC as high as 8 mg/L (beyond which the bacteria have intermediate susceptibility based on clinical breakpoints defined by EUCAST), the target C_max should therefore be 64–80 mg/L as recommended by French guidelines on aminoglycosides use [4, 5].

Previously published data with single daily dose of 15 to 25 mg/kg of amikacin pointed out that this target of C_max ≥ 64 mg/L was not reached for most patients [6‐8]. Thus, in order to optimize the C_max/MIC ratio, it is suggested to increase amikacin loading dose up to 30 mg/kg/day of total body weight (TBW) for severe patients [5].

However, in most clinical studies and in daily practice, the pharmacological efficacy of amikacin is only assessed by C_max measurements (objective of C_max ≥ 64 mg/L based on clinical breakpoints) and not by measured MICs.

To our knowledge, few studies have reported the antibacterial effect of amikacin based on C_max/MIC ratio using measured MIC for each patient profile.

We therefore undertook a retrospective analysis of our prospective database to assess pharmacodynamic target attainment (C_max/MIC ≥ 8)—considering amikacin measured MICs—in critically ill patients empirically treated for documented severe GNB infections.

In this prospective database performed in critically ill patients with documented GNB infections and receiving a 25 mg/kg single daily dose of amikacin, the overall probability of target attainment C_max/MIC ratio ≥ 8 was 93% according to our C_max and MIC distributions.

These data are not in accordance with previous studies that showed a risk of treatment failure, only based on amikacin C_max measurement (64 to 80 mg/L) or C_max/MIC ratio calculated with clinical breakpoints (8 to 10-fold a 8 mg/L MIC) [7, 8]. Most studies concluded that a 25 mg/kg loading dose of amikacin is insufficient in critically ill patients to achieve these pharmacodynamic targets and therefore endanger treatment efficacy.

However, in our study, although only 38.8% patients only achieved an amikacin C_max ≥ 64 mg/L, more than 90% had a C_max/MIC ratio over the pharmacodynamic target (≥ 8) when susceptible strains were involved.

Amikacin C_max measurements ranged from 12.2 to 165.7 mg/L with a median of 55.2 mg/L. This significant dispersion found in several studies conducted in ICUs is explained by the inter-individual variability of pharmacokinetic parameters in these patients [7, 8, 14‐16].

Nevertheless, in our study, the median amikacin C_max was much lower than in the literature (summarized in Additional file 1: Table S4). Indeed, three studies performed in critically ill patients and receiving a 25 mg/kg (TBW) loading dose reported a median amikacin C_max around 70 mg/L [7, 8, 14]: De Montmollin et al. showed that 58% of patients enrolled reached an amikacin C_max ≥ 64 mg/L [7] and Taccone et al. reported 70% of patients achieving this target [8].

Patients characteristics in our study and those reported above seem similar especially in terms of severity (estimated by the SOFA score), but there are significant differences in renal function (renal replacement therapy resort and serum creatinine were lower in our patients). These differences are all the more important as serum creatinine is a covariable of the pharmacokinetic of aminosides, which seems decisive for C_max. Indeed, to our knowledge, two studies identified renal function as a major co-variable of amikacin C_max in ICU patients and this may explain the lower amikacin C_max found in our study [14, 17]. We confirm these results because, in multivariate analysis, after adjustment with the SOFA score on D1 and the administered dose, serum creatinine on D1 was the only factor independently associated with a C_max ≥ 64 mg/L (OR = 1.01 [1.00–1.01], p = 0.004).

In any event, it is even more striking to reach about 90% of PK/PD target attainment with these lower amikacin C_max, which seems related to low amikacin MICs.

Amikacin median MICs of all GNBs identified in our study was 2 mg/L [0.19–16] and the median MIC of Pseudomonas aeruginosa was 3 mg/L [1–8]. These results were comparable to amikacin MICs distribution for GNB published by EUCAST and in the literature [18, 19].

Unsurprisingly, in the group of infectious episodes that did not reach the C_max/CMI ratio target, the median MIC of amikacin was higher (8 mg/L [4–16]). But excluding intermediate susceptibility GNBs, a C_max/CMI ratio ≥ 8 was achieved in 91% of treated infections.

Few other studies reported inconsistencies between calculated C_max or C_max/MIC (using clinical breakpoints) and measured C_max/MIC (using measured MICs). A study based on measured MICs (for some patients enrolled), reported a C_max/MIC ratio ≥ 10 in 93% of cases, with a median loading dose of 29.6 mg/kg, while target amikacin C_max was only achieved in 77% of cases [20]. Similarly, Pajot et al. reported that C_max/MIC ratio ≥ 10 was achieved in 87% patients with documented ventilator-associated pneumonia treated with only 20 mg/kg amikacin [18]. Finally, De Winter et al. in an emergency department, in which a cohort of patients with septic shock/severe sepsis received 25 mg/kg amikacin, showed that C_max/MIC ratio ≥ 8 was achieved in 76% of cases when using critical MICs (EUCAST), and in 95% of cases when using measured MICs [21].

These findings suggest that an exclusive focus on target C_max without MIC measurement is probably not suitable to evaluate the pharmacodynamic efficacy of amikacin therapy in clinical studies. Thus, added value of daily practice amikacin pharmacodynamic assessment is also uncertain, as the probability of pharmacodynamic failure to treat susceptible GNBs with a 25 mg/kg amikacin dose is less than 10% when evaluated by the measured C_max/MIC ratio in several different studies.

Nevertheless, when the MIC is above 4 mg/L, the risk of pharmacological failure is higher (around 40%) and clinicians should keep in mind the key role of the MIC, particularly when Pseudomonas aeruginosa is involved. Thus, dose adjustment could be relevant, based on local epidemiology (MICs distribution and/or P. aeruginosa infections incidence).

Increasing AMK loading dose could, however, lead to withhold the subsequent doses, if needed. Indeed, we reported a C_min ≥ 2.5 mg/L—when measured on D2—in 49/82 (60%) of treated infections and Roger et al. found similar results with a 30 mg/kg loading dose (C_min ≥ 2.5 mg/L in 49% of their patients) [20].

No correlation was found between C_max/MIC ratio and clinical outcome on D8 or D28 in our study. Moore et al. and Kashuba et al. demonstrated, 30 years ago, that the clinical benefit of aminoglycosides for the treatment of GNB infections was optimal if C_max reach 8 to tenfold the MIC [3, 22]. But recent studies in ICU patients have failed to demonstrate a potential impact of aminoglycoside pharmacodynamics on clinical outcome [7, 8, 20, 23]. This negative result should be put into perspective with the various variables that may possibly explain mortality in intensive care. The combination with another antibiotic, most often a β-lactam, could also explain the lack of demonstration.

Even so, the median C_max/MIC ratio in our study greatly exceeded the threshold proposed by Moore et al. [3] (23.1 versus 8), but the proportion of pharmacodynamic failure according to the following criteria (C_max/MIC < 8) was low, around 10%, which represents a small cohort for statistical comparisons.

Our study has several limitations. First, it is a retrospective study, based on a prospective database in 2 participating ICUs and in daily practice, BGN-documented infections are not systematically assessed with measured amikacin MIC. Factors that led to measure the MIC could introduce a selection bias, leading to measure the MIC only in the most difficult clinical situations. Nevertheless, the patients evaluated in our study are similar to those evaluated by De Montmollin et al. and Taccone et al. (in term of SAPS II score and mortality), allowing indirect comparison of our data with the literature [7, 8]. Second, we did not assess the incidence of renal toxicity following amikacin administration. However, 64% of infections were treated with single daily dose of amikacin, minimizing the risk of nephrotoxicity. In addition, data from the literature do not report an increase in renal toxicity following the initial dose [24]. Gàlvez et al. also showed that a dose of 30 mg/kg/day amikacin was not associated with a higher incidence of nephrotoxicity than 15 or 25 mg/kg/day regimen [6]. Finally, we studied the pharmacodynamic profile of amikacin during the first 24 h of administration. This is consistent with usual durations of aminoglycoside treatments (48 to 72 h) and with data reported by Kashuba et al. [22] that highlighted the impact of the first aminoglycoside dose on patient’s clinical outcome.

In critically ill patients treated with amikacin for susceptible Gram-negative bacilli infections, a 25 mg/kg (TBW) single daily dose actually achieved pharmacodynamic target in more than 90% of treated infections. Thus, in light of these results, the current trend of amikacin increase dose in the treatment of severe infections does not appear justified when local GNB ecology and amikacin MICs distribution are similar to ours. The relevance of systematic C_max measurement is also questionable or may require MIC measurement as well to ensure a reliable pharmacodynamic target assessment.