Optimizing the Initial Amikacin Dosage in Adults

Abstract

We report on the pharmacokinetics (PK) and pharmacodynamics (PD) of high-dose (>15 mg/kg of body weight per day) amikacin. A mean (standard deviation [SD]) maximum drug concentration in the serum (C_max) and 24-h area under the concentration-time curve (AUC₂₄) of 101 (49.4) mg/liter and 600 (387) mg · h/liter, respectively, were observed (n = 73) with 28.0 (8.47) mg/kg/day doses. An initial amikacin dose of 2,500 mg in adults weighing 40 kg to 200 kg with therapeutic drug monitoring to adjust the maintenance dose will optimize its PK and PD.

TEXT

Amikacin is an important component of an initial empirical antimicrobial treatment strategy against major Gram-negative pathogens associated with serious infections. The regulatory approved dose of this aminoglycoside is 15 mg/kg of body weight per day as 2 to 3 divided doses (1). Aminoglycoside dosing has shifted from divided daily dosing to single daily dosing in order to optimize its concentration-dependent pharmacokinetic-pharmacodynamic (PK-PD) profile (2, 3). This dosing paradigm seeks to achieve a maximum concentration of drug in the serum (C_max)/MIC of 8 to 10 and an area under the concentration-time curve (AUC)/MIC of 75 (3–5). As a consequence, currently accepted standard doses of aminoglycosides, such as tobramycin, are now administered 5 to 10 mg/kg once daily in patients with normal kidney function, that is roughly 2-fold higher than the regulatory approved dose of 3 to 5 mg/kg/day in divided doses (5).

The typical 4-fold higher MIC₉₀ (4 to 8 mg/liter) of amikacin against Pseudomonas aeruginosa compared to tobramycin (1 to 2 mg/liter) coupled with the dose proportionality of aminoglycosides led our institution to adopt an initial empirical amikacin dose of 24 mg/kg in noncritically ill patients and 30 to 40 mg/kg in critically ill patients. Two amikacin concentrations are measured, at 1 to 2 h and at 8 to 10 h after the end of infusion of this first dose, to permit dose individualization (6). Recent clinical studies corroborate this approach by suggesting that an initial amikacin dose ≥25 mg/kg is likely needed as empirical therapy of certain Gram-negative infections (7–10). Herein, we report on the amikacin exposure profile observed in our patients with this higher-than-regulatory-approved amikacin treatment strategy (1). We also provide a clear rationale for the consideration of an alternate empirical fixed amikacin dosing strategy in line with the current clinical adult total body weight (TBW) distribution.

After institution review board approval, patients ≥21 years of age who received at least one dose of amikacin from 1 January 2012 to 30 June 2014, who had two serial amikacin measurable concentrations, and who had an estimated creatinine clearance (CL_CR) >30 ml/min based on the Cockcroft-Gault equation were included (11). Information pertaining to patient demographics, laboratory values, amikacin dose, administration and infusion times, amikacin concentration, and sample collection times were collected. Amikacin concentrations were measured by an automated turbidometry immunoassay using the UniCel DxC 880i Synchron Access clinical system (Beckman Coulter, Brea, CA). Serum creatinine concentrations were quantified based on an isotopic dilution mass spectrometry referenced method that was not modified during the study period. The concentration-time data were fit by population PK analyses (Adapt 5; BMSR, Los Angeles, CA) using a 1-compartment model and a 2-compartment model (12). The relationship of amikacin PK parameters and exposure to body size were evaluated by visual inspection of scatter plots followed by regression. The final model was selected based on the Akaike information criterion (AIC) (13). The final model was used to generate empirical Bayesian estimates of individual AUC₂₄, C_max, and minimum drug concentration in the serum (C_min) values. The C_max was the estimated concentration at the end of infusion, and the C_min was the estimated concentration 24 h from the start of infusion. Classification and regression tree analysis (CART) was used to identify amikacin doses associated with achievement of a C_max/MIC of ≥10 and an AUC₂₄/MIC of ≥75 based on an MIC₉₀ of 8 mg/liter. These results were used to define doses (1-h infusion) for Monte Carlo simulation (MCS) and permit calculation of the cumulative fraction of response (CFR) for a C_max/MIC of ≥10 and an AUC₂₄/MIC of ≥75 targets and the EUCAST MIC distribution of amikacin against P. aeruginosa (14, 15). Some institutions define amikacin doses based on a calculated dosing weight (DW) algorithm of (i) TBW if less than ideal body weight (IBW); (ii) IBW if TBW is <1.2-fold higher than IBW; or (iii) adjusted body weight (ABW) if TBW is ≥1.2-fold higher than IBW (2). As a result, we also tested the CFR associated with amikacin doses that may be calculated by DW. Descriptive statistics, CART analyses, CFR computation, and figures were generated using Stata/SE, version 13.1.

A total of 218 adult cases received amikacin, and 145 cases were excluded due to an absence of two measurable amikacin concentrations (n = 120), doses <15 mg/kg of TBW/day (n = 8), amikacin administration time not charted (n = 5), and creatinine clearance <30 ml/min (n = 10). Two cases were excluded as outliers due to a very low serum creatinine value and incorrect body size data to appropriately estimate CL_CR. A 2-compartment model was selected as the final model (AIC decreased from 1,780 to 1,767) with a mean (relative standard error) total body clearance (CL), central compartment volume of distribution (V_c), intercompartmental clearance (CL_d), and peripheral compartment volume of distribution (V_p) of 3.22 liters/h (12.3%), 17.7 liters (22.0%), 3.52 liters/h (36.1%), and 22.5 liters (41.7%), respectively, with excellent model prediction of individual concentrations (R² = 0.90).

Patient demographics and Bayesian estimates of amikacin exposures (n = 73) are reported in Table 1 based on receipt of a mean (SD) dose of 2,262 (869) mg or 28.0 (8.47) mg/kg (TBW) or 31.8 (9.60) mg/kg (DW). A majority of patients in the study were white (92.9%) and male (65.9%), and 45.3% were critically ill. About 37.0% of patients were obese based on a body mass index (BMI) of ≥30 kg/m². Most of the patients had TBW of <80 kg (46.6%) or 80 to <120 kg (45.2%) and had DW of <60 kg (32.8%) or 60 to <90 kg (56.2%). The relationship of amikacin V_c and V_p with TBW or DW was weak (R² < 0.1) and not significant (P > 0.1). Similarly, the amikacin CL was correlated to age and serum creatinine but not to TBW or DW. Consistent with this observation, the amikacin C_max and AUC₂₄ were better correlated to amikacin dose in milligrams (R² < 0.32 and 0.39, respectively) than to the dose in milligrams per kilograms (TBW or DW) (R² < 0.13). The relative hazard ratio (RHR) by CART analyses to achieve or not achieve a C_max/MIC of ≥10 was 0.50 and 1.46 for doses above and below 2,200 mg, respectively. Similarly, the RHR to achieve or not achieve AUC₂₄/MIC of ≥75 was 0.38 and 1.89 for doses above and below 2,500 mg, respectively. Only 63.0% and 36.9% of our patients achieved these C_max/MIC and AUC₂₄/MIC targets, respectively.

TABLE 1.

Summary of population demographics, kidney function, and amikacin exposure

Parameter	Mean	SD	Minimum	Maximum
Age	52.5	19.1	24.0	89.0
Height (cm)	168.3	15.4	94.0	203
Total body weight (kg)	83.4	30.3	38.1	193
Dosing weight (kg)	69.3	17.3	30.0	124
Body mass index (kg/m2)	29.6	10.9	15.4	66.1
Creatinine clearance (ml/min)	104.2	45.8	32.5	231
Cmax (mg/liter)	100.9	49.4	23.4	314
Cmin (mg/liter)	10.2	12.5	0.027	58.5
AUC24 (mg · h/liter)	600.2	386.6	124.7	2,194

The mean (SD) MCS predicted C_max and AUC₂₄ values and CFR are provided in Table 2 and centered on TBW-based and DW-based doses and a fixed dosing approach. The CFR exceeded 90% for a C_max/MIC of ≥10 when the amikacin dose was 30 mg/kg of TBW, 40 mg/kg of DW, or 2,500 mg. However, the CFR was consistently higher with a fixed-dose approach compared to weight-based dosing for an AUC₂₄/MIC ≥75. Figure 1 illustrates the observed (dose normalized for comparison) and predicted AUC₂₄ with a single dose of 30 mg/kg or 2,500 mg and shows that the AUC₂₄ increased across the weight categories but was lower in the 40- to 80-kg group (45% of the sample) when dosed on TBW. The same trend was observed when comparing 40 mg/kg of DW to the 2,500 mg dosage (data not illustrated).

TABLE 2.

C_max, AUC₂₄, and CFR predicted based on model-based simulation of amikacin by weight-based and fixed doses and on the EUCAST P. aeruginosa MIC distribution^a

Initial amikacin dose (mg/kg)	Mean (SD) Cmax (mg/liter)	CFR for Cmax/MIC ≥10 (%)	Mean (SD) AUC24 (mg · h/liter)	CFR for AUC24/MIC ≥75 (%)
Total body weight (mg/kg)
15	65.0 (35.9)	72.5	370 (217)	29.5
20	86.6 (47.9)	81.8	494 (289)	41.4
30	130 (71.9)	90.7	741 (433)	58.7
40	173 (95.8)	94.6	987 (577)	70.2
Dosing weight (mg/kg)
15	52.9 (25.0)	66.3	301 (155)	22.1
20	70.6 (33.3)	77.6	401 (206)	33.3
30	106 (49.9)	88.2	602 (309)	51.5
40	141 (66.6)	93.1	802 (413)	64.0
Fixed dose (mg)
1,500	78.5 (37.5)	81.4	435 (194)	69.0
2,000	105 (50.0)	88.1	580 (259)	80.0
2,500	131 (62.5)	92.5	725 (324)	85.7
3,000	157 (75.0)	94.7	870 (389)	89.3

^aSimulation of 5,000 subjects using the population model and covariance matrix with parameter truncation to match observed data distributions. The median (minimum, maximum) simulated values were as follows: CL, 3.28 liters/h (0.423, 9.36 liter/h); V_c, 18.2 liters (1.71, 35.1 liters); CL_d, 3.47 liters/h (0.0783, 8.085 liters/h); V_p, 22.0 liters (1.95, 43.5 liters); total body weight, 83.6 kg (40.0, 189 kg), and a matching distribution of 40 to <80 kg (45.0%) or 80 to <120 kg (43.6%); dosing weight, 69.7 kg (30, 124 kg), and a matching distribution of 30 to <60 kg (28.9%) or 60 to <90 kg (59.2%).

FIG 1.

Box-and-whisker plot of the area under the curve over 24 h (AUC_0–24) based on observed (n = 73) values normalized to total body weight and fixed doses (A) and predicted (n = 5,000) values based on simulated weight-based and fixed doses (B).

Our observations are consistent with the pharmacometric analyses conducted by Rughoo et al. in a sample of 872 patients treated with amikacin (16). Their analysis elegantly demonstrates the relationship of amikacin volume of distribution and weight to be weak and to not serve as a good scalar for dosing (10). We clearly show that the regulatory approved dose of 15 mg/kg/day is expected to be suboptimal based on PK-PD targets. Recent studies have suggested that doses of ≥25 mg/kg are necessary to optimize the PK-PD profile (7–10). However, these studies have not simulated the effect of a fixed dose in their population models, even though weight is not shown to be a proportionate scalar of key PK parameters (7, 8). We illustrate the predicted effects of both approaches that match our observations (Fig. 1). If rapid target attainment and short-term use of amikacin are the empirical goals to optimize effect and safety, respectively, then the use of a single initial 2,500 mg dose of amikacin followed by TDM to individualize the dose is suggested (17). This suggestion, if validated by other groups, not only is simpler across the clinical weight distribution but also should increase the likelihood of effect against serious Gram-negative pathogens.

Our study has several limitations that would be expected of a retrospective design. We included a heterogeneous clinical population who received weight-based doses of amikacin and not a fixed dose, as is proposed by these results. Our model is based on a small and heterogeneous study population but closely matches the PK system parameters reported by recent studies (7, 8). Based on our results, an initial dose of at least 2,500 mg of amikacin increases the expected CFR, and this mathematical expectation can be validated by other research groups. The increasing use of this agent against serious multidrug-resistant Gram-negative infections and the availability of a commercial assay make this suggested shift in the dosing paradigm testable by other research groups prior to broad implementation.