Introduction

Ventilator-associated pneumonia (VAP) is a frequent complication of mechanical ventilation (MV) which leads to significant morbidity and mortality [12]. Aminoglycosides are often used intravenously to treat VAP in addition to a β-lactam antibiotic [34]. Nevertheless systemic therapy is limited by the pulmonary diffusion of the antibiotic and the increasing level of bacterial resistances [56]. Increasing the antibiotic dose is a method used to overcome these limits, but this is associated with systemic toxicity, particularly with aminoglycosides. Administration of antibiotics as an aerosol directly delivered to the respiratory tract is an appealing alternative. This route of administration of aminoglycosides has proven effective in patients with cystic fibrosis, colonized with Pseudomonas aeruginosa [79]. In ventilator-dependent tracheotomized patients nebulized amikacin proved effective in treating tracheobronchial infection [10]. The effectiveness of nebulized amikacin in treating pneumonia, a lung parenchyma infection, has still to be investigated. The low bioavailability of nebulized amikacin may allow the administration of high doses with little systemic toxicity. In animal studies nebulization of 40–50 mg/kg amikacin allowed 3- to 30-fold higher lung parenchyma concentrations than conventional 15 mg/kg intravenous infusion [1114]. To our knowledge, such high doses of nebulized amikacin have never been reported in humans. Optimization of the nebulization technique is a key issue in aerosol therapy during MV because up to 90% of the nebulizer load can be lost in the ventilator circuit, thus reducing the inhalable mass (amount of drug delivered to the lungs) [1517]. Optimization of the nebulization technique in addition to the use of high amikacin doses may constitute an effective complement to standard therapy for VAP, but pharmacological and safety data in humans are lacking.

The aim of this study was to determine the pharmacokinetics and safety of high-dose nebulized amikacin in healthy subjects during noninvasive MV using an optimized nebulization technique. The nebulized doses were chosen to achieve serum levels close to those obtainedafter intravenous administration. The nebulization technique was validated in a preliminary bench study. Some of the results of this study have been previously reported as an abstract [18].

Methods

In vitro study

The vibrating mesh nebulizer Aeroneb Pro (Nektar Therapeutics, San Carlos, Calif., USA) was placed on the inspiratory limb just before the Y piece when used alone (manufacture's recommendations) and 20 cm before when used with the Idehaler spacer (La Diffusion Technique Française, St-Etienne, France), in vertical or horizontal position (Fig. 1). The ultrasonic nebulizer Atomisor Megahertz was placed 40 cm before the Y piece and tested without spacer [14]. Ventilatory pattern was standardized (Servo300 ventilator, Maquet, Rastatt, Germany: volume-controlled mode, square wave-flow pattern, 450 ml tidal volume, 16 breath/min, inspiratory/expiratory ratio 0.5, end expiratory pressure 5 cmH2O, fraction of inspired oxygen 21%). Nebulization was performed continuously. Nebulization yield was determined using the filter method with sodium fluoride as a chemical tracer, as previously described [19]. Briefly, 1 g sulfite-free amikacin (Laboratoire Merck Générique, Lyon, France) was dissolved in a 0.9% NaF solution (8 ml with the Aeroneb Pro and 12 ml with the Atomisor Megahertz nebulizer). An absolute filter (Gelman, Ann Arbor, Mich., USA) was placed between the extremity of the 8 mm endotracheal tube (connected to the Y piece) and the lung model. Filters were changed every 6 min and desorbed for 12 h in 100 ml 25% aqueous TISAB (Reagecon, Shannon, Ireland). The NaF extracted was then assayed by electrochemical analysis (Combination Fluoride Electrode, Orion, Beverly Hills, Calif., USA). Amikacin mass was calculated using the measured NaF mass. Three measurements were made for each setup. The mass median aerodynamic diameter (MMAD) was measured for each setup through laser diffraction (Mastersizer X, Malvern Instruments, Malvern, UK).

Fig. 1
figure 1

Diagram of the nebulization setup

In vivo study

Our institutional review board (Comité consultatif pour la protection de personnes se prêtant à la recherche biomédicale, Tours, France) approved the study protocol. Six healthy subjects (three men) aged 21–30 years and weighting 49–68 kg, without a history of respiratory, renal, or otological disease and with no contraindication to amikacin were included after written informed consent. The protocol consisted of four amikacin dosage sequences separated by a 7-day washout period. Sequence A consisted of a 1-h intravenous infusion of 15 mg/kg amikacin. The three following sequences consisted of nebulization of 40, 50, and 60 mg/kg amikacin (sequences B, C, and D, respectively; Fig. 2). Ten blood samples were collected by forearm venipuncture to measure amikacin serum concentrations over the 24-h period following each drug administration (Fig. 2). After centrifugation at 3,000 g for 10 min the supernatant was stored at –20 °C until completion of the study and analyzed using an enzyme multiplied immunomethod calibrated every 20 measurements (EMIT Amikacin assay, Cobas Mira+, Dade Behring, Paris, France). Amikacin was diluted in saline (1 g/8 ml); the nebulizer was refilled when sputtering until the total dose was nebulized. Throughout nebulization noninvasive MV was performed with a Servo300 ventilator in pressure-support ventilation (PSV) using a mouthpiece (Medi-test, Saclay, France). Subjects lay in a 45° semirecumbent position, PSV was adapted to achieve a respiratory rate of 12–20 per minute, positive end-expiratory pressure set at zero and fraction of inspired oxygen at 21%. The Aeroneb Pro nebulizer with the Idehaler vertical spacer was used (Fig. 1). Hemodynamic and respiratory monitoring was performed throughout amikacin administration and pulmonary functional testing was performed before and after sequence B. Creatinin clearance was calculated before each sequence and at 4 weeks' follow-up. Subjects underwent otological examination with otoacoustic emission test at enrollment and at 4 weeks' follow-up. Pharmacokinetics were analyzed using WinNonLin Professional 4.1 (Pharsight, Mountain View, Calif., USA).

Fig. 2
figure 2

Diagram of the study protocol. The study lasted 3 weeks, and each of the four sequences lasted 24 h. During each sequence pharmacokinetic measurements were performed as indicated for the first sequence. IV, Intravenous infusion; Neb, nebulization

Sequence A

After intravenous administration serum amikacin concentrations were analyzed using a compartmental and a noncompartmental approach [20]. A linear two-compartment model gave the best description of amikacin serum concentrations. Terminal elimination constant (λz) and half-life (t 1 / 2z) were derived from the compartment parameters. The area under the curve of serum amikacin concentrations vs. time after infusion (AUCsinf) was estimated by the trapezoidal method, extrapolated AUCs being estimated by Clastz, where Clast is the last concentration above the quantification limit. The area under the first moment of the amikacin concentration curve after infusion (AUMCsinf) was calculated as \( \sum \left \{ \left( \left[ \text{t}_i\text{C}_i + \text{t}_{i-1}\text{C}_{i-1} \right) / 2\right]\times \left[\text{t}_i-\text{t}_{i-1}\right] \right \} + \left(\text{t}_\text{last}\text{C}_\text{last} / \lambda_\text{z} \right) + \left(\text{C}_\text{last} / \lambda_\text{z}^2 \right) \) where C i and t i are measured concentrations and corresponding times, and tlast is the time of Clast [20]. Mean residence time after infusion (MRTinf), was calculated by: \( \text{MRT}_\text{inf} = \text{AUMC}_\text{s}\text{inf}/\break\text{AUC}_\text{s}\text{inf} \). The MRT of amikacin after bolus injection (MRTbolus) was obtained by: \( \text{MRT}_\text{bolus} = \text{MRT}_\text{inf}-\left(\text{infusion duration} / 2 \right) \) [20].

Sequences B, C, and D

Because of the slow absorption of amikacine during nebulization, leading to complex concentration vs. time curves, a compartmental pharmacokinetic analysis could not be used to describe nebulization sequences. Therefore amikacin serum concentrations measured during sequences B, C, and D were analyzed using a noncompartmental approach. AUCsneb and MRTneb were calculated with the same equations as for the infusion. During nebulization absorption was slow, hiding the elimination phase and thus yielding falsely low values of λz. As elimination kinetics are believed to be the same after infusion and nebulization, the values of λz obtained from sequence A were used for calculations. Mean absorption time(MAT: average time needed for an inhaled molecule of amikacin to reach the systemic circulation) was calculated for each subject by: \( \text{MAT} = \text{MRT}_\text{neb} - \text{MRT}_\text{bolus} - \left(\text{nebulization duration} / 2 \right) \).

Fraction of amikacin absorbed (F,%: fraction of dose placed in the nebulizer that reaches the systemic circulation) was calculated by: \( \text{F} = ( \text{AUC}_\text{s}\text{neb} / \text{D}_\text{neb} ) / (\text{AUC}_\text{s}\text{inf} / \\ \text{D}_\text{inf}) \), with Dneb = amikacin nebulizer load and Dinf =amikacin dose administered by intravenous infusion.

Noncompartmental analysis of amikacin absorption was completed by estimating the pulmonary absorption rate as a function of time through the Iga deconvolution method using the macro constants derived from the two-compartment intravenous model [21]. The instantaneous pulmonary absorption [a(t)] was weighed by the administered doses \( \left [\text{a}(\text{t}) = \text{absorption}\times \text{D}_\text{inf} / \text{D}_\text{neb} \right] \) and expressed as a cumulative fraction of the nebulizer load.

Values were reported as median (interquartile range). Variables were compared using the paired (in vivo study) or unpaired (in vitro study) Wilcoxon rang test (Statview, SAS Institute, Cary, N.C., USA). Probability of randomness below 5% was considered statistically significant.

Results

In vitro study

Placing the Idehaler vertical spacer underneath the Aeroneb Pro nebulizer significantly increased the inhalable mass (31%, 30–32, vs. 8% 8–9 of the nebulizer load) and inhalable output (681 mg/h, 630–742, vs. 235, 217–251; Table 1). Nebulization rate was stable throughout nebulization as no difference was noted in the amount of amikacin deposited on the various filters over time. MMAD was 4 μm (3.9–4.1) with the Atomisor Megahertz and 5 μm (4.9–5.1) with the AeronebPro nebulizer. The Idehaler vertical spacer did not influence the MMAD.

Table 1 Results of the bench study for a 1 g amikacin nebulizer load; values are medians (interquartile range)

In vivo study

All patients completed the four sequences of the study; no side effect occurred. The pulmonary functional tests remained in the normal range. No renal, otological, or respiratory impairment was detected. Serum concentrations after nebulization were less than or equal to those after intravenous infusion throughout the 24 h after administration (Fig. 3). For each individual subject the serum amikacin concentration after nebulization remained lower or equal (± 5 mg/l) to the concentrations after intravenous infusion at each time point. All 24-h concentrations were below the quantification limit of the assay (2.5 mg/l). Peak serum concentrations and AUCs values after intravenous infusion were significantly higher than after nebulization (Table 2). The pulmonary absorption rate was similar with the three nebulization doses, with MAT values above 2 h (n.s.; Table 2). No differences in the fraction of the nebulized dose absorbed were observed between the nebulization sequences: 13% (11–14), 12% (11–17) and 10% (8–15) with sequences B, C, and D, respectively (n.s.; Fig. 4, Table 2). Therefore the pulmonary absorption was not saturated in the healthy subjects, which allowed assuming linear absorption kinetics. This biodisponibility corresponded to 41% (35–45), 39% (35–55), and 32% (26–48) of the inhalable mass considering the 31% yield of the nebulization technique (bench study). Nebulization durations were: 53 min (52–55), 1 h 09 min (1, 02–1, 16), and 1 h 15 min (1, 13–1, 17) with sequence B, C, and D, respectively.

Fig. 3
figure 3

Mean amikacin serum concentrations during the 24 h period following the beginning of administration. Bars, Standard error of the mean

Table 2 Comparison between nebulization and intravenous infusion sequences; values are medians (interquartile range) (Peak, peak amikacin serum concentration; AUC s , area under the serum amikacin concentration vs. time curve; MAT, mean absorption time; MRT, mean residence time; F, fraction absorbed expressed as proportion of the nebulizer load; IV, 1-h intravenous infusion; Neb, nebulization using the vertical spacer)
Fig. 4
figure 4

Mean cumulative systemic absorption of amikacin as a function of time. Absorption is expressed as a fraction of the nebulizer load. The curves reached a plateau (corresponding to the fraction absorbed on the y axis) after about 6 h, which indicates that most of the systemically available amikacin had reached the serum. Bars, Standard error of the mean

Discussion

In this study the administration of up to 60 mg/kg amikacin by a nebulization technique yielding a 31% inhalable mass was associated with serum concentrations equal to or less than those obtained after intravenous infusion of 15 mg/kg amikacin, no side effects being noted in the six subjects studied. Otological and renal toxicity of aminoglycosides occurs through a saturable mechanism and is therefore not related to peak serum concentration. The AUCs is a better indicator of potential otological or renal toxicity [2223]. In our study after nebulization the AUCs remained significantly lower than after intravenous administration. Therefore nebulization of up to 60 mg/kg amikacin seems to carry a risk of renal or otological toxicity inferior to that of 15 mg/kg intravenous amikacin. Indeed, the fraction of amikacin absorbed after nebulization of 60 mg/kg was 10% of the nebulizer load; therefore the systemic circulation was exposed to a total mass of amikacin comparable to the intravenous infusion of 6 mg/kg. This limited biodisponibility may allow the use of high doses in order to increase pulmonary drug levels, but as discussed below this may not fully apply in patients with infected lungs.

The bench study revealed a higher inhalable mass when the Aeroneb Pro nebulizer was used with the Idehaler spacer. These results confirm studies reporting an increase in inhalable mass with the use of horizontal spacers [152425]. This increase in nebulization efficiency may be related to aerosol homogenization within the spacer. Our results suggest that the use of a vertically placed spacer further increases inhalable mass; higher aerosol droplet deposition during expiration in the horizontal spacer or inspiratory limb compared to a lower deposition in the vertical spacer may explain these results.

Our in vivo results are compatible with those reported by Goldstein et al. [14] in healthy ventilated piglets. The authors reported a peak serum concentration of 12 mg/l after nebulization of approx. 50 mg/kg amikacin and a ratio between AUCs after nebulization and after intravenous infusion of 58% [14]. Our measurements showed a peak serum concentration of 9.2 (7.6–10.2) mg/l and a ratio between AUCs after nebulization and after intravenous infusion of 46% with sequence C (50 mg/kg; Table 2). The lower serum concentrations in our study might be explained by higher oropharyngeal impaction of the aerosol in our study than endotracheal tube impaction, differences in ventilatory pattern, and interspecies differences in mucocilliary clearance and alveolocapillary diffusion. Caution is needed when comparing those study results because Goldstein et al. [14] reported a nebulization yield of 38% (measured by the mass balance technique), slightly higher than the yield of our nebulization setup (31%). We tested the Atomisor Megahertz nebulizer in the same conditions as reported by Goldstein et al. (40 cm before the Y piece on the inspiratory limb) to be able to compare our results with the animal data [1114]; using the filter impaction method, the nebulization yield measured was much lower (19%); Table 1.

The systemic absorption after antibiotic nebulization is believed to occur mainly at the alveolar level as it exhibits the thinnest interface between aerial and vascular compartments. Therefore systemic pharmacokinetics reflect deposition of the nebulized drug in the distal pulmonary parenchyma [2627]. Goldstein et al. reported very high lung parenchyma amikacin concentrations [14]. As we found similar amikacin serum pharmacokinetics, we hypothesize that lung parenchyma concentrations, albeit not measured in our healthy subjects, were high.

MAT is the average time spend by molecules in the lung before diffusion to the serum. With all sequences MAT was above 2 h (Table 2), and therefore it may be hypothesized that amikacin molecules remained for a sufficiently long time in the pulmonary parenchyma to allow potential antibacterial activity. Furthermore, in addition to the systemically absorbed amikacin, the molecules eliminated by mucocilliary clearance contribute to the antibacterial activity. Aminoglycosides exhibit a concentration-dependent bactericidal activity with a marked postantibiotic effect. Achieving high tissue concentrations once daily seems a pharmacokinetic strategy adapted to amikacin pharmacodynamics and has proven effective after intravenous administration [28]. In contrast, clinical studies performed with nebulized aminoglycosides in cystic fibrosis or ventilator-dependent patients evaluated multiple daily administrations of relatively low doses [710]; Palmer et al. [10] reported effective treatment of tracheobronchial infection with 400 mg amikacin nebulized every 8 h in ventilated tracheotomized patients. Whether once daily administration of high-dose nebulized amikacin is as effective or superior for treating lung infections warrants further clinical studies. Nevertheless in an animal model of pneumonia the reduction in lung bacterial burden was greater after 45 mg/kg nebulized amikacin once daily for 2 days than after standard intravenous treatment (15 mg/kg once daily) [13].

The results of our study should be applied only very cautiously, and after confirmatory clinical studies, to patients with VAP. Amikacin absorption may indeed be higher in patients with VAP than in our study, leading to potential risks of toxicity, for various reasons. First, pneumonia induces ruptures of the alveolocapillary barrier resulting in an increased diffusion from the lung to the serum. Using the above animal model of Escherichia coli pneumonia, Goldstein et al. [1314] observed higher serum peak concentrations than in their healthy lung animal model. Second, patients with VAP are ventilated through an endotracheal tube, which was not the case with the subjects in our study; oropharyngeal deposition of amikacin is very likely to have reduced the mass of amikacin delivered to the lungs. On the other hand, poor lung aeration may induce low alveolar concentrations and lower systemic absorption of amikacin in patients with VAP [11]. Thus, as demonstrated in animal models [29], the pharmacokinetics of nebulized amikacin may be less favorable in patients with infected lungs than in our study (lower antibiotic deposition in the poorly aerated infected lung segments with concomitant higher systemic absorption and toxicity). Studies in patients with infected lungs are therefore mandatory to draw conclusions regarding optimal dose and to assess short and long-term tolerance in patients. The subjects of our study were ventilated in PSV to maximize comfort, making it impossible to control inspiration/expiration ratio, tidal volume, or inspiratory flow which have been shown to influence aerosol delivery during MV [1516]. Respiratory rate (14 bpm, 12–15) and tidal volume (11 ml/kg, 9–12) measured were within the range usually recommended for optimal nebulization during MV [29], but lung protective ventilation of patients with VAP which may present acute respiratory distress syndrome requires lower tidal volumes. Our results may therefore not strictly apply to this specific patient population.

The nebulization setup that we investigated produced a high inhalable amikacin output, which allowed nebulizing high doses of amikacin in a short period of time (median nebulization duration of 1 h 15 min for 60 mg/kg placed in the nebulizer). This is of interest as nebulization during MV requires modifications of the ventilatory parameters and discontinuation of the inhaled gas humidification; thus the duration of nebulization should be kept as short as possible [1517]. In contrast, the Atomisor Megahertz nebulizer, albeit exhibiting a significantly higher inhalable mass than the Aeroneb Pro nebulizer, showed a similar inhalable output (Table 1). Inspiratory phase synchronized nebulization, which is an effective method to increase the inhalable mass, presents the same shortcoming because it does not increase the inhalable output and therefore results in significantly increased nebulization duration [30].

This study carried out in subjects with healthy lungs showed that, using an optimized nebulization technique with a vertical spacer (Idehaler) placed underneath the Aeroneb Pro nebulizer, doses of 60 mg/kg amikacin are probably associated with high lung parenchyma concentrations of amikacin during a sufficiently long period of time for bactericidal activity. No side effect was observed in the limited number of healthy subjects studied. This dose could be used as a basis for clinical studies investigating the potential benefits of once daily administration of high-dose nebulized amikacin in the treatment of VAP.