Amikacin nebulization for the adjunctive therapy of gram-negative pneumonia in mechanically ventilated patients: a systematic review and meta-analysis of randomized controlled trials

Treatment of ventilated patients with gram-negative pneumonia (GNP) is often unsuccessful. We aimed to assess the efficacy and safety of nebulized amikacin (NA) as adjunctive therapy to systemic antibiotics in this patient population. PubMed, Embase, China national knowledge infrastructure, Wanfang, and the Cochrane database were searched for randomized controlled trials (RCTs) investigating the effect of NA as adjunctive therapy in ventilated adult patients with GNP. Heterogeneity was explored using subgroup analysis and sensitivity analysis. The Grading of recommendations assessment, development, and evaluation approach was used to assess the certainty of the evidence. Thirteen RCTs with 1733 adults were included. The pooled results showed NA had better microbiologic eradication (RR = 1.51, 95% CI 1.35 to 1.69, P < 0.0001) and improved clinical response (RR = 1.23; 95% CI 1.13 to 1.34; P < 0.0001) when compared with control. Meanwhile, overall mortality, pneumonia associated mortality, duration of mechanical ventilation, length of stay in ICU and change of clinical pneumonia infection scores were similar between NA and control groups. Additionally, NA did not add significant nephrotoxicity while could cause more bronchospasm. The use of NA adjunctive to systemic antibiotics therapy showed better benefits in ventilated patients with GNP. More well-designed RCTs are still needed to confirm our results.


Study selection.
Studies were considered eligible if they met the following criteria: (1) design: RCTs; (2) population: adult (≥ 18 years old) critically ill patients with MV (tracheal intubation or tracheostomy) and diagnosed of GNP (caused by susceptible or resistant pathogens); (3) intervention: patients were randomized to either NA group or control group (aerosolized placebo or no drug), both of which were given alongside intravenous antibiotics during the treatment period (decided by the attending physician based on available culture results or clinical guidelines provided); and (4) predefined outcomes: clinical response, mortality, microbiologic eradication, clinical pulmonary infection score (CPIS), duration of MV and length of stay in ICU. We excluded studies as following: (1) the main focus was children or pregnant women, (2) with any different therapy other than NA between two groups, (3) use of NA as monotherapy, (4) studies focused on in vitro or cystic fibrosis or just pharmacokinetic/pharmacodynamic, (5) available only in abstract form or meeting reports, and 6) studies without reporting predefined treatment outcomes.

Data extraction and outcomes. Data extraction was undertaken by H-BH and JPQindependently for
included studies on study design, patient inclusion criteria, NA and control group regimens, microbiological and clinical cure criteria, as well as predefined outcomes. Authors were contacted where data were unclear or unavailable. The primary outcome was the clinical response (defined as a complete or partial resolution of clinical signs and symptoms of infection, according to the criteria by each study author). Secondary outcomes included overall mortality (defined as ICU or hospital or 28-day mortality, the longest follow-up reported was preferred), pneumonia associated mortality, microbiologic eradication (defined as no growth of the causative pathogen from any samples taken [e.g., sputum, throat swab or bronchoalveolar lavage fluid] after treatment, regardless of the clinical outcome), change of CPIS from baseline after treatment (∆CPIS), length of stay in ICU, duration of MV and adverse events of bronchospasm and nephrotoxicity. Discrepancies were identified and resolved through discussion. Quality assessment. The two investigators also independently assessed the quality of RCTs using the risk of bias tool recommended by the Cochrane Handbook for Systematic Reviews of Interventions 21 . We also used Jadad score to assess the quality of included trials 22 . The quality of evidence resulting from the present meta-analysis was evaluated using the Grading of recommendations assessment, development, and evaluation

Statistical analysis.
The results from all relevant studies were combined to estimate the pooled risk ratio (RR) and associated 95% confidence intervals (CI) for dichotomous outcomes. As to the continuous outcomes, mean differences (MD) and 95% CI were estimated as effective results. Some studies reported median as the measure of treatment effect, with accompanying interquartile range (IQR). Before data analysis, we estimated mean from median and standard deviations (SD) from IQR using the methods described in previous studies 24 . Heterogeneity was tested by using the I 2 statistic. An I 2 < 50% was considered to indicate insignificant heterogeneity and a fixed-effect model was used, whereas a random-effect model was used in cases of significant heterogeneity (I 2 > 50%) using the Mantel-Haenszel method 25 . Testing the robustness of our outcomes and exploring the potential influence factors, we performed sensitivity analyses by omitting one study in each turn to investigate the influence of a single study on the overall pooled estimate of each predefined outcome. Also, subgroup analyses were performed concerning the primary outcome by pooling studies with the following: (1) types of nebulizers (Jet or ultrasonic or vibrating nebulizer); (2) dose of NA (≥ 800 mg/day or < 800 mg/ day); (3) proportion of patients with drug-resistant bacteria (including multidrug-resistant (MDR), extensively drug-resistant (XDR) or pan drug-resistant (PDR) bacteria) (100% or < 100%); (4) study design (blinded or un-blinded), and estimated models (fixed-effect or random effect models). All analyses were performed using Review Manager, Version 5.3. The quality assessment of the evidence was evaluated by GRADE profiler software version 3.6 (GRADE Working Group, 2004Group, -2007.
Studies characteristics and quality assessment. The main characteristics of included RCTs and predefined outcomes are shown in Table 1   www.nature.com/scientificreports/ technique, including nebulizer position, ventilator settings, humidifier, respiratory mode, and sedation during the nebulization period (Appendix 5). The Cochrane risk of bias score for each study is summarized in Appendix 6, Fig. S1A and S1b. Four studies 11,14,19,26 were assessed to be at low risk of bias overall and nine studies 20,[27][28][29][30][31][32][33][34] were at high risk of bias overall. The median Jadad score of the included studies was 2.6 (range from 1 to 5, see Appendix 7). Using GRADE methodology, we evaluated the evidence for pooled data for clinical response rate, overall mortality, pneumonia associated mortality, microbiologic eradication, ∆CPIS, duration of MV, length of stay in ICU, nephrotoxicity, and bronchospasm to be moderate, moderate, moderate, low, low, very low, low, respectively (Table 2). Assessment of publication bias using visually inspecting funnel plots and modified Galbraith tests showed no potential publication bias among the included RCTs (Appendix 8, Fig. S2a and S2b) (Appendix 8, Fig. S2). Table 1. Characteristics of the studies included in current systemic review and meta-analysis. APACHE II = acute physiology and chronic health evaluation II, AUC 0-24 h = area under the concentration-time curve from 0 to 24 h, CAP = community acquired pneumonia, CPIS = clinical pulmonary infection score, Cmax = maximum concentration, DB = double blind, GNB = gram-negative pneumonia, HAP = hospitalacquired pneumonia, h = hours, HCAP = healthcare-associated pneumonia, ICU = intensive care unit, IVAB = intravenous antibiotics, MC = multi-centers, Mix-ICU = medical-surgical intensive care unit, NA = nebulized amikacin, NR = not reported, PR = prospective randomized, SD = standard deviation, SC = single-center, VAP = ventilator-associated pneumonia. a ITT = intention-to-treat analysis, b defined as multidrug-resistant or extensively drug-resistant or pandrug-resistant gram-negative pneumonia.  (Fig. 2). In the sensitivity analysis, exclusion of any single trial did not significantly alter the overall combined RR (P value ranging from 1.22 to 1.37, with I 2 from 31 to 53%), whereas most subgroup analyses based on types of nebulizers, NA dose, sample size, study quality or study design confirmed similar improved clinical response among groups. However, the use of NA did   Table 3).

Discussion
The present meta-analysis assessed the role of NA as adjunctive therapy in ventilated patients with GNP. We found NA has a better microbiologic eradication and improve the clinical response. Meanwhile, NA did not affect mortality, ∆CPIS, and duration of MV or ICU stay. Additionally, NA did not add significant nephrotoxicity, while it could cause more bronchospasm.
To date, several recent meta-analyses and guidelines have suggested favorable clinical response of aerosolized antibiotics in ventilated pneumonia [3][4][5] . However, pooled results of different study designs (RCTs and observational studies), various antibiotics (aminoglycosides, colistin, and vancomycin), and different therapy strategies (adjunctive and substitution) might contribute to the significant heterogeneity among the included studies. Meanwhile, observational studies have the risk of overrated pooled estimates. To address these limitations, we focused specifically on NA used as adjunctive therapy in ventilated GNP, expanded the sample size by including recent published RCTs, and conducted robust data analyses and quality evaluation. We found NA is effective as such a therapeutic strategy for GNP. Therefore, our findings support and expand the suggestions in previous meta-analyses and guidelines.
To facilitate comparison with the previous meta-analyses 4,5 , we chose clinical response as the primary outcome. Indeed, from a research and clinical standpoint, the clinical response may be a more reliable parameter compared with other important clinical outcomes (e.g., CPIS, microbiologic eradication or mortality, duration of MV, and ICU stay). For instance, the CPIS was originally designed for VAP diagnosis, rather than assessing the response to treatment 35 , whereas mortality is an outcome not only related to GNP, but it is also influenced by many other prognostic factors (e.g., underlying diseases, the severity of illness or immunity of the host). Furthermore, clinical response was the most reported outcome and might provide more evidence to aid in the clinical decision. Table 3. Subgroup analysis on primary outcome of clinical response. NA = nebulized amikacin; CI = confidence interval; GNP = gram-negative pneumonia. www.nature.com/scientificreports/ Our results showed NA exhibited better clinical response. However, we should interpret this finding with caution. First, we found moderate heterogeneity among the pooled trials in this outcome. This heterogeneity could be caused by different pathogenic bacteria and the definition of clinical response between the pooled trials. Subgroup-analysis of studies with large sample size and double blinding also could not confirm this benefit of NA. Second, we could not demonstrate a significant reduction in mortality, ICU LOS, and ventilated duration. Additionally, although NA resulted in better microbiologic eradication, the eradication data varied widely among the pooled studies (ranging from 29 to 71%) 11,14,18,24 , which means these data can be susceptible to some clinical factors, such as microbiological detection technique, the proportion of drug-resistant GNP, systemic antibiotics therapy, or airway secretions or antibiotics contained in bronchoalveolar lavage fluid. Of note, the positive detection of microbial culture may be affected by colonization with bacteria, and the correlation has been demonstrated to be poor between the positive cultures alone and histologically confirmed pneumonia 36 . Thus, microbiologic eradication based on microbial culture does not necessarily mean the eradication of deep parenchymal pneumonia.
Several included studies with high quality, though reporting the negative results, provided information concerning the specific treatments in NA. This might help to explain the opposite results among the included studies. On the one hand, the severity and extension of pulmonary infection might affect the lung deposition of NA. In ventilated animal models with pneumonia, lung tissue concentrations of NA were markedly lower in pulmonary segments with confluent pneumonia and lung abscess compared to that in the early stages of lung infection. However, most patients of included RCTs received NA only after their time-consuming VAP/GNP diagnosis procedures. This, to some extent, delays the administration of NA in the early stages of GNP. Furthermore, most of these patients also received a prolonged course of MV and/or intravenous amikacin before receiving NA. This might contribute to an increase in airway biofilms and bacterial resistance, thus making lung infection treatment more difficult and ineffective.
On the other hand, several critical factors, such as aerosol particle size, type of nebulizer, physical characteristics of the carrying gas, and respiratory settings during the implementation of NA can also influence lung deposition of NA. By and large, to increase the efficiency of aerosol delivery, ultrasonic or vibrating mesh nebulizers producing low flow turbulence, volume-control mode with the constant inspiratory flow and appropriate end-inspiratory pause (representing about 20% of the duty cycle) are preferred; whereas heating and humidification that increase the diameter of the aerosol particles (> 5 μm), decelerating flows, spontaneous modes or ventilator-patient asynchrony during NA period should be avoided. In one RCT focusing on nebulized antibiotics in VAP, the authors chose vibrating mesh nebulizers and filled out the well-designed checklist before NA to standardize and optimize the nebulization procedure. However, the total extrapulmonary (nebulizer chamber, the inspiratory limb of the respiratory circuit, and the expiratory filter) depositions of amikacin were as high as 40%. Therefore, it can be conceivable that in clinical practice, as shown in the included RCTs in the current study (Appendix 4), the efficiency of actual aerosol delivery may be lower. However, this may also mean that there is still ample space for improvement in nebulized techniques in the future.
This study has several limitations. First, most of included studies 14,18,24 had a sample size of fewer than 100 patients, which might be subject to overestimation of effect size. Second, definitions and timing assessment of microbiologic eradication, the dose of amikacin used, as well as disease severity varied among the included RCTs. This might lead to observed heterogeneity, thus impairing the robustness of our findings. Third, the duration of MV before NA, time to start NA, and pathogens varied across included RCTs. The original plan of subgroup analysis to further explore trials based on the above diversities was hampered by insufficient data. Finally, the results of some subgroup analyses should be interpreted with caution due to insufficient studies, i.e., type of nebulizers or study design.

Conclusion
In summary, based on the current evidence, the use of NA adjunctive to systemic antibiotics therapy showed better benefits in ventilated patients with GNP. However, the overall quality of included studies is poor and more well-designed RCTs are still needed to confirmed our results.