Usefulness of β-lactam and macrolide combination therapy for treating community-acquired pneumonia patients hospitalized in the intensive care unit: Propensity score analysis of a prospective cohort study.

INTRODUCTION
Whether β-lactam and macrolide combination therapy reduces mortality in severe community-acquired pneumonia (SCAP) patients hospitalized in the intensive care unit (ICU) is controversial. The aim of the present study was to evaluate the usefulness of β-lactam and macrolide combination therapy for SCAP patients hospitalized in the ICU.


METHODS
A prospective, observational, cohort study of hospitalized pneumonia patients was performed. Hospitalized SCAP patients admitted to the ICU within 24 h between October 2010 and October 2017 were included for analysis. The primary outcome was 30-day mortality, and secondary outcomes were 14-day mortality and ICU mortality. Inverse probability of treatment weighting (IPTW) analysis as a propensity score analysis was used to reduce biases, including six covariates: age, sex, C-reactive protein, albumin, Pneumonia Severity Index score, and APACHE II score.


RESULTS
A total of 78 patients were included, with 48 patients in the non-macrolide-containing β-lactam therapy group and 30 patients in the macrolide combination therapy group. β-lactam and macrolide combination therapy significantly decreased 30-day mortality (16.7% vs. 43.8%; P = 0.015) and 14-day mortality (6.7% vs. 31.3%; P = 0.020), but not ICU mortality (10% vs 27.1%, P = 0.08) compared with non-macrolide-containing β-lactam therapy. After adjusting by IPTW, macrolide combination therapy also decreased 30-day mortality (odds ratio, 0.29; 95%CI, 0.09-0.96; P = 0.04) and 14-day mortality (odds ratio, 0.19; 95%CI, 0.04-0.92; P = 0.04), but not ICU mortality (odds ratio, 0.34; 95%CI, 0.08-1.36; P = 0.13).


CONCLUSIONS
Combination therapy with β-lactam and macrolides significantly improved the prognosis of SCAP patients hospitalized in the ICU compared with a non-macrolide-containing β-lactam regimen.


Background
Community-acquired pneumonia (CAP) is a leading cause of morbidity and mortality among infectious diseases worldwide [1].
Severe CAP (SCAP) patients hospitalized in the intensive care unit (ICU) are reported to have a high mortality of up to 25-50% [2][3][4]. To improve the prognosis of SCAP patients, comprehensive treatment including antimicrobial therapy, mechanical ventilatory support, vasopressor drug support, and nutrition is essential, and early and appropriate antibiotic therapy is particularly important [5,6].
Some CAP guidelines recommend that β-lactam combination therapy including β-lactam and macrolides or β-lactam and quinolones should be administered to SCAP patients hospitalized in the ICU [1,7]. Many previous studies have reported that βlactam and macrolide combination therapy (macrolide combination therapy) was useful for reducing mortality in SCAP patients compared with β-lactam monotherapy or β-lactam and quinolones [8][9][10][11][12][13]. However, these were all prospective cohort studies including prospective and retrospective parts, and there have been no randomized, controlled trials (RCTs) that de nitively showed the clinical usefulness of macrolide combination therapy so far. In addition, Adrie et al showed that dual therapy including βlactam plus macrolide or uoroquinolone did not signi cantly reduce 60-day mortality compared with β-lactam monotherapy for CAP patients requiring ICU admission [14]. Therefore, whether macrolide combination therapy truly improves the prognosis of SCAP patients compared with other regimens is still controversial. Furthermore, all previous studies that showed the usefulness of macrolide combination therapy were conducted in the United States and in European countries, and no study was conducted in other areas or countries, including Asian countries.
The aim of the present study was to investigate whether macrolide combination therapy signi cantly reduced the mortality of SCAP patients hospitalized in the ICU compared with a non-macrolide-containing antibiotic regimen. Propensity score analysis was used to reduce some biases.

Study design and setting
This observational, prospective cohort study enrolled consecutive patients with pneumonia, including CAP and healthcareassociated pneumonia (HCAP), hospitalized at the Kurashiki Central Hospital, which is a 1,166-bed tertiary hospital, between October 2010 and October 2017. The patients were diagnosed as having pneumonia based on the Infectious Diseases Society of America/American Thoracic Society guidelines [1]. Brie y, patients were diagnosed as having pneumonia if they had at least one of the following clinical symptoms (fever, cough, sputum, dyspnea, and pleuritic chest pain), plus at least one nding of coarse crackles on auscultation or elevated in ammatory biomarkers including C-reactive protein or white blood cell count, in addition to new in ltrates on chest radiography. Patients with CAP hospitalized in the ICU within 24 hours of admission were included. The exclusion criteria were age < 15 years, HCAP [15], hospital-acquired pneumonia, and patients treated without a β-lactam. Patients diagnosed with Legionella pneumonia on admission were also excluded because they were all treated with uoroquinolones or macrolides and not treated with a β-lactam. This study was performed as a clinical study of pneumonia (UMIN000004353) and was approved by the institutional review board of Kurashiki Central Hospital (approval number 3398). Based on the Ethical Guidelines for Medical and Health Research Involving Human Subjects of the Ministry of Health, Labour and Welfare, the research subjects were noti ed or the public was made aware of information concerning the research on a website. All patients gave their informed consent to participate in this study by being given opportunities to refuse to participate.
In all patients, pneumonia severity on admission was assessed using the CURB-65 score [confusion, urea > 7 mmol/L, respiratory rate ≥ 30 breaths per minute, low blood pressure (systolic < 90 mmHg or diastolic ≤ 60 mmHg), and age ≥ 65 years] [16] and the Pneumonia Severity Index (PSI) [17]. The APACHE II score, which was previously reported to be able to predict prognosis well in ICU patients, was also evaluated [18]. The antibiotics administered to all patients were at the discretion of the attending physician.
Criteria for admission to the intensive care unit and patient management Pneumonia patients were usually treated in the ICU on admission if at least one of the following criteria was met: (1) need for mechanical ventilatory support including non-invasive positive pressure ventilation and invasive positive pressure ventilation; (2) need for vasopressor drug therapy; or (3) unstable condition that would lead to mechanical ventilatory support or vasopressor drug support early, although the patients did not need such support on admission.
In our hospital, a semi-closed ICU system was adopted. In brief, a member of the Department of Respiratory Medicine was the attending physician of hospitalized CAP patients in the ICU and selected the antibiotic regimen. On the other hand, intensivists provided comprehensive patient care, including management of mechanical ventilation, circulatory dynamics, and nutrition.

Antibiotic regimens
The non-macrolide-containing therapy group (non-macrolide group) was de ned as including β-lactam monotherapy or β-lactam + non-macrolide antibiotic (quinolones, tetracyclines, glycopeptides) combination therapy. The macrolide combination group was de ned as including β-lactam + macrolides or β-lactam + macrolides + other antibiotics. De-escalation and the duration of antimicrobial agent treatment were at the discretion of the attending physicians.

Microbiologic examinations
Sputum and blood for cultures and blood for measuring serum antibodies were collected on admission to detect causative pathogens. A causative microorganism was identi ed according to a previous report [19].
In summary, a causative pathogen was identi ed if any one of the following criteria was satis ed: (1) positive sputum culture of more than 1 + on a qualitative test or ≥ 10 5 on a quantitative test, in the context of a signi cant Gram stain; (2) positive blood culture, excluding bacterial contamination; (3) positive pleural uid culture; (4) positive urinary antigen test for Streptococcus pneumoniae or Legionella pneumophila serogroup 1; (5) seroconversion or a 4-fold increase in the antibodies for Mycoplasma pneumoniae or Chlamydophila pneumoniae; and (6) ≥ 1:320 on single particle agglutination antibody test for M. pneumoniae (FUJIREBIO; Tokyo, Japan) or ≥ 2.0 the cut-off index on a C. pneumoniae IgM antibody test (Hitazyme assay; Hitachi Chemical, Tokyo, Japan).

Outcomes
The primary outcome of this study was 30-day mortality, and secondary outcomes were 14-day mortality and ICU mortality.

Statistical analysis
Nominal variables are expressed as numbers and percentages, whereas continuous variables are expressed as medians and interquartile range. Nominal variables were analyzed using Fisher's exact test, and continuous variables were analyzed using the non-parametric Mann-Whitney U-test.
For comparisons of 30-day mortality, 14-day mortality, and ICU mortality between the non-macrolide group and the macrolide combination group, propensity score (PS) methods were used to reduce bias and the effects of patients' confounding factors on the treatment outcomes. The PS was de ned as the probability that a patient would be assigned to a particular therapy, based on the patients' baseline covariates. Inverse probability of treatment weighting (IPTW) was selected for PS analysis, because it was reported to result in a lower mean squared error when estimating the effect of treatment [20]. The PS was estimated by multivariate logistic regression analysis involving six covariates, including age, sex, C-reactive protein, albumin, PSI score, and APACHE II score. These covariates were selected based on previous reports analyzing prognostic factors for CAP patients in the ICU [21][22][23][24]. All statistical analyses were two-tailed, and a P value of < 0.05 was considered signi cant. Analyses were conducted by R (version 3.0.3, Vienna, Austria).

Patients' characteristics
In this prospective cohort, a total of 1,544 CAP patients were hospitalized, and 78 patients were nally included in the analysis ( Fig. 1). Table 1 shows the baseline characteristics of the non-macrolide group and the macrolide combination group. The two groups did not differ signi cantly in age, sex, comorbidities, vital signs, or laboratory examinations on admission. Of the indices of pneumonia severity, only the PSI score was signi cantly different between the groups, and the CURB-65 and APACHE II score showed no signi cant differences. The rates of mechanical ventilatory support and vasopressor drug support were not signi cantly different between the two groups. The 30-day mortality rate and the 14-day mortality rate were signi cantly lower in the macrolide combination group than in the non-macrolide group (16.7% vs. 43.8%, P = 0.015; 6.7% vs. 31.3%, P = 0.020), but not the ICU mortality rate (10% vs. 27.1%, P = 0.08) (Fig. 2). There were 0 and 5 deaths within 30 days of admission in the macrolide combination and non-macrolide groups, respectively. These patients were transferred to a medical ward from the ICU because a do not attempt resuscitation order was issued in the ICU based on the determination that no further recovery was likely. Data are presented as medians (interquartile range) or n (%).
b Diagnosed according to the GOLD de nition [37]. Patients who were already diagnosed and treated as having COPD at other hospitals and had emphysema on chest computed tomography were also included.
Initial antibiotic therapy regimen in the non-macrolide-containing therapy and the macrolide combination therapy groups Table 2 shows the initial antimicrobial agents in the non-macrolide and macrolide combination groups. In the non-macrolide group, the most frequently used antibiotic therapy was sulbactam/ampicillin monotherapy (35.4%), followed by ceftriaxone monotherapy (12.5%). On the other hand, in the macrolide combination group, the most frequently used antibiotic therapy was sulbactam/ampicillin + azithromycin (26.7%), followed by ampicillin + azithromycin (20.0%). Etiology of pneumonia The distribution of the causative pathogens in the present study is shown in Table 3. In the non-macrolide group, the most common causative microorganism was S. pneumoniae (14.6%), followed by Haemophilus in uenzae (10.4%), Moraxella catarrhalis (10.4%), and methicillin-susceptible Staphylococcus aureus (MSSA) (10.4%). However, in the macrolide combination group, the most common pathogen was S. pneumoniae (43.3%), followed by MSSA (13.3%). a There were 7 and 3 patients with multiple etiologies in the non-macrolide therapy group and the macrolide combination group, respectively; therefore, the sum of the infection rates is over 100%.
Outcome after adjustment by propensity score analysis Table 4 shows the standardized mean differences of the six covariates before and after adjustment by IPTW analysis. The standardized mean differences were less than 0.1 for all covariates. After adjusting by IPTW analysis, the macrolide combination group had signi cantly decreased 30-day mortality (odds ratio 0.29, 95% con dence interval 0.09-0.96, P = 0.04) and 14-day mortality (odds ratio 0.19, 95% con dence interval 0.04-0.92, P = 0.04), but ICU mortality was not signi cantly decreased (odds ratio 0.34, 95% con dence interval 0.08-1.36, P = 0.13) ( Table 5).

Discussion
The present study showed that macrolide combination therapy signi cantly decreased 30-day and 14-day mortality, but not ICU mortality compared with β-lactam monotherapy or β-lactam and non-macrolide-containing combination therapy. After adjusting by IPTW analysis, macrolide combination therapy also signi cantly improved 30-day and 14-day mortality, though ICU mortality was not signi cantly decreased.
Many previous studies reported that macrolide combination therapy improved prognosis compared with non-macrolidecontaining regimens [8][9][10][11][12][13]. However, all of the studies were observational cohort studies including prospective and retrospective parts, and there were no RCTs. In addition, there have been few studies that used propensity scores to reduce various biases [11].
Thus, whether macrolide combination therapy truly reduces mortality in SCAP patients hospitalized in the ICU is controversial, although a previous meta-analysis study stated that "this meta-analysis supported the use of macrolides as a rst-line combination treatment in critically ill patients with severe CAP" [25]. In the present study, IPTW analysis was used to reduce as much as possible some previously known biases related to prognostic factors. Therefore, the present study's results con rm those of the previous studies.
Regarding ICU mortality, the present study did not show the usefulness of macrolide combination therapy. The reason was thought to be due to the fact that 5 patients died in the medical ward within 30 days of admission in the non-macrolide combination group. These patients were transferred to a medical ward from the ICU due to a do not attempt resuscitation order.
In the guidelines for CAP in the United States and Europe [1,7], combination antibiotic therapy including β-lactam and macrolides or β-lactam and quinolones was recommended for SCAP patients hospitalized in the ICU. Based on these guidelines and the results of the present study, macrolide combination therapy appears to be more suitable for SCAP patients than β-lactam monotherapy, although Japanese CAP guidelines state that β-lactam monotherapy is a valid antibiotic regimen for SCAP patients hospitalized in the ICU [26].
There is another question related to which antibiotic regimen including β-lactam and macrolides or quinolones is better for treatment of SCAP patients. Some studies showed that guideline-concordant therapy was important to improve the prognosis of SCAP, but not speci c treatment regimens [10,27,28]. In the present study, mortality was not compared between β-lactam and macrolide combination therapy and β-lactam and quinolone combination therapy because only a few patients were treated using β-lactam and quinolone combination therapy (n = 12). Therefore, whether macrolide combination therapy signi cantly reduces mortality compared with β-lactam and quinolone combination therapy remains unknown.
However, there are some reasons that macrolide combination therapy may be preferred to quinolone combination therapy. First, Martin-Loeches et al reported that macrolide combination therapy signi cantly reduced 30-day ICU mortality compared to βlactam and quinolone combination therapy in patients treated using mechanical ventilatory support (26.1% vs. 46.3%, P = 0.04) in a prospective cohort study. They also showed that macrolide combination therapy signi cantly improved outcomes in severe sepsis and septic shock patients [12]. On the other hand, Mortensen et al showed that initial empiric antibiotic therapy with βlactam and uoroquinolone combination therapy was associated with signi cantly increased 30-day mortality for SCAP patients compared with other guideline-concordant therapy using a propensity score analysis in a retrospective cohort study [29]. Second, macrolides have some immunomodulatory effects [30,31]. Anderson et al reported that macrolides signi cantly decreased pneumolysin production by S. pneumoniae, but uoroquinolones and ceftriaxone did not [32]. In addition, a meta-analysis of 28 RCTs showed that empiric atypical coverage did not signi cantly improve the prognosis in hospitalized CAP patients [33], and macrolides and β-lactam antibiotics have no synergistic effects [34,35]. Therefore, we think that immunomodulatory effects of macrolides have an important role in improving the prognosis, as stated above. Indeed, high levels of both pro-and antiin ammatory cytokines have been reported to cause a poor prognosis for SCAP patients, even in patients who received appropriate antibiotic therapy [36].
The present study has some limitations. First, it was a single-center study with a relatively small number of patients, and whether the present results can be generalized to other areas and other countries is unknown. Second, this study was also an observational cohort study like many previous studies, and a future RCT is needed to con rm these results.
Regardless of the above limitations, the present study has some strengths. It was the rst study of macrolide combination therapy in an Asian country. All previous studies that showed the e cacy of macrolide combination therapy were conducted in the United States and Europe. Therefore, the results of the present study are thought to be signi cant because they show the usefulness of macrolide combination therapy in other countries. Furthermore, the present study reduced some biases as much as possible using propensity score analysis, although it was not an RCT.

Conclusions
In conclusion, macrolide combination therapy is useful for reducing mortality in SCAP patients hospitalized in the ICU compared with β-lactam monotherapy or non-macrolide-containing β-lactam combination therapy. To improve the prognosis of SCAP patients, among the guideline-concordant therapies, macrolide combination therapy may contribute to better survival.

Abbreviations
CAP community-acquired pneumonia; CURB-65:confusion, urea > 7 mmol/L, respiratory rate ≥ 30 breaths per minute, low blood pressure (systolic < 90 mmHg or diastolic ≤ 60 mmHg), and age ≥ 65 years; HCAP:healthcare-associated pneumonia; ICU:intensive care unit; IPTW:inverse probability of treatment weighting; MSSA:methicillin-susceptible Staphylococcus aureus; PS:propensity score; PSI:Pneumonia Severity Index; RCT:randomized, controlled trial; SCAP:severe community-acquired pneumonia Declarations Ethics approval and consent to participate This study was approved by the institutional review board of Kurashiki Central Hospital (approval number 3398). All patients gave their informed consent to participate in this study. The study was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.

Consent for publication
Not applicable Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
All authors have no con icts of interest to declare.

Funding
This research did not receive any speci c grant from funding agencies in the public, commercial, or not-for-pro t sectors.
Authors' contributions AI served as the principal author, had full access to all data in the study, and takes responsibility for the integrity and accuracy of the data and data analysis. AI, TI, HT, YN, FT, AY, YW, HI, and TO contributed to the study conception and design and to the acquisition of data. AI, TI, and HT contributed to the analysis and interpretation of data. AI, TI, HT, YN, FT, AY, YW, HI, and TO contributed to the drafting and revision of the manuscript and approved the nal version to be submitted for consideration for publication. Figure 1 Study owchart Figure 2 30-day mortality, 14-day mortality, and ICU mortality in the non-macrolide-containing group and the macrolide combination group A. 30-day mortality, B. 14-day mortality, C. ICU mortality The non-macrolide-containing group includes β-lactam monotherapy and β-lactam and non-macrolide combination therapy. ICU, intensive care unit