Antibiotic Therapy for Adults Hospitalized With Community-Acquired Pneumonia A Systematic Review

Clinical Review & Education Review Antibiotic Therapy for Adults Hospitalized With Community-Acquired Pneumonia A Systematic Review Jonathan S. Lee, MD; Daniel L. Giesler, MD, PharmD; Walid F. Gellad, MD, MPH; Michael J. Fine, MD, MSc IMPORTANCE Antibiotic therapy is the cornerstone of medical management for community-acquired pneumonia. OBJECTIVE To assess the associations between 3 key aspects of antibiotic therapy (optimal time to antibiotic initiation, initial antibiotic selection, and criteria for the transition from intravenous to oral therapy) and short-term mortality in adults hospitalized with community-acquired pneumonia. Author Audio Interview at jama.com Supplemental content at jama.com CME Quiz at jamanetworkcme.com and CME Questions page 607 EVIDENCE REVIEW Bibliographic databases of MEDLINE, EMBASE, and the Cochrane Collaboration were searched for studies of adults hospitalized with radiographically confirmed community-acquired pneumonia published from January 1, 1995, until November 5, 2015. FINDINGS Twenty studies (17 observational and 3 randomized trials) met eligibility criteria. Among 8 observational studies identified, the 4 largest (study populations of 2878 to 1 170 022) found that antibiotic initiation within 4 to 8 hours of hospital arrival was associated with relative reductions of 5% to 43% in mortality; the 4 smallest studies (study populations of 451 to 2076) found no associations between the timing of antibiotic initiation and mortality. One cluster randomized trial (n = 1737) demonstrated noninferiority of β-lactam monotherapy (n = 506) vs β-lactam plus macrolide combination therapy (n = 566), with an absolute adjusted difference of 2.5% (90% CI, 0.6% to 5.2%) in 90-day mortality favoring β-lactam monotherapy. A second randomized trial (n = 580) failed to demonstrate noninferiority of β-lactam monotherapy vs β-lactam plus macrolide combination therapy, with an absolute difference of 7.6% (1-sided 90% CI upper limit, 13.0%) in attainment of clinical stability on hospital day 7 favoring β-lactam plus macrolide combination therapy. Six of 8 observational studies (study populations of 1188 to 24 780) found that β-lactam plus macrolide combination therapy was associated with relative reductions of 26% to 68% in short-term mortality and all 3 observational studies (study populations of 2068 to 24 780) reported that fluoroquinolone monotherapy was associated with relative reductions of 30% to 43% in mortality compared with β-lactam monotherapy. One randomized trial (n = 302) reported significantly reduced hospital length of stay (absolute difference, 1.9 days; 95% CI, 0.6 to 3.2 days), but no differences in treatment failure when objective clinical criteria were used to decide when to transition patients from intravenous to oral therapy. CONCLUSIONS AND RELEVANCE In adults hospitalized with community-acquired pneumonia, antibiotic therapy consisting of β-lactam plus macrolide combination therapy or fluoroquinolone monotherapy initiated within 4 to 8 hours of hospital arrival was associated with lower adjusted short-term mortality, supported predominantly by low-quality observational studies. One randomized trial supports the use of objective clinical criteria to guide the transition from intravenous to oral antibiotic therapy. JAMA. 2016;315(6):593-602. doi:10.1001/jama.2016.0115 Author Affiliations: Division of General Internal Medicine, Department of Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania (Lee, Giesler, Gellad, Fine); now with Division of General Internal Medicine, Department of Medicine, University of California, San Francisco (Lee); Center for Health Equity Research and Promotion, Veterans Affairs Pittsburgh Healthcare System, Pittsburgh, Pennsylvania (Gellad, Fine). Corresponding Author: Michael J. Fine, MD, MSc, Center for Health Equity Research and Promotion, VA Pittsburgh Healthcare System, 151C University Dr, Bldg 30, Pittsburgh, PA 15240 (michael.fine@va.gov). Section Editors: Edward Livingston, MD, Deputy Editor, and Mary McGrae McDermott, MD, Senior Editor. (Reprinted) 593

Clinical Review & Education Review Antibiotic Therapy for Community-Acquired Pneumonia in Adults Each year in the United States, community-acquired pneumonia is responsible for approximately 600 000 hospital admissions and generates $10.6 billion in health care costs. 1,2 Community-acquired pneumonia and influenza combined represent the eighth leading cause of death in the United States. 3 Patients with community-acquired pneumonia typically present with fever, shortness of breath, and cough. Common pulmonary signs are dullness to percussion, crackles, and egophony. 4 In patients without recent hospitalizations or exposure to outpatient health care facilities, these clinical findings along with radiographic evidence of lung consolidation are diagnostic of communityacquired pneumonia. 5 Advanced age, altered mental status, comorbid conditions, and vital sign abnormalities are associated with an increased risk of mortality, and patients with a greater number of these characteristics oftenrequirehospitalization. 6,7 Arecentstudyofolderpatients( 65 years) hospitalizedwithcommunity-acquiredpneumoniafrom2006 to 2010 documented an overall 30-day mortality of 6.5%. 8 Despite advancements in diagnostic testing, specific pathogens are not detected in most patients with community-acquired pneumonia. 9 Although recent studies have demonstrated that viral pathogens are common in community-acquired pneumonia, 10,11 patients are considered to have bacterial pneumonia at the time of admission due in part to the absence of widely available point-ofcare testing for the full range of pathogens. Therefore, initiation of empirical antibiotic therapy covering the most likely bacterial pathogens is standard of care, and the timing of initiation, empirical selection, and the transition from intravenous to oral routes of therapy are key aspects of such treatment. 12 The reported associations between the time to initiate antibiotic therapy after hospital arrival and short-term mortality have varied in prior studies of community-acquired pneumonia, 13-15 and the optimal time to deliver the initial dose of antibiotic therapy after hospital arrival remains uncertain. Using the timeliness of antibiotic initiation as a national hospital quality metric previously led to concerns of antibiotic overuse in patients initially suspected of having community-acquired pneumonia. 16 Although typical and atypical bacterial pathogens can cause moderate to severe community-acquired pneumonia requiring hospitalization, the importance of including atypical coverage for Legionella pneumophila, Mycoplasma pneumoniae, and Chlamydia pneumoniae in the initial empirical selection of antibiotic therapy remains unclear. Broad-spectrum antibiotics that cover both atypical and typical bacterial pathogens could conceivably improve patient outcomes; however, this therapeutic approach may also increase the risk of antimicrobial resistance and adverse drug effects. 17-19 In addition, because most adults hospitalized with communityacquired pneumonia are moderately to severely ill, empirical antibiotic therapy is typically initiated intravenously. Few studies have tested the effectiveness or safety of using objective clinical criteria to guide the transition from intravenous to oral antibiotic treatment. This transition from intravenous to oral therapy is a necessary step for hospital discharge and has the potential to safely reduce hospital length of stay and use of health care resources. This review provides an evidence-based assessment of the following clinical questions pertaining to antibiotic therapy for adults hospitalized with community-acquired pneumonia outside intensive care unit (ICU) settings: (1) Is earlier initiation of antibiotic therapy after hospital arrival associated with better outcomes? (2) Is initial empirical therapy that covers both typical and atypical bacterial pathogens (β-lactam plus macrolide combination therapy or fluoroquinolone monotherapy) associated with better outcomes than therapy covering typical pathogens only (β-lactam monotherapy)? (3) Which clinical criteria can be used to determine when patients can be safely transitioned from intravenous to oral therapy? Take-Home Points Antibiotic therapy should be initiated within 4 to 8 hours of hospital arrival for patients with radiographically confirmed pneumonia and moderate to high levels of illness severity at presentation. Initial first-line antibiotic therapy should consist of β-lactam plus macrolide combination therapy or fluoroquinolone monotherapy for hospitalized patients with community-acquired pneumonia treated outside an intensive care unit setting. Patients meeting all of the following criteria for at least 24 hours can be transitioned from intravenous to oral therapy: 1. Absence of mental confusion 2. Ability to take oral medication 3. Temperature lower than 38.3 C 4. Hemodynamic stability (heart rate <100 beats/min and systolic blood pressure >90 mm Hg) 5. Respiratory rate lower than 25 breaths/min 6. Oxygensaturationhigherthan90%, arterialoxygenpartialpressure higher than 60 mm Hg while breathing in normal room air or low-flow supplemental oxygen by nasal cannula, or return to baseline oxygen level for patients receiving long-term oxygen therapy. Methods The databases of MEDLINE, EMBASE, and the Cochrane Collaboration were searched for English-language studies of adults hospitalized with radiographically confirmed community-acquired pneumonia published from January 1, 1995, until November 5, 2015. Studies consisting predominantly of patients managed outside ICU settings were included, and those that restricted enrollment to select patient subgroups (eg, health care associated or bacteremic pneumonia) were excluded. Studies that included patients admitted from skilled nursing facilities that are now defined as health care associated pneumonia, but were previously classified as community-acquired pneumonia, were included. Additional studies were identified by reviewing the bibliographies of retrieved articles and the 2007 practice guidelines for community-acquired pneumonia developed by the Infectious Diseases Society of America and the American Thoracic Society (IDSA/ATS). 12 Two of the study authors (J.S.L. and D.L.G.) independently reviewed articles for inclusion and assessed quality using the Grades of Recommendation, Assessment, Development and Evaluation criteria. 20 Disagreements were resolved by consensus among all authors. For question 1, only studies that compared patient outcomes by time to initiate antibiotic therapy were included. For question 2, only 594 JAMA February 9, 2016 Volume 315, Number 6 (Reprinted) jama.com

Antibiotic Therapy for Community-Acquired Pneumonia in Adults Review Clinical Review & Education studies evaluating antibiotic coverage with β-lactam plus macrolide combination therapy or fluoroquinolone monotherapy vs β-lactam monotherapy regimens were included. The review was limited to studies evaluating the following first-line antibiotics recommended by the 2007 IDSA/ATS practice guidelines: β-lactams, macrolides (azithromycin, clarithromycin, erythromycin) and respiratory fluoroquinolones (moxifloxacin, gemifloxacin, levofloxacin). In addition, the reporting of adverse drug effects and the periods after hospital admission used to define initial empirical therapy were also assessed for question 2. For both questions 1 and 2, findings from randomized trials were emphasized and observational studies were only included if patient outcomes were adjusted for baseline severity of illness or age and comorbidity. For question 3, only clinical trials that directly compared the use of clinical criteria to determine when to transition from intravenous to oral antibiotic therapy with a fixed 7- to 10-day duration of intravenous antibiotic therapy were included. For observational studies, the associations of treatment differences and outcomes (adjusted for potential confounders) were reported as in the original studies. For clinical trials, unadjusted and adjusted outcomes were assessed as reported in the original studies. Further details of study eligibility criteria, search strategies, and article identification and selection appear in the emethods and in efigures 1, 2, and 3 in the Supplement. Results Optimal Length of Time to Initiate Antibiotic Therapy After Hospital Arrival Five retrospective studies, 8,13-15,21 2 prospective cohort studies, 22,23 and 1 secondary analysis of a cluster randomized trial 24 were identified evaluating the associations between the time to initiate antibiotic therapy and patient outcomes (Table 1). Five of these studies assessed 30-day mortality and 3 assessed in-hospital mortality. All 8 studies were assessed as having low quality. The 4 largest studies demonstrated statistically significant associations between the timing of initial antibiotic therapy and reduced short-term mortality (Table 1 and Figure 1). A retrospective study 13 of 14 069 older patients ( 65 years) reported that receipt of antibiotics within 8 hours or less (vs >8 hours) of hospital arrival was associated with a relative reduction of 15% for death at 30 days, adjusting for illness severity with the pneumonia severity index. A subsequent study 15 of 13 771 older patients ( 65 years) that used a similar design and severity adjustment method also reported that antibiotic initiation within 4 hours or less of arrival was associated with a relative reduction of 15% for 30-day mortality. A more recent retrospective study 8 of 1 170 022 older patients ( 65 years) reported that antibiotic initiation within 6 hours of hospital arrival was associated with a relative reduction of 5% for 30-day mortality, adjusting for patient demographics and comorbid conditions. In addition, a retrospective cohort study 21 of 2878 patients aged 19 years or older reported that antibiotic initiation within 8 hours of hospital arrival was associated with a relative reduction of 43% for hospital mortality. The 4 smallest studies 14,22-24 found no significant associations between time to initiate antibiotic therapy and short-term mortality in adults of all ages. Initial Empirical Selection of Antibiotic Therapy Two randomized controlled trials, 25,26 4 retrospective studies, 27-30 3 prospective cohort studies, 31-33 and 2 secondary analyses of combined cohort study and clinical trial data 34,35 were identified of β-lactam plus macrolide combination therapy vs β-lactam monotherapy (n = 10 studies) and fluoroquinolone monotherapy vs β-lactam monotherapy (n = 4 studies) (Table 2). The period used to define initial empirical antibiotic therapy varied from 24 to 48 hours after hospital admission in the observational studies. All studies assessed mortality at varying short-term follow-up intervals; one assessed clinical stability on hospital day 7. The specific names of the antibiotics and the antibiotic classes used in each study appear in the etable in the Supplement. Two recently published high-quality randomized trials that used different designs and primary outcomes demonstrated conflicting results regarding the effectiveness of initial antibiotic regimens. First, 580 inpatients with community-acquired pneumonia were randomly allocated to receive β-lactam monotherapy or β-lactam plus macrolide combination therapy in a noninferiority trial conducted at 6 Swiss hospitals. 25 More patients in the β-lactam monotherapy group (41.2%) than in the combination therapy group (33.6%) did not reach clinical stability on hospital day7(p =.07). The absolute difference between groups was 7.6% and because the upper limit of the 1-sided 90% CI was 13.0% and exceeded the predefined 8% noninferiority boundary, this trial did not demonstrate noninferiority of β-lactam monotherapy. Patients in the monotherapy group also had a nonsignificantly higher 30-day mortality compared with those in the combination therapy group (4.8% vs 3.4%, respectively; P =.42). Second, in a pragmatic, cluster randomized noninferiority trial, 2283 patients with clinically suspected community-acquired pneumonia admitted outside the ICU setting at 7 Dutch hospitals were randomly allocated at the hospital level in rotating 4-month blocks to treatment with β-lactam monotherapy, β-lactam plus macrolide combination therapy, or fluoroquinolone monotherapy. 26 In intention-to-treat analyses of the subgroup of 1737 (76.1%) patients with radiographically confirmed community-acquired pneumonia, the absolute-adjusted 90-day mortality was lower by 2.5% (90% CI, 0.6% to 5.2%) in the β-lactam monotherapy group than in the β-lactam plus macrolide combination therapy group and higher by 0.7% (90% CI, 3.4% to 1.8%) in the β-lactam monotherapy group than in the fluoroquinolone monotherapy group. Based on a prespecified noninferiority boundary of 3% for 90-day mortality, the trial demonstrated that β-lactam monotherapy was noninferior to β-lactam plus macrolide combination therapy, but was not noninferior to fluoroquinolone monotherapy in this subgroup with radiographically confirmed community-acquired pneumonia. Six of 8 low-quality observational studies demonstrated benefits for β-lactam plus macrolide combination therapy compared with β-lactam monotherapy. A retrospective study 28 of 12 945 older patients ( 65 years) showed that compared with β-lactam monotherapy, third- and second-generation combination therapies of cephalosporin plus macrolide were associated with relative reductions of 26% and 29%, respectively, for 30-day adjusted mortality (Table 2 and Figure 2). Another retrospective study 27 of 24 780 older patients ( 65 years) also jama.com (Reprinted) JAMA February 9, 2016 Volume 315, Number 6 595

Clinical Review & Education Review Antibiotic Therapy for Community-Acquired Pneumonia in Adults Table 1. Description of Studies Evaluating Time to Initiate Antibiotic Therapy and Medical Outcomes for Patients Hospitalized With Community-Acquired Pneumonia Source Study Design Threshold Evaluated <4 h Period of Data Collection Study Sites Patients Illness Severity Risk Eligible Adjustment Study Age, y a Age, y a PSI Class b No. (%) Method Outcomes c Quality d Houck et al, 15 2004 Retrospective 1998-1999 3463 13 771 65 NR IV-V 9777 (71) PSI 30-d Mortality Low Waterer et al, 23 2006 Prospective 1998-2001 1 451 Adults 58 (mean) IV-V 131 (29) Individual covariates In-hospital mortality Low Lee et al, 24 2011 Prospective e 2001 32 2076 18 74 (median) IV-V 1007 (49) PSI 30-d Mortality Low Simonetti et al, 22 2012 Prospective 2001-2009 1 1274 Adults NR IV-V 712 (56) Individual covariates Threshold Evaluated <6 h 30-d Mortality Low Lee et al, 8 2014 Retrospective 2006-2010 4740 1 170 022 65 80 (mean) Elixhauser Elixhauser index 30-d Mortality Low conditions f Threshold Evaluated <8 h Meehan et al, 13 1997 Retrospective 1994-1995 3555 14 069 65 79 (mean) IV-V 9613 (68) PSI 30-d Mortality Low Dedier et al, 14 2001 Retrospective 1997-1998 38 1062 18 64 (median) IV-V 555 (52) PSI In-hospital mortality Low Arnold et al, 21 2007 Retrospective 2001-2006 39 2878 19 65 (mean) IV-V 50-60 g PSI In-hospital mortality Low Simonetti et al, 22 2012 Prospective 2001-2009 1 1274 Adults NR IV-V 712 (56) Individual covariates Abbreviations: NR, not reported; PSI, pneumonia severity index. a ForthestudiesbyWatereretal 23 and Simonetti et al, 22 the populations consisted of adults, but no age eligibility criteria were specified. Mean or median study population ages are presented due to variability in the reporting of age. Some values were estimated based on available data. b The PSI is a validated prediction rule for pneumonia prognosis that stratifies patients into 5 risk classes for 30-day mortality based on 20 patient demographic and clinical characteristics. In the original model derivation and validation, 30-day mortality ranged from 0.1% to 0.4% for class I to 27.0% to 31.1% for class V. The study by Leeetal 8 used the Elixhauser index, which is a validated comorbidity index of 30 conditions for use with administrative data. Some values were estimated based on available data. c The in-hospital and 30-day mortality outcomes include deaths from all causes during the index hospitalization and within 30 days of admission, respectively. 30-d Mortality Low d Independently classified by 2 of the study authors as high, moderate, low, or very low using the Grades of Recommendation, Assessment, Development and Evaluation system. Classification disagreements were resolved by consensus among all study authors. e This study was a secondary analysis of a cluster randomized trial that was not originally designed to assess the association between time to initiate antibiotic therapy and medical outcomes for patients hospitalized with community-acquired pneumonia. This was considered as a prospective cohort study because it was not designed to answer the study question. f There were a mean of 3.7 conditions per patient. g Proportion range of patients in PSI risk classes visually estimated from a figure published in the original study. 596 JAMA February 9, 2016 Volume 315, Number 6 (Reprinted) jama.com

Antibiotic Therapy for Community-Acquired Pneumonia in Adults Review Clinical Review & Education Figure 1. Studies Assessing Initiation of Antibiotic Therapy Within Various Time Thresholds and Short-term Mortality for Patients Hospitalized With Community-Acquired Pneumonia <Time Threshold >Time Threshold Source Threshold evaluated <4 h Outcome Patients No. (%) Who Died Patients No. (%) Who Died Adjusted OR (95% CI) Houck et al, 15 2004 30-d Mortality 8388 973 (11.6) 5383 684 (12.7) 0.85 (0.76-0.95) Waterer et al, 23 2006 In-hospital mortality NR NR NR NR 0.54 (0.20-1.19) Lee et al, 24 2011 30-d Mortality 1619 107 (6.6) 443 34 (7.7) 0.74 (0.48-1.13) Simonetti et al, 22 2012 30-d Mortality 477 33 (6.9) 797 37 (4.6) 1.12 (0.38-3.33) Threshold evaluated <6 h Lee et al, 8 2014 30-d Mortality 1102555 122 384 (11.1) 67 467 7421 (11.0) 0.95 (0.93-0.98) Threshold evaluated <8 h Meehan et al, 13 1997 30-d Mortality NR NR NR NR 0.85 (0.75-0.96) Dedier et al, 14 2001 In-hospital mortality 809 NR 253 NR 1.69 (0.78-3.66) Arnold et al, 21 2007 In-hospital mortality NR NR NR NR 0.57 (0.44-0.74) Simonetti et al, 22 2012 30-d Mortality 1030 58 (5.6) 244 12 (4.9) 1.58 (0.64-3.88) Favors Earlier Treatment Favors Less Early Treatment 0.1 1.0 Adjusted OR (95% CI) 10 Time threshold evaluated <4 h. Some values were estimated based on available data. NR indicates not reported; OR, odds ratio. reported that β-lactam plus macrolide combination therapy was associated with a relative reduction of 30% for adjusted 30-day mortality. A third retrospective study 29 of 10 069 older patients ( 65 years) hospitalized during 3 separate years reported that β-lactam plus macrolide combination therapy was associated with a lower adjusted30-daymortalityin1993, butnotin1995or1997. Threesmaller observational studies 30,32,34 in adults of all ages also reported that various β-lactam plus macrolide combinations were associated with lower adjusted short-term mortality compared with β-lactam monotherapy. Only a secondary analysis 35 of 451 patients prospectively enrolled in a multicenter study and a prospective cohort study 33 of 1854 patients demonstrated no significant association with 30-day mortality. All 3 of the observational studies 27,31,34 reviewed consistently favored the use of fluoroquinolone monotherapy compared with β-lactam monotherapy (Figure 2). A retrospective study 27 of 24 780 older patients ( 65 years) and 2 smaller studies 31,34 in adults of all ages reported significant associations between fluoroquinolone monotherapyandreducedmortalitycomparedwithβ-lactammonotherapy at follow-up intervals ranging from the end of therapy to 6 months following presentation. Only the 2 randomized trials reported adverse events related to antibiotic therapy. The Swiss trial found no differences in adverse events between the β-lactam plus macrolide combination and β-lactam monotherapy groups, whereas the Dutch trial reported a higher frequency of change in the class of antibiotics due to adverse effects for patients assigned to β-lactam plus macrolide combination therapy (7.2%) than for those assigned to β-lactam (1.7%) orfluoroquinolonemonotherapy(2.9%); however, no testing for statistical significance was reported. Transition From Intravenous to Oral Antibiotic Therapy Only 1 high-quality randomized controlled trial 36 that used objective clinical criteria to evaluate the transition from intravenous to oral antibiotic therapy in patients hospitalized for community-acquired pneumonia was identified. In this trial, 302 patients with moderate to severe illness severity hospitalized at 7 Dutch hospitals were randomized to early transition or usual therapy. On hospital day 3, patients in the intervention group were transitioned to oral antibioticsiftheywereclinicallystableusingthefollowingcriteria: respiratory rate lower than 25 breaths/min, oxygen saturation higher than 90% or arterial oxygen partial pressure higher than 55 mm Hg, hemodynamic stability, greater than 1 C decrease in temperature in patients with a fever, absence of mental confusion, and the ability to tolerate oral medications. Patients in the intervention group completed 10 total days of antibiotic therapy and those in the control group received a total of 7 days of intravenous therapy. The duration of intravenous antibiotic therapy (absolute difference, 3.4 days; 95% CI, 2.8-3.9 days) and hospital length of stay (absolute difference, 1.9 days; 95% CI, 0.6-3.2 days) were significantly shorter in the early transition group than in the conventional therapy group, with no significant difference between groups in treatment failure defined as death, continued hospitalization at day 28, or clinical deterioration. Mortality did not differ between groups, but the study was not powered to detect a difference in this outcome. Discussion Ourevidencesynthesisoftheinitiation, selection, andtransitionfrom intravenous to oral antibiotic therapy for adults with communityacquired pneumonia managed outside the ICU hospital setting identified an association between initiating antibiotic therapy within 4 to 8 hours of hospital arrival and reduced mortality in 4 of the 8 observational studies. Although 2 recent multicenter randomized trials 25,26 reached differing conclusions regarding the noninferiority of β-lactam monotherapy compared with β-lactam plus macrolide combination therapy, 6 of 8 observational studies 27-30,32,34 reported an association between β-lactam plus macrolide combination therapy and reduced short-term mortality. One randomized trial 26 and 3 observational studies 27,31,34 support the superiority of fluoroquinolone monotherapy compared with jama.com (Reprinted) JAMA February 9, 2016 Volume 315, Number 6 597

Clinical Review & Education Review Antibiotic Therapy for Community-Acquired Pneumonia in Adults Table 2. Description of Studies Comparing β-lactam Plus Macrolide Combination Therapy or Respiratory Fluoroquinolone Monotherapy vs β-lactam Monotherapy for Patients Hospitalized With Community-Acquired Pneumonia a Period Illness Severity Risk of Data Study Eligible Class or Adjustment Study Source Study Design Collection Sites Patients Age, y Age, y b Score c No. (%) Method Outcomes d Quality e β-lactam Plus Macrolide Combination Therapy vs β-lactam Monotherapy Gleason et al, 28 1999 Retrospective 1994-1995 NR 12 945 65 79 (mean) PSI IV-V 8841 (68) PSI 30-d Mortality Low Houck et al, 29 2001 Retrospective 1993, 1995, NR 10 069 65 79 (mean) PSI IV-V 6681 (66) PSI 30-d Mortality Low 1997 García Vázquez et al, 32 2005 Prospective 1996-2001 1 1188 15 68 (mean) PSI IV-V 772 (65) PSI In-hospital mortality Low Paul et al, 35 2007 Prospective f 2002-2004 3 451 18 67 (mean) PSI IV-V 298 (66) PSI 30-d Mortality Low Bratzler et al, 27 2008 Retrospective 1998-2001 NR 24 780 65 NR PSI IV-V 17 619 (64) g PSI 30-d Mortality Low Blasi et al, 34 2008 2002-2004 31 2847 19 79 (mean) PSI IV-V 2847 (100) PSI Mortality at the end Retrospective/ Low prospective f of antibiotic therapy Tessmer et al, 33 2009 Prospective 2002-2006 10 1854 18 66 (mean) CRB-65 score 519 (28) CRB-65 30-d Mortality Low 2 Rodrigo et al, 30 2013 Retrospective 2009-2011 72 5240 17 73 and 76 (medians) CURB-65 score 3 1513 (29) CURB-65 In-hospital mortality at 30 d Low Garin et al, 25 2014 Noninferiority RCT 2009-2013 6 580 18 76 (median) PSI IV 240 (41) None Clinical stability at day 7 High and 30-d mortality Postma et al, 26 2015 Noninferiority RCT 2011-2013 7 2283 h 18 70 (median) PSI III i PSI 90-d Mortality High Fluoroquinolone Monotherapy vs β-lactam Monotherapy Bratzler et al, 27 2008 Retrospective 1998-2001 NR 24 780 65 NR PSI IV-V 17 619 (64) g PSI 30-d Mortality Low Blasi et al, 34 2008 2002-2004 31 2847 19 79 (mean) PSI IV-V 2847 (100) PSI Mortality at the end Retrospective/ Low prospective f of antibiotic therapy Ewig et al, 31 2011 Prospective 2002-2007 10 2068 18 64 (mean) CRB-65 score 578 (28) CRB-65 6-mo Mortality Low 2 Postma et al, 26 2015 Noninferiority RCT 2011-2013 7 2283 h 18 70 (median) PSI III i PSI 90-d Mortality High Abbreviations: CURB-65, confusion, urea nitrogen, respiratory rate, blood pressure, 65 years of age and older; CRB-65, confusion, respiratory rate, blood pressure, 65 years of age and older; NR, not reported; PSI, pneumonia severity index; RCT, randomized controlled trial. a Intervention group regimens consisted of β-lactam plus macrolide combination therapy or respiratory fluoroquinolone monotherapy and provided coverage for atypical and typical bacterial pathogens. All comparisons were made with β-lactam monotherapy providing coverage for typical pathogens only. Detailed antibiotic regimens used for each study are available in the etable in the Supplement. b Mean or median study population ages are presented due to variability in the reporting of age. For Rodrigo et al, 30 73 years was the median age for the β-lactam plus macrolide combination therapy group and 76 years was for the β-lactam monotherapy group. Some values were estimated based on available data. c The PSI, CURB-65, and CRB-65 are validated prediction rules for pneumonia prognosis. The PSI stratifies patients into 5 risk classes for 30-day mortality based on 20 patient demographic and clinical characteristics. In the original model derivation and validation, 30-day mortality ranged from 0.1% to 0.4% for class I and 27.0% to 31.1% for class V. The CURB-65 uses 5 variables to stratify patients on a 6-point scale (0-5) and 30-day mortality in the original model derivation ranged from 0.7% to 9.2% for scores of 0 to 2 and 14.5% to 40% for scores of 3 or greater. The CRB-65 uses 1 less variable to stratify patients on a 5-point scale (range, 0-4) and 30-day mortality in the original model derivation ranged from 1.2% to 5.3% for scores of 0 to 1 and 12.2% to 32.9% for scores of 2 or greater. Some values were estimated based on available data. d Deaths from all causes during the index hospitalization (in-hospital mortality) and deaths within 14 days, 30 days, 90 days, and 6 months of admission. Mortality at the end of antibiotic therapy includes deaths prior to the completion of antibiotic therapy. The 30-day in-hospital mortality outcome includes inpatient deaths within 30 days of hospital admission. When studies reported multiple time points for mortality, 30-day mortality was preferentially selected, if available. Clinical stability was defined as a heart rate of less than 100 beats/min, systolic blood pressure of greater than 90 mm Hg, temperature of less than 38.0 C, respiratory rate of less than 24 breaths/min, and oxygen saturation of greater than 90% while breathing room air. e Independently classified by 2 of the study authors as high, moderate, low, or very low using the Grades of Recommendation, Assessment, Development and Evaluation system. Classification disagreements were resolved by consensus among all study authors. f These studies were secondary analyses of combined data from cohort studies and clinical trials that were not designed to assess the associations between antibiotic selection and medical outcomes for patients hospitalized with community-acquired pneumonia. They were considered as cohort studies because they were not designed to answer the study question. g Data are for the entire 27 730 patient cohort, which includes 2950 patients admitted to an intensive care unit. h Of these 2283 patients, 1737 (76.1%) had radiographically confirmed community-acquired pneumonia. i Mean PSI score of 85. 598 JAMA February 9, 2016 Volume 315, Number 6 (Reprinted) jama.com

Antibiotic Therapy for Community-Acquired Pneumonia in Adults Review Clinical Review & Education Figure 2. Studies Assessing Short-term Mortality for β-lactam Plus Macrolide Combination Therapy or Respiratory Fluoroquinolone Monotherapy vs β-lactam Monotherapy for Patients Hospitalized With Community-Acquired Pneumonia Source Outcome β-lactam plus macrolide combination therapy Gleason et al, 28 1999 Second-generation cephalosporin plus macrolide Third-generation cephalosporin plus macrolide Houck et al, 29 2001 Cephalosporin or β-lactam/ β-lactamase inhibitor plus macrolide c Cephalosporin or β-lactam/ β-lactamase inhibitor plus macrolide d Cephalosporin or β-lactam/ β-lactamase inhibitor plus macrolide e β-lactam Plus Macrolide or Fluoroquinolone Monotherapy Patients No. (%) Who Died β-lactam Monotherapy Patients No. (%) Who Died Adjusted OR (95% CI) a 30-d Mortality 544 46 (8.4) 3430 511 (14.9) 0.71 (0.52-0.96) b 30-d Mortality 1139 104 (9.1) 3430 511 (14.9) 0.74 (0.60-0.92) b 30-d Mortality 312 26 (8.3) 1740 242 (13.9) 0.42 (0.25-0.69) 30-d Mortality 561 48 (8.6) 1982 234 (11.8) 0.93 (0.62-1.41) 30-d Mortality 870 89 (10.2) 1758 244 (13.9) 0.87 (0.63-1.19) García Vázquez et al, 32 2005 In-hospital mortality 918 63 (6.9) 270 36 (13.3) 0.50 (0.31-0.81) Paul et al, 35 2007 30-d Mortality 282 21 (7.4) 169 37 (21.9) 0.69 (0.32-1.48) Bratzler et al, 27 2008 30-d Mortality 5963 338 (5.7) 4463 376 (8.4) 0.7 (0.6-0.9) Blasi et al, 34 2008 End of therapy mortality 330 19 (5.7) 452 73 (16.2) 0.32 (0.19-0.56) Tessmer et al, 33 2009 30-d Mortality 946 42 (4.4) 908 78 (8.6) 1.04 (0.66-1.63) Rodrigo et al, 30 2013 30-d In-hospital 3239 745 (23.0) 2001 536 (26.8) 0.72 (0.60-0.85) mortality Garin et al, 25 2014 30-d Mortality 289 10 (3.4) 291 14 (4.8) 0.71 (0.32-1.59) f Postma et al, 26 2015 30-d Mortality 566 NR 506 NR 1.37 (0.88-2.13) g Fluoroquinolone monotherapy Bratzler et al, 27 2008 30-d Mortality 5045 318 (6.3) 4463 376 (8.4) 0.7 (0.6-0.9) Blasi et al, 34 2008 End of therapy mortality 363 33 (9.1) 452 73 (16.2) 0.59 (0.37-0.94) Ewig et al, 31 2011 6-mo Mortality 365 NR 1703 NR 0.57 (0.35-0.92) b Postma et al, 26 2015 90-d Mortality 665 NR 506 NR 0.91 (0.58-1.42) g Favors β- Lactam Plus Macrolide or Fluroquinolone Monotherapy Favors β- Lactam Monotherapy 0.1 1.0 Adjusted OR (95% CI) 10 Some values were estimated based on available data. NR indicates not reported; OR, odds ratio. a Unless otherwise indicated. b Hazard ratio not adjusted OR. c Data collected in 1993. d Data collected in 1995. e Data collected in 1997. f Calculated using available data and is unadjusted. g Data are for subgroup with radiographically confirmed pneumonia. β-lactam monotherapy. A single randomized trial 36 with limited power found that using objective clinical criteria to decide when to transitionfromintravenoustooralantibioticsreducedhospitallength of stay with no significant differences in medical outcomes. For time to initiate antibiotic therapy, all 8 eligible studies were observational in design and received grades indicating low quality. The 4 studies 8,13,15,21 demonstrating a significant association between more rapid initiation of antibiotic therapy and reduced mortality were the largest identified, and 3 of these 8,13,15 were conducted in older patients ( 65 years) with greater illness severity at presentation. In contrast, the 4 smallest studies, 14,22-24 which included adults of all ages with lesser illness severity, found no significant associations between timeliness of antibiotic initiation and mortality. For initial antibiotic selection, 2 of the eligible studies 25,26 were recently conducted high-quality randomized trials and 9 were lowquality observational studies. 27-35 The Swiss and Dutch trials reached opposite conclusions regarding the noninferiority of β-lactam monotherapy compared with β-lactam plus macrolide combination therapy using distinctly different primary outcomes. Even though the Dutch trial demonstrated that β-lactam monotherapy was noninferior to β-lactam plus macrolide combination therapy using mortality at 90 days as the primary outcome, certain aspects of the trial design diminish confidence in the findings. The Swiss trial, which failed to establish the noninferiority of β-lactam monotherapy using clinical stability on hospital day 7 as the primary outcome, used well-defined criteria to justify deviation from the randomly assigned antibiotic therapy (eg, lack of fever resolu- jama.com (Reprinted) JAMA February 9, 2016 Volume 315, Number 6 599

Clinical Review & Education Review Antibiotic Therapy for Community-Acquired Pneumonia in Adults Box. 2007 Guideline Recommendations From the Infectious Diseases Society of America and the American Thoracic Society Regarding Antibiotic Therapy for Patients Hospitalized With Community-Acquired Pneumonia Treated Outside Intensive Care Unit Settings a Question 1: Is earlier initiation of antibiotic therapy after hospital arrival associated with better medical outcomes? Answer: For patients admitted through the emergency department, the first antibiotic dose should be administered while the patient is still in the emergency department. Question 2: Is empirical β-lactam plus macrolide combination therapy or fluoroquinolone monotherapy associated with better medical outcomes than β-lactam monotherapy? Answer: Empirical antibiotic therapy should cover both typical and atypical bacterial pathogens with the following first- and second-line regimens. First-line regimens: Respiratory fluoroquinolone monotherapy (moxifloxacin, gemifloxacin, or levofloxacin) or β-lactam (cefotaxime, ceftriaxone, or ampicillin) and macrolide (azithromycin, clarithromycin, or erythromycin) combination therapy Second-line regimens: β-lactam (cefotaxime, ceftriaxone, or ampicillin) and doxycycline combination therapy Question 3: Which clinical criteria can be used to determine when patients can be safely transitioned from intravenous to oral therapy? Answer: Patients should be transitioned from intravenous to oral therapy when they are hemodynamically stable and improving clinically, are able to ingest medications, and have a normally functioning gastrointestinal tract. a Adapted from the 2007 practice guidelines for community-acquired pneumonia developed by the Infectious Diseases Society of America and the American Thoracic Society. 12 tion, isolation of a resistant organism), whereas the Dutch trial permitted a much broader range of accepted deviation from the randomly assigned antibiotic classes. In the Dutch trial, 27% of patients in the β-lactam monotherapy group received an antibiotic regimen covering atypical pathogens as initial therapy and 39% received such regimens at some point during their hospitalization. This large deviation from the randomly assigned treatment could contribute to an overestimation of the relative effectiveness of β-lactam monotherapy and to the finding of noninferiority. In addition, the noninferiority boundary for mortality of 3% used in this trial may be inappropriately large, considering that short-term mortality for patients hospitalized with community-acquired pneumonia ranges from 5% to 9%. 8,37,38 Previous reviews of the timeliness of antibiotic initiation and antibiotic selection have reached conflicting conclusions. A clinical review 39 of antibiotic initiation that evaluated 3 of the 8 studies included in this review and other studies that did not meet our inclusion criteria reached conclusions similar to those presented in this review. In contrast, a prior systematic review 40 that evaluated 10 studies, including 6 without severity adjustment that were not included in this review, concluded that the evidence did not support an association between rapid antibiotic initiation and mortality. Reviews of initial antibiotic selection also reached conflicting conclusions based in part on the types of studies included in their analyses. 41,42 Findings from the observational studies that we reviewed were consistent with a recent meta-analysis 41 with less stringent inclusion criteria comprising 16 observational studies, which demonstrated 33% lower odds of mortality with β-lactam plus macrolide combination therapy compared with β-lactam monotherapy. In contrast, a Cochrane review 42 of 28 predominately industry-funded randomized trials found no benefit in shortterm mortality or clinical success with empirical antibiotic regimens containing atypical coverage compared with regimens providing typical coverage only. However, none of the studies considered in this review met our inclusion criteria primarily because the antibiotic regimens evaluated (sparfloxacin, temafloxacin) were not recommended by the 2007 IDSA/ATS practice guidelines. Findings from a single trial of the transition from an intravenous to oral antibiotic considered in our evidence synthesis are supported by a previous meta-analysis 43 and a prospective cohort study 44 demonstrating consistent relationships between objective criteria for clinical stability and subsequent patient outcomes. In the meta-analysis, pooled data from 6 randomized trials evaluating the transition from intravenous to oral antibiotic therapy 2 to 4 days after antibiotic initiation showed reduced hospital length of stay in the early transition group compared with the conventional intravenous therapy group, with no significant between-group differences in treatment success or short-term mortality. In a prospective cohort study evaluating the time course for resolution of vital sign abnormalities and risk for clinical deterioration in patients hospitalized for community-acquired pneumonia, the median time to normalization of vital signs, mental status abnormalities, and ability to maintain oral intake was 3 days. Moreover, once normalization occurred, the risk of clinical deterioration requiring transfer to an ICU or telemetry monitoring unit was 1% or less across all pneumonia severity index classes. These findings suggest that the transition from intravenous to oral antibiotic therapy prior to hospital day 3 may be safe once objective criteria for clinical stability are met. The take-home messages from our systematic review were informed by multiple studies published after the release of the 2007 IDSA/ATS guidelines (Box). These guidelines recommend that hospitalized patients receive the first antibiotic dose in the emergency department, without a specific time window for initiation. Based on low-quality observational evidence, we recommend initiating antibiotic therapy within 4 to 8 hours of hospital arrival in patients with radiographically confirmed community-acquired pneumonia and moderate to severe illness severity at presentation. The IDSA/ATS guidelines recommend initial selection of β-lactam plus macrolide combination therapy or fluoroquinolone monotherapy for patients with community-acquired pneumonia managed outside the ICU hospital setting. Based predominately on lowquality, observational evidence incorporating findings from 8 studies published since the release of these guidelines, we recommend these same regimens covering atypical and typical bacterial pathogens compared with β-lactam monotherapy covering typical pathogens only. 600 JAMA February 9, 2016 Volume 315, Number 6 (Reprinted) jama.com

Antibiotic Therapy for Community-Acquired Pneumonia in Adults Review Clinical Review & Education The divergent conclusions from the 2 recent randomized trials concerning the noninferiority of β-lactam monotherapy, and the aforementioned limitations of the Dutch trial, preclude the highquality evidence from these trials from overriding consistent evidence from observational studies demonstrating improved survival with β-lactam plus macrolide combination therapy or fluoroquinolone monotherapy compared with β-lactam monotherapy. In addition, the IDSA/ATS guidelines recommend that hospitalized patients be switched from intravenous to oral antibiotic therapy when they are determined to be hemodynamically stable and improving clinically, are able to take oral medications, and have a normally functioning gastrointestinal tract. Based on a single highquality randomized trial supplemented by a prospective cohort study of the natural history of time to reach clinical stability for community-acquired pneumonia, we recommend treatment transition when patients meet explicit criteria for clinical stability derived from these studies. Our review has limitations. First, it does not address new or emerging therapies for community-acquired pneumonia or changing antimicrobial resistance patterns over time. Second, many of the studies included in our review were conducted prior to 2000. Nevertheless, the multiple more recent observational studies included in our review were not of higher quality than the older observational studies, and the most recent high-quality trials did not provide consistent evidence informing antibiotic selection. Third, relatively few studies informed each clinical question, and few of these studies were high-quality randomized controlled trials. Fourth, for initial antibiotic selection, we only evaluated antibiotics recommended by the IDSA/ATS guidelines and currently available in the United States, which limits the generalizability of our findings. Even though a randomized trial to assess the optimal timing of antibiotic initiation is unlikely to be conducted due to ethical concerns of delaying therapy in hospitalized patients, future prospective studies assessing a wide range of prognostic variables and detailed statistical adjustment for severity of illness could further inform this question. Given that macrolides and fluoroquinolones have been associated with increased antibiotic resistance and adverse drug effects, additional comparative effectiveness studies would better define the risks and benefits of such broad-spectrum antibiotic regimens. In addition, trials are needed that compare the transition from intravenous to oral antibiotic therapy guided by objective criteria for clinical stability or biomarkers of bacterial infection beginning on day 1 of hospitalization. Conclusions In adults hospitalized with community-acquired pneumonia, antibiotic therapy consisting of β-lactam plus macrolide combination therapy or fluoroquinolone monotherapy initiated within 4 to 8 hours of hospital arrival was associated with lower adjusted shortterm mortality, supported predominantly by low-quality observational studies. One randomized trial supports the use of objective clinical criteria to guide the transition from intravenous to oral antibiotic therapy. ARTICLE INFORMATION Author Contributions: Dr Lee had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Lee, Gellad, Fine. Acquisition, analysis, or interpretation of data: All authors. Drafting of the manuscript: Lee, Giesler. Critical revision of the manuscript for important intellectual content: Lee, Gellad, Fine. Statistical analysis: Lee, Giesler. Administrative, technical, or material support: Lee. Study supervision: Fine. Conflict of Interest Disclosures: The authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported. Submissions: We encourage authors to submit papers for consideration as a Review. Please contact Edward Livingston, MD, at Edward.livingston@jamanetwork.org or Mary McGrae McDermott, MD, at mdm608@northwestern.edu. REFERENCES 1. Pfuntner A, Wier LM, Steiner C; Agency for Healthcare Research and Quality. Most frequent conditions in US hospitals. http://www.hcup-us.ahrq.gov/reports/statbriefs/sb162.jsp. Accessed September 4, 2015. 2. Pfuntner A, Wier LM, Steiner C; Agency for Healthcare Research and Quality. Costs for hospital stays in the United States. http://www.hcup-us.ahrq.gov/reports/statbriefs/sb168-hospital-costs -United-States-2011.jsp. Accessed September 4, 2015. 3. US Centers for Disease Control and Prevention. FastStats: leading causes of death. http://www.cdc.gov/nchs/fastats/leading-causes-of-death.htm. Accessed September 4, 2015. 4. Metlay JP, Kapoor WN, Fine MJ. Does this patient have community-acquired pneumonia? diagnosing pneumonia by history and physical examination.jama. 1997;278(17):1440-1445. 5. American Thoracic Society; Infectious Diseases Society of America. Guidelines for the management of adults with hospital-acquired, ventilator-associated, and healthcare-associated pneumonia. Am J Respir Crit Care Med. 2005;171(4): 388-416. 6. Lim WS, van der Eerden MM, Laing R, et al. Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study. Thorax. 2003;58 (5):377-382. 7. Fine MJ, Auble TE, Yealy DM, et al. A prediction rule to identify low-risk patients with community-acquired pneumonia. N Engl J Med. 1997;336(4):243-250. 8. Lee JS, Nsa W, Hausmann LR, et al. Quality of care for elderly patients hospitalized for pneumonia in the United States, 2006 to 2010. JAMA Intern Med. 2014;174(11):1806-1814. 9. Jain S, Self WH, Wunderink RG, et al; CDC EPIC Study Team. Community-acquired pneumonia requiring hospitalization among US adults. N Engl J Med. 2015;373(5):415-427. 10. Jennings LC, Anderson TP, Beynon KA, et al. Incidence and characteristics of viral community-acquired pneumonia in adults. Thorax. 2008;63(1):42-48. 11. Lieberman D, Shimoni A, Shemer-Avni Y, Keren-Naos A, Shtainberg R, Lieberman D. Respiratory viruses in adults with community-acquired pneumonia. Chest. 2010;138 (4):811-816. 12. Mandell LA, Wunderink RG, Anzueto A, et al; Infectious Diseases Society of America; American Thoracic Society. Infectious Diseases Society of America/American Thoracic Society consensus guidelines on the management of community-acquired pneumonia in adults. Clin Infect Dis. 2007;44(suppl 2):S27-S72. 13. Meehan TP, Fine MJ, Krumholz HM, et al. Quality of care, process, and outcomes in elderly patients with pneumonia. JAMA. 1997;278(23): 2080-2084. 14. Dedier J, Singer DE, Chang Y, Moore M, Atlas SJ. Processes of care, illness severity, and outcomes in the management of community-acquired pneumonia at academic hospitals. Arch Intern Med. 2001;161(17):2099-2104. 15. Houck PM, Bratzler DW, Nsa W, Ma A, Bartlett JG. Timing of antibiotic administration and outcomes for Medicare patients hospitalized with community-acquired pneumonia. Arch Intern Med. 2004;164(6):637-644. 16. Wachter RM, Flanders SA, Fee C, Pronovost PJ. Public reporting of antibiotic timing in patients with pneumonia: lessons from a flawed performance measure. Ann Intern Med. 2008;149(1):29-32. jama.com (Reprinted) JAMA February 9, 2016 Volume 315, Number 6 601