|

 

Antibiotic treatment strategies for community-acquired pneumonia in adults

Antibiotic Treatment Strategies for Community-Acquired Pneumonia in Adults Douwe F. Postma, M.D., Cornelis H. van Werkhoven, M.D., Leontine J.R. van Elden, M.D., Ph.D., Steven F.T. Thijsen, M.D., Ph.D., Andy I.M. Hoepelman, M.D., Ph.D., Jan A.J.W. Kluytmans, M.D., Ph.D., Wim G. Boersma, M.D., Ph.D., Clara J. Compaijen, M.D., Eva van der Wall, M.D., Jan M. Prins, M.D., Ph.D., Jan J. Oosterheert, M.D., Ph.D., and Marc J.M. Bonten, M.D., Ph.D., for the CAP-START Study Group* From the Julius Center for Health Scienc- The choice of empirical antibiotic treatment for patients with clinically suspected es and Primary Care (D.F.P., C.H.W., community-acquired pneumonia (CAP) who are admitted to non–intensive care M.J.M.B.) and the Departments of Inter-nal Medicine and Infectious Diseases unit (ICU) hospital wards is complicated by the limited availability of evidence. We (D.F.P., A.I.M.H., J.J.O.) and Medical Mi- compared strategies of empirical treatment (allowing deviations for medical rea- crobiology (M.J.M.B.), University Medi- sons) with beta-lactam monotherapy, beta-lactam–macrolide combination therapy, cal Center Utrecht, and the Departments of Internal Medicine (D.F.P.), Pulmonol- or fluoroquinolone monotherapy.
ogy (L.J.R.E.), and Medical Microbiology METHODS
(S.F.T.T.), Diakonessenhuis Utrecht, Utrecht, the Department of Medical Mi- In a cluster-randomized, crossover trial with strategies rotated in 4-month periods, crobiology, Amphia Ziekenhuis Breda, we tested the noninferiority of the beta-lactam strategy to the beta-lactam–macro- Breda (J.A.J.W.K.), the Department of lide and fluoroquinolone strategies with respect to 90-day mortality, in an inten- Pulmonology, Medisch Centrum Alk-maar, Alkmaar (W.G.B.), the Department tion-to-treat analysis, using a noninferiority margin of 3 percentage points and a of Internal Medicine, Kennemer Gasthuis two-sided 90% confidence interval.
Haarlem, Haarlem (C.J.C.), the Depart-ment of Pulmonology, Spaarne Zieken- huis, Hoofddorp (E.W.), and the Depart- A total of 656 patients were included during the beta-lactam strategy periods, 739 dur- ment of Internal Medicine, Academic Medical Center Amsterdam, Amsterdam ing the beta-lactam–macrolide strategy periods, and 888 during the fluoroquinolone (J.M.P.) — all in the Netherlands. Ad- strategy periods, with rates of adherence to the strategy of 93.0%, 88.0%, and 92.7%, dress reprint requests to Dr. van Werk- respectively. The median age of the patients was 70 years. The crude 90-day mortality hoven at the University Medical Center Utrecht, Julius Center for Health Scienc- was 9.0% (59 patients), 11.1% (82 patients), and 8.8% (78 patients), respectively, during es and Primary Care, P.O. Box 85500, these strategy periods. In the intention-to-treat analysis, the risk of death was higher 3508 GA Utrecht, the Netherlands, or at by 1.9 percentage points (90% confidence interval [CI], −0.6 to 4.4) with the beta- c . h .vanwerkhoven@ umcutrecht . nl.
lactam–macrolide strategy than with the beta-lactam strategy and lower by 0.6 per- *A complete list of investigators in the centage points (90% CI, −2.8 to 1.9) with the fluoroquinolone strategy than with the Community-Acquired Pneumonia — Study on the Initial Treatment with An- beta-lactam strategy. These results indicated noninferiority of the beta-lactam strategy. tibiotics of Lower Respiratory Tract In- The median length of hospital stay was 6 days for all strategies, and the median time fections (CAP-START) Study Group is to starting oral treatment was 3 days (interquartile range, 0 to 4) with the fluoroqui- provided in the Supplementary Appen-dix, available at NEJM.org.
nolone strategy and 4 days (interquartile range, 3 to 5) with the other strategies.
Drs. Postma and van Werkhoven contrib- uted equally to this article.
Among patients with clinically suspected CAP admitted to non-ICU wards, a strategy N Engl J Med 2015;372:1312-23.
of preferred empirical treatment with beta-lactam monotherapy was noninferior to strategies with a beta-lactam–macrolide combination or fluoroquinolone monotherapy Copyright 2015 Massachusetts Medical Society. with regard to 90-day mortality. (Funded by the Netherlands Organization for Health Research and Development; CAP-START ClinicalTrials.gov number, NCT01660204.) n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. Antibiotic Treatment for Community-Acquired Pneumonia Community-acquired pneumonia (CAP) from February 2011 through August 2013 (see is a leading cause of hospitalization and the Supplementary Appendix, available with the death worldwide.1-3 Most guidelines rec- full text of this article at NEJM.org). The design ommend that antibiotic treatment be based on and rationale of the study have been described the severity of disease at presentation, assessed elsewhere,18 and the data are reported in accor- either on the basis of the level of care needed or dance with Consolidated Standards of Reporting on the basis of a prognostic risk score.4-6 For Trials (CONSORT) statements for cluster-random- patients with clinically suspected CAP who are ized and noninferiority studies.19,20 Additional admitted to a non–intensive-care-unit (ICU) ward, study details are provided in the study protocol guidelines recommend either combination ther- and statistical analysis plan, which are available apy with a beta-lactam plus a macrolide or plus at NEJM.org. The study protocol was approved ciprofloxacin or monotherapy with moxifloxacin by the ethics review board at the University Medical or levofloxacin for empirical treatment. These Center Utrecht (reference number 10/148), by the guidelines have increased the use of macrolides local institutional review boards, and by the anti- and fluoroquinolones, although these antibiotic biotic committee at each participating hospital.
classes have been associated with increased de- velopment of resistance.7,8 The evidence in support Eligibility and Recruitment of Patients
of these recommendations is limited.9-13 The rec- Patients 18 years of age or older with clinically ommendation to add a macrolide to a beta-lactam suspected CAP who required antibiotic treatment is based on observational studies, which are and hospitalization in a non-ICU ward were eli- prone to confounding by indication.14 Although gible for the study (Table 1). Patients with cystic fluoroquinolones have been evaluated in random- fibrosis were not eligible. Hospital G (see the ized, controlled trials, their superiority over beta- Supplementary Appendix) included only patients lactam monotherapy has not been shown.15,16 with a CURB-65 score greater than 2 (the CURB- Moreover, the results of randomized, controlled 65 score is calculated by assigning 1 point each trials may be affected by in-hospital antibiotic for confusion, uremia [blood urea nitrogen ≥20 mg exposure that occurs before randomization17,18 per deciliter], high respiratory rate [≥30 breaths per and often have restrictive inclusion criteria, which minute], low systolic blood pressure [<90 mm Hg] limit the generalizability of their results to daily or diastolic blood pressure [≤60 mm Hg], and an age of 65 years or older, with a higher score in- We therefore assessed whether a strategy of dicating a higher risk of death within 30 days).21 preferred empirical treatment with beta-lactam We used on-site training of research nurses monotherapy is noninferior to either preferred throughout the study to ensure the standardiza- beta-lactam–macrolide combination therapy or tion of case definitions.
preferred fluoroquinolone monotherapy, with re- Emergency department registries were screened gard to 90-day all-cause mortality among adults daily for eligible patients by research nurses or with clinically suspected CAP who are admitted physicians. Obtaining informed consent before to non-ICU wards. These strategies allowed for intervention was deemed unnecessary, because deviation from the assigned antibiotic therapy patients did not undergo randomization individu- for medical reasons, so as not to compromise care. ally, and all the antibiotics we studied are used We performed a pragmatic, cluster-randomized, in current practice.22 Written informed consent crossover trial to overcome confounding by indi- obtained within 72 hours after admission was re- cation and the effects of prerandomization anti- quired for data collection.
biotic therapy.
During consecutive periods of 4 months, beta- lactam monotherapy, beta-lactam with a macro- Study Design and Oversight
lide, or fluoroquinolone monotherapy was used as The Community-Acquired Pneumonia — Study the preferred empirical treatment for eligible on the Initial Treatment with Antibiotics of Lower patients. The order of strategies was randomized Respiratory Tract Infections (CAP-START) was separately in each hospital, without washout peri- performed in seven hospitals in the Netherlands, ods. Patients were treated and assessed in ac- n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. Table 1. Definitions.
Community-acquired pneumonia (CAP) (working diagnosis): The presence of at least two of the diagnostic clinical cri- teria and in-hospital treatment with antibiotics for clinically suspected CAP as documented by the treating physi- cian. Patients with two or more criteria and an obvious nonrespiratory source of infection were not considered to have a working diagnosis of CAP, nor were patients who had recently been hospitalized (for >48 hours in the previ- ous 2 weeks) or who resided in long-term care facilities.
Radiologically confirmed CAP: A working diagnosis of CAP plus the presence of a new or increased infiltrate on chest radiography or computed tomography (CT) and at least two other clinical criteria.
Diagnostic clinical criteria
Production of purulent sputum or a change in the character of sputum Temperature >38°C or <36.1°C Auscultatory findings consistent with pneumonia, including rales, evidence of pulmonary consolidation (dullness on percussion, bronchial breath sounds, or egophony), or both Leukocytosis (>10×109 white cells per liter or >15% bands) C-reactive protein level more than 3 times the upper limit of the normal range Dyspnea, tachypnea, or hypoxemia New or increased infiltrate on chest radiography or CT scan Beta-lactam strategy: Preferred empirical treatment with amoxicillin, amoxicillin plus clavulanate, or a third-generation cephalosporin. Penicillin was not allowed as empirical beta-lactam monotherapy.
Beta-lactam–macrolide strategy: Preferred empirical treatment with penicillin, amoxicillin, amoxicillin plus clavulanate, or a third-generation cephalosporin in combination with azithromycin, erythromycin, or clarithromycin Fluoroquinolone strategy: Preferred empirical treatment with moxifloxacin or levofloxacin * Strategies were based on the recommendations in the Dutch guideline on treatment of CAP that was available before the start of the study.23 cordance with the strategy that was applicable on treatment with the assigned antibiotic, irrespec- the admission date. Clinicians were repeatedly tive of subsequent switches of antibiotic treat- informed of the current antibiotic strategy by lo- ment to a nonassigned antibiotic.
cal investigators and with the use of newsletters and presentations.
The antibiotics allowed during each treat- Computer-generated randomization was per- ment strategy period (Table 1) were based on the formed in blocks of six, each containing a se- 2005 Dutch guideline.23 Physicians were encour- quence of the three antibiotic strategies. Hospi- aged to apply the assigned treatment strategies tals were assigned to their sequence after for the full treatment of patients with suspected approval of the study by the hospital antibiotic CAP, unless there were medical reasons not to, committee. Two hospitals that had closely col- such as adverse events or de-escalation of antibi- laborating medical staff were randomized as otic treatment (e.g., because of a switch to tar- one cluster. All the hospitals planned to partici- geted treatment when a causative pathogen had pate until the calculated sample size was met or been identified). Adherence to the strategy was for a maximum of 2 years (Fig. S1 in the Supple- defined as treatment in accordance with the as- mentary Appendix).
signed strategy or deviation from the strategy for medical reasons (i.e., motivated deviation), Outcomes
irrespective of subsequent switches of antibiotic The primary outcome was all-cause mortality treatment to a nonassigned antibiotic. Adher- within 90 days after admission. The secondary ence to the antibiotic was defined as initial outcomes were the time to starting oral treat- n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. Antibiotic Treatment for Community-Acquired Pneumonia ment, length of hospital stay, and occurrence of rate, heart rate, and confusion at admission; the minor or major complications during the hospi- values for these variables were assumed to be tal stay. All outcomes were measured at the in- normal when data were missing. The analyses dividual patient level.
were performed with the use of R software, version 3.0.2 (R Project for Statistical Computing).28 Data Collection
Data on clinical presentation, laboratory and
microbiologic test results, the antibiotic agents used, complications, and clinical outcome were Enrollment
retrieved from medical records. Nonroutine data A total of 3325 patients were eligible for inclusion were recorded by research nurses directly after the in the study, and 2283 (69%) gave consent. The patient's inclusion. When the reasons for devia- median age of the patients was 70 years (inter- tions from the assigned empirical treatment were quartile range, 59 to 79). Among the patients who not clear in the medical chart, research nurses re- were not included, the median age was 74 years quested information from responsible physicians. (interquartile range, 63 to 83) during the beta- The 90-day mortality was determined from the lactam strategy periods, 74 years (interquartile patient record database of each participating range, 61 to 82) during the beta-lactam–macro- hospital or from the municipal personal records lide strategy periods, and 74 years (interquartile database (see the Supplementary Appendix).
range, 61 to 83) during the fluoroquinolone strat- egy periods, and the reasons for noninclusion were similar across strategies (Fig. 1). The base- Details about the calculation of sample size are line characteristics of included patients were provided in the Supplementary Appendix. Analy- similar among strategy periods, and blood and ses were performed in accordance with the in- sputum cultures and urinary antigen testing for tention-to-treat principle, with adjustment for clus- Streptococcus pneumoniae and Legionel a pneumophila tering. Differences among the groups in 90-day were performed with similar frequency (Ta- mortality were assessed with the use of a mixed- ble 2). The microbial causes of CAP were similar effects logistic-regression analysis, including hos- in the three treatment groups. S. pneumoniae was pitals as a fixed effect and each strategy period the pathogen detected most frequently (in 15.9% per hospital as a random intercept.24 We estimated of patients), followed by Haemophilus influenzae (in absolute risk differences among strategies by aver- 6.8%); atypical pathogens were found in 2.1% of aging the computed individual risks for each treat- the patients (Table S1 in the Supplementary Ap- ment group, and confidence intervals were calcu- pendix). Resistance to the initiated antibiotic lated with the use of 2000 bootstrapping samples.25 treatment was highest with the beta-lactam strat- Noninferiority was assessed in a one-sided test at egy (Table S2 in the Supplementary Appendix).
a significance level of 0.05 with the use of two- Six hospitals completed 6 randomized strategy sided 90% confidence intervals.
periods each; enrollment was discontinued in one Differences in the length of hospital stay and hospital after 4.5 periods, when the intended the time to starting oral administration of anti- number of patients per treatment group had biotics were tested with mixed-effects Cox pro- been reached. Changeovers from one treatment portional-hazards models.26 The frequencies of strategy period to the next occurred as planned major and minor complications were compared except in one hospital: because of unforeseen by means of mixed-effects multinomial regression. fluoroquinolone supply problems, 4 weeks of the Post hoc analyses of the strategy-adherent and first fluoroquinolone period were exchanged with antibiotic-adherent populations were performed the subsequent beta-lactam–macrolide period for all outcomes. We performed sensitivity analy- (Fig. S1 in the Supplementary Appendix).
ses that included only patients with radiologi- cally confirmed CAP (Table 1) and that assessed Strategy Adherence and Antibiotic Use
30-day mortality instead of 90-day mortality, and Rates of adherence to the strategies and to anti- we calculated two-sided 95% confidence intervals. biotic treatment are shown in Figure 1. Antibiotic Missing data were imputed by multiple imputa- use during each strategy period is summarized in tion,27 with the exception of data on respiratory Table S3 in the Supplementary Appendix, and n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. 3325 Patients were eligible 993 Were assigned to receive 1055 Were assigned to receive 1277 Were assigned to receive 337 (34.0%) Were excluded 316 (30.0%) Were excluded 389 (30.5%) Were excluded 134 (13.5%) Declined to parti- 78 (7.4%) Declined to parti- 133 (10.4%) Declined to parti- 96 (9.7%) Were discharged 123 (11.7%) Were discharged 98 (7.7%) Were discharged before consent was given before consent was given before consent was given 88 (8.9%) Were unable to 80 (7.6%) Were unable to 130 (10.2%) Were unable to 19 (1.9%) Had unknown 35 (3.3%) Had unknown 28 (2.2%) Had unknown 656 Were included in study 739 Were included in study 888 Were included in study 610 (93.0%) Were in the strategy-adher- 650 (88.0%) Were in the strategy-adher- 823 (92.7%) Were in the strategy-adher- 468 (71.3%) Were in the antibiotic- 538 (72.8%) Were in the antibiotic- 712 (80.2%) Were in the antibiotic- adherent population adherent population adherent population 142 (21.6%) Had motivated deviation 112 (15.2%) Had motivated deviation 111 (12.5%) Had motivated deviation 46 (7.0%) Were nonadherent 89 (12.0%) Were nonadherent 65 (7.3%) Were nonadherent 2 (0.3%) Had missing data 1 (0.1%) Had missing data 1 (0.1%) Had missing data 59 (9.0%) Were in the intention-to- 82 (11.1%) Were in the intention-to- 78 (8.8%) Were in the intention-to- 52 (8.5%) Were in the strategy- 68 (10.5%) Were in the strategy- 70 (8.5%) Were in the strategy- adherent population adherent population adherent population 42 (9.0%) Were in the antibiotic- 55 (10.2%) Were in the antibiotic- 53 (7.4%) Were in the antibiotic- adherent population adherent population adherent population Figure 1. Inclusion of Patients, Rates of Adherence, and Mortality.
The strategy-adherent population was the population that underwent treatment in accordance with the assigned strategy or had devia-
tion from the strategy for medical reasons (i.e., motivated deviation), irrespective of subsequent switches of antibiotic treatment to a
nonassigned antibiotic; the antibiotic-adherent population was the population that underwent initial treatment with the assigned antibi-
otic, irrespective of subsequent switches of antibiotic treatment to a nonassigned antibiotic.
antibiotic adherence is summarized in Figure S3 Deviations were made for 565 patients (24.8%); in the Supplementary Appendix. The number of a total of 200 of these deviations had no docu- patients empirically treated with antibiotic cov- mented medical reason. The most frequent medi- erage for atypical pathogens (i.e., macrolides, fluo- cal reasons for deviation from the beta-lactam roquinolones, and doxycycline) during the beta- strategy were the perceived need to cover atypical lactam strategy periods was 67% less than the pathogens (53 patients, 8.1%), a contraindication number treated with atypical coverage during the to beta-lactams (21 patients, 3.2%), and a recent beta-lactam–macrolide strategy periods and 69% start of treatment with another antibiotic class less than the number during the fluoroquino- or a lack of response to preadmission treatment lone strategy periods, and the cumulative num- with beta-lactams (27 patients, 4.1%) (Table S4 ber of days with atypical coverage was 57% and in the Supplementary Appendix). Among pa- 62% less, respectively.
tients receiving the assigned therapy, switches to n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. Antibiotic Treatment for Community-Acquired Pneumonia Table 2. Baseline Characteristics of Patients in the Intention-to-Treat Population.*
Antibiotic Treatment Strategy
Beta-Lactam–Macrolide Median age (interquartile range) — yr Male sex — no. (%) Median duration of symptoms (interquartile Received antibiotics before admission — no./total no. (%) Current smoker — no./total no. (%) Past smoker — no./total no. (%) Received influenza vaccination — no./ Received pneumococcal vaccination — no./total no. (%) Dependency in ADL — no./total no. (%)† Had one or more hospital stays in the previous 271/653 (41.5) year — no./total no. (%) Had coexisting condition — no. (%) Cardiovascular disease Other chronic pulmonary disease Diabetes mellitus HIV/AIDS — no. (%) Chronic renal failure or nephrotic Receiving immunosuppressive therapy — no. (%) Underwent organ or bone marrow transplantation — no. (%) Median CURB-65 score (interquartile range)§‖ Had radiologically confirmed CAP — no. (%) Blood culture obtained — no. (%) Sputum culture obtained — no. (%) PUAT performed — no. (%) LUAT performed — no. (%) * Plus–minus values are means ±SD. ADL denotes activities of daily living, COPD chronic obstructive pulmonary disease, LUAT legionella urinary antigen test, PCV13 13-valent pneumococcal conjugate vaccine (received in the Community Acquired Pneumonia Immunization Trial in Adults [CAPITA]), PPSV23 23-valent pneumococcal polysaccharide vaccine, PSI Pneumonia Severity Index, and PUAT pneumococcal urinary antigen test.
† This category includes patients who were not able to perform ADL autonomously.
‡ Active cancer was defined as a solid or hematologic cancer treated with radiotherapy or chemotherapy within the previ- ous 5 years.
§ When data were missing, values were assumed to be normal. A total of 6.3% of data points used to calculate the PSI score had missing values, and 11.3% of data points used to calculate the CURB-65 score had missing values.
¶ The PSI score uses 20 clinical measures to predict risk of death within 30 days, with results ranging from 0.1% (in pa- tients with a score of 0–50) to 27.0% (in patients with a score >131).
‖ The CURB-65 score is calculated by assigning 1 point each for confusion, uremia (blood urea nitrogen ≥20 mg per deci- liter), high respiratory rate (≥30 breaths per minute), low systolic blood pressure (<90 mm Hg) or diastolic blood pres- sure (≤60 mm Hg), and an age of 65 years or older, with a higher score indicating a higher risk of death within 30 days.
n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. A Intention-to-Treat Analysis
B Intention-to-Treat Analysis (radiologically confirmed CAP)
−0.04 ∆ −0.02 −0.04 ∆ −0.02 Other Strategy Better
Other Strategy Better
C Strategy-Adherent Analysis
D Strategy-Adherent Analysis (radiologically confirmed CAP)
−0.04 ∆ −0.02 −0.04 ∆ −0.02 Other Strategy Better
Other Strategy Better
E Antibiotic-Adherent Analysis
F Antibiotic-Adherent Analysis (radiologically confirmed CAP)
−0.04 ∆ −0.02 −0.04 ∆ −0.02 Other Strategy Better
Other Strategy Better
other antibiotic classes because of perceived in- Primary Outcome
sufficient clinical recovery were made for 41 All-cause mortality at 90 days could not be as- patients (8.8%) during the beta-lactam strategy sessed for four patients; these patients were in- periods, for 33 patients (6.1%) during the beta- cluded only in secondary analyses (Fig. 1). The lactam–macrolide strategy periods, and for 26 pa- absolute difference in the adjusted risk of death tients (3.7%) during the fluoroquinolone strategy between the beta-lactam strategy and the beta- periods. Other reasons for switching antibiotic lactam–macrolide strategy was 1.9 percentage classes are provided in Table S5 in the Supplemen- points (90% confidence interval [CI], −0.6 to tary Appendix.
4.4) in favor of the beta-lactam strategy, and the n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. Antibiotic Treatment for Community-Acquired Pneumonia val for the comparison of the beta-lactam strat- Figure 2 (facing page). Noninferiority Plots.
egy with the fluoroquinolone strategy crossed the The noninferiority plots show crude and adjusted ab-solute risk differences for 90-day mortality associated noninferiority margin (Fig. 2, and Table S6, S7, with the beta-lactam–macrolide combination and flu- and S8 in the Supplementary Appendix).
oroquinolone monotherapy strategies, as compared with the beta-lactam monotherapy strategy, in analysis of the intention-to-treat population, the strategy-ad- The median length of hospital stay was 6 days for herent population, and the antibiotic-adherent popula-tion, as well as for the sensitivity analysis including all strategies, but the upper quartile was higher only patients with radiologically confirmed communi- during the beta-lactam–macrolide strategy peri- ty-acquired pneumonia (CAP). To allow for one-sided ods (Table 3). The median duration of intrave- testing of noninferiority, 90% confidence intervals nous treatment was 3 days during the fluoroqui- were calculated (shown in black); 95% confidence in- nolone strategy periods and 4 days during the tervals are also provided (shown in red). Confidence intervals within the gray-shaded area are noninferior. other strategy periods (Table 3). The proportions The crude analyses take into account cluster-period ef- of patients whose treatment started with oral fects and center effects. The adjusted analyses are ad- antibiotics were 27% during the fluoroquinolone ditionally corrected for Pneumonia Severity Index strategy periods, as compared with 13% and 10% score (a score that uses 20 clinical measures, includ- during the beta-lactam and beta-lactam–macro- ing age and sex, to predict the risk of death within 30 days, with results ranging from 0.1% [in patients with lide strategy periods, respectively. There were no a score of 0–50] to 27.0% [in patients with a score significant differences among the three strate- >131]); smoking status; presence of chronic pulmo- gies in the incidence of major or minor compli- nary diseases, chronic cardiovascular diseases, diabe- cations (Table 3).
tes mellitus, or immunosuppression; previous treat-ment with antibiotics; and number of hospitalizations during the previous year. The analysis of the antibiotic- adherent population is further adjusted for duration of symptoms; dependency in activities of daily living; In this pragmatic, cluster-randomized, crossover presence or absence of hematologic cancer, nonhema- trial, a strategy of preferred empirical treatment tologic cancer, or immunosuppression; C-reactive pro- with beta-lactam monotherapy was noninferior tein level; whole-blood leukocyte count; and tempera-ture at hospital admission. The noninferiority margin to strategies of treatment with beta-lactam–mac- is −3 percentage points (shown as Δ). The intracluster rolide combination therapy and with fluoroqui- correlation for cluster-period effects in the primary nolone monotherapy among patients with sus- analysis was 4.5×10−7. Exact numbers are provided in pected CAP who were admitted to non-ICU wards. Table S6 in the Supplementary Appendix, and survival Moreover, there were no clinically relevant dif- curves are shown in Figure S4 in the Supplementary Appendix. BLM denotes beta-lactam–macrolide combi- ferences among treatment strategies in the length nation therapy and FQL fluoroquinolone monotherapy.
of hospital stay or in reported complications. The median time to starting oral treatment was shorter with the fluoroquinolone strategy, main- absolute difference between the beta-lactam strat- ly because more patients during those strategy egy and the fluoroquinolone strategy was −0.6 periods started with oral empirical treatment at percentage points (90% CI, −2.8 to 1.9) in favor admission, but this did not result in a decreased of the fluoroquinolone strategy. These confi- length of hospital stay.
dence intervals do not include the prespecified Our approach differs from those of previous margin of a 3–percentage-point higher 90-day studies in four aspects. First, this study addressed mortality, thus demonstrating the noninferiority treatment strategies, rather than individual anti- of the beta-lactam strategy to the beta-lactam– biotics, in the treatment of patients hospitalized macrolide and fluoroquinolone strategies (Fig. 2). with a clinical suspicion of CAP. To reflect daily In the strategy-adherent and antibiotic-adher- medical practice, we allowed for deviations from ent populations, the absolute adjusted risk dif- the assigned therapy for medical reasons. To ferences were similar to those in the intention- minimize confounding, all the patients were to-treat population. Similar estimates were obtained included in the intention-to-treat analysis. Al- in sensitivity analyses of patients with radiologi- though deviations and switches reduced the dif- cally confirmed CAP and in analyses of 30-day ferences among treatment strategies, empirical mortality. The two-sided 95% confidence inter- atypical coverage was reduced by 67% during the n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. Table 3. Effects of Antibiotic Treatment Strategies on Secondary Outcomes.*
Antibiotic Treatment Strategy
Beta-Lactam Beta-Lactam–Macrolide Median length of stay (IQR) — days† Rate ratio for discharge alive (95% CI)‡ 0.86 (0.77–0.96) 1.03 (0.93–1.15) 0.87 (0.78–0.97) 1.04 (0.94–1.16) 0.86 (0.77–0.96) 1.03 (0.93–1.15) 0.86 (0.77–0.97) 1.04 (0.93–1.16) 0.84 (0.74–0.96) 1.04 (0.92–1.17) 0.84 (0.74–0.95) 1.03 (0.92–1.17) Time to starting oral treatment§ Receipt of oral antibiotics as initial in-hospital therapy — no. (%) Median time receiving IV antibiotic treatment Rate ratio for starting oral treatment (95% CI)¶ 0.95 (0.84–1.08) 1.28 (1.13–1.44) 0.97 (0.86–1.09) 1.29 (1.15–1.46) 0.94 (0.82–1.07) 1.30 (1.15–1.48) 0.94 (0.83–1.08) 1.33 (1.17–1.51) 0.93 (0.78–1.10) 1.47 (1.24–1.73) 0.93 (0.79–1.11) 1.52 (1.28–1.80) Minor — no. (%) Major — no. (%) Unknown — no. (%) Odds ratio (95% CI)** 1.06 (0.76–1.48) 1.02 (0.73–1.41) 1.06 (0.74–1.52) 1.03 (0.73–1.46) 1.20 (0.82–1.77) 1.03 (0.71–1.51) * Crude analyses take into account cluster-period effects and center effects. Adjusted analyses are additionally corrected for PSI score (including age and sex); smoking status; presence of chronic pulmonary disease, chronic cardiovascular disease, diabetes mellitus, or immunosuppression; previous receipt of antibiotics; and number of hospitalizations in the previous year. IQR denotes interquartile range, and IV intravenous.
† The length of stay was unknown for 5 patients in the beta-lactam strategy group (0.8%), 2 patients in the beta-lactam–macrolide strategy group (0.3%), and 5 patients in the fluoroquinolone strategy group (0.6%), who were transferred to other hospitals.
‡ Rate ratios are from a Cox proportional-hazards model predicting the day of discharge as the event of interest. A rate ra- tio below 1 indicates a longer length of stay. The survival curve is shown in Figure S5 in the Supplementary Appendix.
§ The duration of intravenous treatment was unknown for 1 patient in the fluoroquinolone strategy group (0.1%) who was transferred to another hospital while receiving intravenous treatment.
¶ Rate ratios are from a Cox proportional-hazards model predicting the end of intravenous treatment or the start of oral treatment as the event of interest. A rate ratio below 1 indicates a longer duration of intravenous treatment. The sur- vival curve is shown in Figure S6 in the Supplementary Appendix.
‖ Major complications include in-hospital death, respiratory insufficiency, ICU admission, organ failure, and septic shock. A detailed description of complications is provided in Table S9 in the Supplementary Appendix.
** Odds ratios (all crude analyses) are from a mixed-effects ordinal logistic-regression model with no, minor, or major complications as the dependent variable.
n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. Antibiotic Treatment for Community-Acquired Pneumonia beta-lactam strategy periods as compared with Finally, the primary end point was 90-day all- the beta-lactam–macrolide strategy periods and cause mortality, because CAP is associated with by 69% during the beta-lactam strategy periods high long-term mortality and this is a patient- as compared with the fluoroquinolone strategy relevant outcome that is not susceptible to obser- periods. The number of in-hospital days with vation bias.17,29,30 An unplanned sensitivity analysis atypical coverage was also reduced during the with 30-day mortality as the outcome yielded beta-lactam strategy periods, by 57% and 62%, similar results. Among the secondary outcomes, respectively. In the post hoc analysis of the strat- complications, which were extracted from the egy-adherent and antibiotic-adherent populations, medical records, might have been misclassified the beta-lactam strategy remained noninferior to and subject to observation bias.
the beta-lactam–macrolide strategy. In the crude The noninferiority of the beta-lactam strategy analysis of the antibiotic-adherent population, to the beta-lactam–macrolide strategy was appar- the lower limit of the confidence interval crossed ent in all analyses. These findings, together with −3 percentage points for the comparison be- the slightly longer length of hospital stay with the tween beta-lactam and fluoroquinolone mono- latter strategy, reported associations with the de- therapy; however, after adjustment for confound- velopment of resistance,7,8 and possible increased ers, the lower limit of the confidence interval fell risks of cardiac events,31,32 indicate that the addi- within the defined margins of noninferiority.
tion of macrolides for empirical treatment of CAP Second, we used a cluster-randomized design should be reconsidered. In a recent randomized, that allowed for an immediate start of the as- controlled trial, the noninferiority of beta-lactam signed empirical treatment strategy. The cross- monotherapy to beta-lactam–macrolide combina- over component increased the efficiency of the tion therapy with respect to clinical stability at trial by allowing comparisons of the effect of the day 7 could not be shown, although superiority strategies within each cluster and ensuring that of the beta-lactam–macrolide combination ther- all hospitals used all three strategies, a design apy was not shown, either. Moreover, 30-day and that minimized the possibility of confounding. 90-day all-cause mortality and length of hospital Despite the risk of selection bias that is inherent stay were similar with the two therapies.33 Dif- to cluster-randomized studies, the baseline char- ferences between that study and the current study acteristics of the patients were similar among include the strict criteria for eligibility and for strategies, and statistical adjustment for potential switching therapy in cases of clinical deteriora- confounders changed the findings only mini- tion in that study.
mally. Differential inclusion of patients across Some aspects of our study require explana- treatment groups was unlikely, given the similar tion. In the noninferiority design, we used one- age patterns for nonincluded patients and simi- sided testing with an alpha significance level of lar enrollment rates. We were not allowed to 0.05. With 95% confidence intervals — that is, collect data on other characteristics of the pa- an alpha level of 0.025 — the noninferiority of tients who were not included. The pathogens beta-lactams to fluoroquinolones was not shown found were similar among strategy groups, but (Fig. 2); however, there was no clear trend to- the resistance of pathogens to the actual treat- ward superiority for fluoroquinolones in any of ment was highest during the beta-lactam strat- the other adjusted analyses.
egy periods. This did not appear to lead to a Differences in the numbers of eligible pa- worse outcome, possibly because not all were tients per strategy resulted from cluster random- proven causative pathogens and because of anti- ization. The beta-lactam and fluoroquinolone biotic switches.
strategies were assigned more frequently during Third, all patients for whom the antibiotic the 2011–2012 and 2012–2013 winter seasons, strategy might have been used in daily practice respectively, and more patients were hospital- were eligible for enrollment, which increases the ized during 2012–2013 winter months. However, generalizability of the results. Although this the proportions of patients included were simi- could increase the heterogeneity of the popula- lar throughout the seasons and among strategies tion and the potential for bias toward noninferi- (Fig. 1, and Fig. S2 in the Supplementary Ap- ority, the effect estimates were similar in the pendix). Although a low incidence of atypical sensitivity analysis that included only patients infections during the 2011–2012 winter season with radiologically confirmed CAP.
could have favored the beta-lactam strategy, na- n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. tional surveillance data showed a higher inci- available. All 5 patients had a good clinical out- dence of Mycoplasma pneumoniae infections, most- come. Higher incidences of community-acquired ly CAP, during that period,34 for which the Pseudomonas aeruginosa or methicillin-resistant beta-lactam strategy might have been less effec- Staphylococcus aureus infections would require the tive. The outbreak of Q fever in the Netherlands adaptation of all three treatment strategies.
ended before the start of the current study,35 and In conclusion, among patients with suspected the distribution of pathogens was similar to CAP who were admitted to non-ICU wards, we those in other studies that have relied on routine found that a strategy of preferred empirical treat- ment with beta-lactam monotherapy that al- Regional differences in microbial causes could lowed for deviations for medical reasons was reduce the generalizability of our findings. How- noninferior to strategies with beta-lactam–mac- ever, resistance of S. pneumoniae to penicillin,39 rolide combination therapy or fluoroquinolone which rarely occurs in the Netherlands, is un- monotherapy in terms of 90-day all-cause mor- likely to influence the outcome in patients with tality. In addition, beta-lactam monotherapy was pneumonia treated with beta-lactam antibiot- not associated with a longer length of hospital ics.40 The prevalence of S. pneumoniae resistance stay or a higher incidence of complications.
to macrolides was 4.2% in the Netherlands in 2011.39 The incidence of legionella in this study Supported by a grant (171202002) from the Netherlands Or- was less than 1%. A higher incidence could in- ganization for Health Research and Development.
fluence the effectiveness of empirical treatment No potential conflict of interest relevant to this article was with beta-lactam monotherapy, which stresses Disclosure forms provided by the authors are available with the importance of rapid testing in patients with the full text of this article at NEJM.org.
risk factors for Legionnaires' disease. In the cur- We thank Richard Wunderink (Northwestern University Fein- berg School of Medicine, Chicago), Robert Weinstein and David rent study, rapid urinary antigen testing for le- Schwartz (Stroger Hospital and Rush University Medical Center, gionella was performed in 492 patients (75%) Chicago), and Arno Hoes (University Medical Center, Utrecht, during the beta-lactam strategy periods; 5 of the the Netherlands) for helpful suggestions regarding an earlier version of the manuscript, and the research nurses in the par- patients (1%) tested positive, 2 of whom received ticipating hospitals for their efforts in patient recruitment and ciprofloxacin empirically because of a high clini- data collection.
cal suspicion. For the other 3 patients, antibiotic This article is dedicated to the memory of Reinier Veenhoven, an investigator for this study in Spaarne Hospital and Kennemer therapy was adjusted after test results became Gasthuis, who died in October 2013.
References
1. Thomas CP, Ryan M, Chapman JD, et Party on Antibiotic Policy and Dutch As-
therapy for atypical pathogens improve al. Incidence and cost of pneumonia in sociation of Chest Physicians) guidelines outcomes for patients with CAP? Infect medicare beneficiaries. Chest 2012; 142: on the management of community- Dis Clin North Am 2013; 27: 99-114.
acquired pneumonia in adults. Neth J Med 11. Morris AM, Ovens H. Community-
2. Welte T, Torres A, Nathwani D. Clini-
2012; 70: 90-101.
acquired pneumonia. N Engl J Med 2014; cal and economic burden of community- 7. Fuller JD, Low DE. A review of Strep- 370: 1862.
acquired pneumonia among adults in Eu- tococcus pneumoniae infection treatment 12. Spellberg B. Community-acquired
rope. Thorax 2012; 67: 71-9.
failures associated with fluoroquinolone pneumonia. N Engl J Med 2014; 370: 1861- 3. Yu H, Rubin J, Dunning S, Li S, Sato resistance. Clin Infect Dis 2005; 41: 118-21.
R. Clinical and economic burden of com- 8. Malhotra-Kumar S, Lammens C, 13. Wunderink RG, Waterer GW. Com-
munity-acquired pneumonia in the Medi- Coenen S, Van Herck K, Goossens H. Ef- munity-acquired pneumonia. N Engl J care fee-for-service population. J Am fect of azithromycin and clarithromycin Med 2014; 370: 1863.
Geriatr Soc 2012; 60: 2137-43.
therapy on pharyngeal carriage of macro- 14. Grobbee DE, Hoes AW. Confounding
4. Lim WS, Baudouin SV, George RC, et lide-resistant streptococci in healthy vol-
and indication for treatment in evaluation al. BTS guidelines for the management of unteers: a randomised, double-blind, pla- of drug treatment for hypertension. BMJ community acquired pneumonia in adults: cebo-controlled study. Lancet 2007; 369: 1997; 315: 1151-4.
update 2009. Thorax 2009; 64: Suppl 3: iii1- 482-90.
15. Eliakim-Raz N, Robenshtok E, Shefet
9. Oosterheert JJ, Bonten MJ, Hak E, D, et al. Empiric antibiotic coverage of
5. Mandell LA, Wunderink RG, Anzueto Schneider MM, Hoepelman IM. How atypical pathogens for community-
A, et al. Infectious Diseases Society of good is the evidence for the recommend- acquired pneumonia in hospitalized America/American Thoracic Society con- ed empirical antimicrobial treatment of adults. Cochrane Database Syst Rev 2012; sensus guidelines on the management of patients hospitalized because of commu- 9: CD004418.
community-acquired pneumonia in adults. nity-acquired pneumonia? A systematic 16. Mills GD, Oehley MR, Arrol B. Effec-
Clin Infect Dis 2007; 44: Suppl 2: S27-S72.
review. J Antimicrob Chemother 2003; 52: tiveness of beta lactam antibiotics com- 6. Wiersinga WJ, Bonten MJ, Boersma 555-63.
pared with antibiotics active against atyp- WG, et al. SWAB/NVALT (Dutch Working 10. File TM Jr, Marrie TJ. Does empiric ical pathogens in non-severe community
n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved. Antibiotic Treatment for Community-Acquired Pneumonia acquired pneumonia: meta-analysis. BMJ models including logistic regression. Health 34. Dutch National Institute for Public
2005; 330: 456.
Serv Res 2009; 44: 288-302.
Health and the Environment. Virologische 17. Powers JH. Reassessing the design, 26. Therneau TM, Grambsch PM, Pan-
weekstaten, 2014 (http://www .rivm .nl/ conduct, and analysis of clinical trials of kratz VS. Penalized survival models and Onderwerpen/ V/ Virologische_weekstaten/ therapy for community-acquired pneu- frailty. J Comput Graph Statist 2003; 12: Rapportages/ Open_rapportages monia. Clin Infect Dis 2008; 46: 1152-6.
_virologische_weekstaten/ Virologische 18. van Werkhoven CH, Postma DF, 27. van Buuren S, Groothuis-Oudshoorn _uitslagen_per_week_sinds_2006_grafiek).
Oosterheert JJ, Bonten MJM. Antibiotic K. Mice: multivariate imputation by 35. van der Hoek W, Morroy G, Renders
treatment of moderate-severe communi- chained equations in R. J Stat Softw 2011; NH, et al. Epidemic Q fever in humans in ty-acquired pneumonia: design and ra- 45: 1-67. (http://www .jstatsoft .org/ v45/ i03/ the Netherlands. Adv Exp Med Biol 2012; tionale of a multicentre cluster-ran- paper).
984: 329-64.
domised cross-over trial. Neth J Med 28. R Foundation. The R project for sta- 36. Ewig S, Hecker H, Suttorp N, Marre
2014; 72: 170-8.
tistical computing (http://www .r-project R, Welte T. Moxifloxacin monotherapy 19. Campbell MK, Piaggio G, Elbourne .org).
versus β-lactam mono- or combination DR, Altman DG. Consort 2010 statement: 29. Spellberg B, Talbot GH, Brass EP, therapy in hospitalized patients with com-
extension to cluster randomised trials. Bradley JS, Boucher HW, Gilbert DN. Posi- munity-acquired pneumonia. J Infect 2011; BMJ 2012; 345: e5661.
tion paper: recommended design features 62: 218-25.
20. Piaggio G, Elbourne DR, Pocock SJ, of future clinical trials of antibacterial 37. Huijts SM, Pride MW, Vos JM, et al.
Evans SJ, Altman DG. Reporting of non- agents for community-acquired pneumo- Diagnostic accuracy of a serotype-specific inferiority and equivalence randomized nia. Clin Infect Dis 2008; 47: Suppl 3: S249- antigen test in community-acquired pneu- trials: extension of the CONSORT 2010 S265.
monia. Eur Respir J 2013; 42: 1283-90.
statement. JAMA 2012; 308: 2594-604.
30. Bruns AH, Oosterheert JJ, Cucciolillo 38. Tessmer A, Welte T, Martus P, Schnoor
21. Lim WS, van der Eerden MM, Laing R, MC, et al. Cause-specific long-term mor-
M, Marre R, Suttorp N. Impact of intrave- et al. Defining community acquired pneu- tality rates in patients recovered from nous beta-lactam/macrolide versus beta- monia severity on presentation to hospi- community-acquired pneumonia as com- lactam monotherapy on mortality in hos- tal: an international derivation and vali- pared with the general Dutch population. pitalized patients with community-acquired dation study. Thorax 2003; 58: 377-82.
Clin Microbiol Infect 2011; 17: 763-8.
pneumonia. J Antimicrob Chemother 2009; 22. Huijts SM, van Werkhoven CH, Boers-
31. Ray WA, Murray KT, Hall K, Arbogast 63: 1025-33.
ma WG, et al. Guideline adherence for PG, Stein CM. Azithromycin and the risk 39. Antimicrobial resistance surveillance
empirical treatment of pneumonia and pa- of cardiovascular death. N Engl J Med in Europe: annual report of the European tient outcome: treating pneumonia in the 2012; 366: 1881-90.
Antimicrobial Resistance Surveillance Netherlands. Neth J Med 2013; 71: 502-7.
32. Schembri S, Williamson PA, Short Network (EARS-Net). Stockholm:
23. Schouten JA, Prins JM, Bonten MJ, et PM, et al. Cardiovascular events after clar-
pean Centre for Disease Prevention and al. Revised SWAB guidelines for antimi- ithromycin use in lower respiratory tract Control, 2012 (http://www .ecdc .europa crobial therapy of community-acquired infections: analysis of two prospective .eu/ en/ publications/ publications/ antimi- pneumonia. Neth J Med 2005; 63: 323-35.
cohort studies. BMJ 2013; 346: f1235.
24. Rietbergen C, Moerbeek M. The de-
33. Garin N, Genné D, Carballo S, et al. rope-2011 .pdf).
sign of cluster randomized crossover tri- β-Lactam monotherapy vs β-lactam- 40. Klugman KP. Bacteriological evidence
als. J Educ Behav Stat 2011; 36: 472-90.
macrolide combination treatment in mod- of antibiotic failure in pneumococcal lower 25. Kleinman LC, Norton EC. What's the erately severe community-acquired pneu-
respiratory tract infections. Eur Respir J risk? A simple approach for estimating monia: a randomized noninferiority trial. Suppl 2002; 36: 3s-8s.
adjusted risk measures from nonlinear JAMA Intern Med 2014; 174: 1894-901.
Copyright 2015 Massachusetts Medical Society. posting presentations from medical meetings online
Online posting of an audio or video recording of an oral presentation at a medical meeting, with selected slides from the presentation, is not considered prior publication. Authors should feel free to call or send e-mail to the Journal's Editorial Offices if there are any questions about this policy.
n engl j med 372;14 nejm.org April 2, 2015 The New England Journal of Medicine Downloaded from nejm.org at MEDISCH CENTRUM ALKMAAR on April 7, 2015. For personal use only. No other uses without permission. Copyright 2015 Massachusetts Medical Society. All rights reserved.

Source: http://www.foreestmedicalschool.nl/media/1308/nejm-2015cap-start.pdf