You are on page 1of 6

Clinical Microbiology and Infection 24 (2018) 1171e1176

Contents lists available at ScienceDirect

Clinical Microbiology and Infection


journal homepage: www.clinicalmicrobiologyandinfection.com

Original article

Are third-generation cephalosporins associated with a better


prognosis than amoxicillineclavulanate in patients hospitalized in the
medical ward for community-onset pneumonia?*
E. Batard 1, 2, *, F. Javaudin 1, 2, E. Kervagoret 1, E. Caruana 2, Q. Le Bastard 1, 2,
G. Chapelet 1, 3, N. Goffinet 2, E. Montassier 1, 2
1)
Universit
e de Nantes, Microbiotas Hosts Antibiotics Bacterial Resistances (MiHAR), Institut de Recherche en Sant
e 2, Nantes, France
2)
CHU Nantes, Emergency Department, Nantes, France
3)
CHU Nantes, Clinical Gerontology Department, Nantes, France

a r t i c l e i n f o a b s t r a c t

Article history: Objectives: We aimed to assess whether treatment with ceftriaxone/cefotaxime is associated with lower
Received 13 February 2018 in-hospital mortality than amoxicillineclavulanate in pati0ents hospitalized in medical wards for
Received in revised form community-onset pneumonia.
14 June 2018
Methods: We conducted a retrospective and multicentre study of patients hospitalized in French medical
Accepted 18 June 2018
Available online 28 June 2018
wards for community-onset pneumonia between 2002 and 2015. Treatments with ceftriaxone/cefo-
taxime or amoxicillineclavulanate were defined by their start in the emergency department for a
Editor: M. Paul duration of 5 days or more with no other b-lactam. A logistic regression analysis was performed on the
overall population, and a propensity score analysis was restricted to patients treated with either cef-
Keywords: triaxone/cefotaxime or amoxicillineclavulanate.
Amoxicillineclavulanate Results: 1698 patients (median age, 80 y) were included, of which 716 and 198 were treated with amoxi-
Cefotaxime cillineclavulanate and ceftriaxone/cefotaxime, respectively. In-hospital mortality was 10% (9e12%). In
Ceftriaxone multivariate analysis, factors associated with in-hospital mortality were treatment with ceftriaxone/cefo-
Community-onset pneumonia
taxime (aOR 2.9; (1.4e5.7)), pneumonia severity index class 4 or 5 (aOR 7.8 (4.3e15.7)), do-not-resuscitate
Mortality
order (aOR 8.7 (5.2e14.6)) and fluid therapy (aOR 6.3 (2.5e15.1)). The propensity score analysis was per-
formed on 178 patients treated with ceftriaxone/cefotaxime matched with 178 patients treated with
amoxicillineclavulanate; no significant association between treatment with ceftriaxone/cefotaxime and in-
hospital mortality was found (OR 1.5 (0.7e3.0)).
Conclusion: In the largest study aiming to compare amoxicillineclavulanate and ceftriaxone/cefotaxime
in community-onset pneumonia, ceftriaxone/cefotaxime was not associated with lower in-hospital
mortality than amoxicillineclavulanate. Our results suggest that ceftriaxone/cefotaxime should not be
preferred over amoxicillineclavulanate for patients hospitalized in medical wards with community-
onset pneumonia. E. Batard, Clin Microbiol Infect 2018;24:1171
© 2018 European Society of Clinical Microbiology and Infectious Diseases. Published by Elsevier Ltd. All
rights reserved.

Introduction treatment of patients with comorbidities and/or age of 65 years


who are hospitalized in the medical ward for a community-
Ceftriaxone, cefotaxime, and amoxicillineclavulanate are acquired pneumonia [1,2]. These antibiotics are frequently used
equally recommended by French and European guidelines for the for this condition [3e5]. It has long been suggested that narrow-
spectrum b-lactams be preferred over third-generation cephalo-
sporins (3GCs) to treat community-acquired pneumonia and to
*
This study has been presented as an oral flash presentation (O1035) at the preserve the microbial ecology of hospitals [6]. Resistance to 3GCs
ECCMID 2018 in Madrid, Spain. in the Enterobacteriaceae is associated with exposure to antibiotics
* Corresponding author. E. Batard, MiHAR Lab (Microbiotas Hosts Antibiotics and
 IRS2, 22 Boulevard Benoni-
in general and to cephalosporins and fluoroquinolones in particular
Bacterial Resistance), Institut de Recherche en Sante
Goullin, 44200 Nantes, France. [7e12]. Conversely, most studies have found no relationship be-
E-mail address: eric.batard@univ-nantes.fr (E. Batard). tween amoxicillineclavulanate use and resistance mediated by

https://doi.org/10.1016/j.cmi.2018.06.021
1198-743X/© 2018 European Society of Clinical Microbiology and Infectious Diseases. Published by Elsevier Ltd. All rights reserved.
1172 E. Batard et al. / Clinical Microbiology and Infection 24 (2018) 1171e1176

extended-spectrum b-lactamase (ESBL) or resistance to 3GCs in the therapies. Exposures to macrolides (including telithromycin and
Enterobacteriaceae [13e18]. A panel of French experts recently synergistins) and to fluoroquinolones were defined as a treatment
listed b-lactams by increasing order of spectrum width and for 1 day. Exposure to macrolides and fluoroquinolones was
resistance-promoting potential [19]. In this six-rank classification, defined alternatively by a treatment duration of 5 days or more.
amoxicillineclavulanate (rank #2) had a narrower spectrum and Medical data from the entire duration of the hospitalization were
lower resistance-promoting potential than 3GCs (rank #3). Hence, collected by trained and supervised abstractors into an electronic
from an ecological point of view, it may be preferable to treat database.
community-onset pneumonia with amoxicillineclavulanate rather
than 3GCs, although this point remains to be demonstrated. Statistical analysis
However, 3GCs have been increasingly used to treat this condition,
although ceftriaxone has never been shown to be superior to other Descriptive statistics were expressed as median (25the75th
b-lactamsdespecially amoxicillineclavulanatedin community- percentiles) and proportions (95%CI). First, we tested in the overall
onset pneumonia [20]. population the association between in-hospital mortality and
Here, we aim to assess whether treatment with ceftriaxone/ treatment, using logistic regression. All variables with a P
cefotaxime is associated with a better prognosis than value < 0.2 on univariate analysis were included in the multivariate
amoxicillineclavulanate in patients hospitalized in the medical analysis and were selected using an automated backward proced-
ward for community-onset pneumonia. ure. Interactions were added when necessary. Subgroup analyses
were performed on pneumonia severity index (PSI) class 5 patients,
Methods and on patients who had no DNR order in the ED. A sensitivity
analysis was performed on patients treated in centres where the
Study design, setting and participants mortality rate was not significantly different from the others.
Second, as the empirical selection of antibiotic might be influ-
This study was retrospectively conducted in two populations of enced by the presence of underlying comorbidities and illness
patients treated for community-onset pneumonia. The first study severity, we simulated the experimental comparison of ceftriaxone/
population originated from a French academic centre (Centre cefotaxime and amoxicillineclavulanate with a propensity score
Hospitalier Universitaire de Nantes, Nantes, France), the study analysis. This analysis was restricted to the subgroup of patients
period ranging from January 2002 to December 2015. Patients who treated with either amoxicillineclavulanate or cefotaxime/ceftri-
were hospitalized between 2002 and 2013 have been described axone, as defined earlier. The propensity score (PS) was defined as
previously [5,20]. The second study population originated from a the individual probability of being treated with ceftriaxone or
previously reported multicentre study in seven other French hos- cefotaxime. Cases and controls were 1:1 matched without
pitals, including two academic centres [5]. Local hospital guidance replacement and with a calliper set at 0.2. The balance in the dis-
limited the use of 3GCs for community-onset pneumonia in two of tribution of baseline covariates was checked using the standardized
the eight hospitals [5]. Eligible patients were selected from the mean difference (SMD) [21]. All statistical tests were two-tailed,
institution's databases using the following criteria: age of 18 and p 0.05 was considered statistically significant. Statistical an-
years, admission through the emergency department (ED), transfer alyses were performed using R software 3.2.3 (2015-12-10, http://
from the ED to any acute medical ward (except for intensive- and CRAN.R-project.org) with MASS, tableone and Matching libraries.
intermediate-care units), and a main diagnosis of pneumonia
according to the 10th International Classification of Diseases at Ethical issues
hospital discharge. In order to cover both community-acquired and
healthcare-associated pneumonia and to exclude hospital-acquired This study was approved by the ethics committee of Centre
pneumonia, patients transferred from another acute-care hospital Hospitalier Universitaire de Nantes (Number 2014-04-01). Ac-
to the ED were not eligible. Among eligible patients, we randomly cording to French law, no consent was required from patients in
selected 100 cases per year (except for 168 cases in 2013) using a this retrospective study.
computer-generated random number list [5,20]. Among these
randomly selected cases, inclusion and exclusion criteria were Results
subsequently applied. Patients were included if the diagnosis of
pneumonia was mentioned on the ED chart's conclusion and if an After case selection (Supplementary file, Fig. S1), 1698 patients
antibacterial agent was administered in the ED. Patients were were included. Among these, 956 originated partly from a previ-
excluded if any other acute infectious disease was diagnosed or ously described and ongoing monocentric population (from
suspected in the ED chart's conclusion. 2002e2015), and 742 originated from seven other hospitals (2013).
Patient characteristics are reported in Table 1. Blood culture was
Outcome, exposure variables and data sources sampled in the ED for 1161 patients, and was sterile, truly positive
and contaminated in 1071 (92.2% (90.5e93.7%)), 73 (6.3%
The primary outcome was in-hospital mortality. The exposure (5.0e7.9%)) and 17 (1.5% (0.9e2.4%)) cases, respectively. Blood
was a three-level variable: (a) treatment with a 3GC, defined as a cultures drawn outside the ED (i.e. in the medical ward) were truly
treatment with ceftriaxone or cefotaxime started in the ED and positive in five additional cases, and were contaminated in three
administered for at least 5 days without any other b-lactam; (b) additional cases (Supplementary file, Table S1). Only 869 patients
treatment with amoxicillineclavulanate, started in the ED and (51%) received one unique antibacterial agent during their hospital
administered for at least 5 days without any other b-lactam; (c) stay, and 484 (29%), 242 (14%) and 103 (6%) received two, three and
other treatment. This classification was based on the b-lactam four or more antibiotics; 198 patients were treated in the ED with
exposure, and patients in each class could have received non-b- ceftriaxone/cefotaxime for at least 5 days without any other b-
lactam antibiotics (e.g., a macrolide or a fluoroquinolone). In lactam, and 716 received amoxicillineclavulanate in the ED
addition, exposures to a 3GC and to amoxicillineclavulanate were without any other b-lactam for at least 5 days (Table 1). Among the
alternatively defined as a treatment with a 3GC (or amox- 784 patients of the ‘other treatment’ group, 393 (50%) and 480
icillineclavulanate) for 1 day, irrespective of other antibacterial (61%) were treated for at least 1 day with ceftriaxone/cefotaxime
E. Batard et al. / Clinical Microbiology and Infection 24 (2018) 1171e1176 1173

Table 1
Patient characteristics.

Description All patients Amoxicillineclavulanate 3GC p

Number 1698 716 (42%) 198 (12%) e


Age (years), median (IQR) 80 (67e87) 82 (73e88) 80 (69e86) 0.08
Male, n (%) 930 (55%) 395 (55%) 111 (56%) 0.89
Nursing home resident, n (%) 430 (1659) (25%) 179 (704) (25%) 62 (191) (31%) 0.09
Neoplastic disease, n (%) 260 (15%) 105 (15%) 41 (21%) 0.052
Liver disease, n (%) 44 (3%) 17 (2%) 8 (4%) 0.30
Congestive heart failure, n (%) 282 (17%) 137 (19%) 35 (18%) 0.72
Cerebrovascular disease, n (%) 251 (15%) 120 (17%) 36 (18%) 0.72
Renal disease, n (%) 144 (8%) 70 (10%) 21 (11%) 0.83
History of pneumonia, n (%) 248 (15%) 100 (14%) 35 (18%) 0.23
Immunocompromised, n (%) 192 (11%) 56 (8%) 39 (20%) <0.001
Altered mental status, n (%) 393 (1545) (23%) 169 (669) (24%) 49 (161) (25%) 0.81
Respiratory rate 30/min, n (%) 426 (1426) (25%) 164 (598) (23%) 55 (168) (28%) 0.18
Systolic blood pressure <90 mmHg, n (%) 70 (1689) (4%) 26 (713) (4%) 10 (197) (5%) 0.48
Temperature <35 C or 40 C, n (%) 66 (1682) (4%) 20 (709) (3%) 17 (196) (9%) <0.001
Pulse  125 bpm, n (%) 145 (1690) (9%) 49 (713) (7%) 18 (197) (9%) 0.36
Arterial pH < 7.35, n (%) 80 (734) (5%) 27 (283) (4%) 16 (103) (8%) 0.019
Blood urea nitrogen 11 mmol/L, n (%) 454 (1628) (27%) 192 (692) (27%) 66 (185) (33%) 0.086
Sodium <130 mmol/L, n (%) 85 (1676) (5%) 36 (711) (5%) 8 (194) (4%) 0.70
Glucose 14 mmol/L, n (%) 63 (1576) (4%) 25 (673) (3%) 10 (173) (5%) 0.42
Haematocrit <30%, n (%) 109 (1680) (6%) 45 (711) (6%) 15 (195) (8%) 0.63
Partial pressure of arterial oxygen <8 kPA or oxygen 394 (1672) (23%) 162 (701) (23%) 59 (196) (30%) 0.046
saturation <90%, n (%)
Pleural effusion, n (%) 97 (6%) 45 (6%) 6 (3%) 0.11
PSI class 1, n (%) 103 (6%) 17 (2%) 11 (6%)
PSI class 2, n (%) 180 (11%) 47 (7%) 14 (7%)
PSI class 3, n (%) 308 (18%) 151 (21%) 25 (13%)
PSI class 4, n (%) 699 (41%) 341 (48%) 86 (43%)
PSI class 5, n (%) 408 (24%) 160 (22%) 62 (31%) 0.002
Antibacterial treatment in the preceding 3 months, n (%) 174 (10%) 47 (7%) 35 (18%) <0.001
Antibacterial therapy before the arrival in the ED, n (%) 331 (19%) 89 (12%) 60 (30%) <0.001
Fluid therapy, n (%) 198 (12%) 47 (7%) 43 (22%) <0.001
Non-invasive ventilation, n (%) 15 (1%) 3 (0%) 4 (2%) 0.068
DNR order, n (%) 134 (8%) 31 (4%) 20 (10%) 0.003
Fluoroquinolone, n (%) 349 (21%) 77 (11%) 46 (23%) <0.001
Macrolide, streptogramin or ketolide, n (%) 290 (17%) 48 (7%) 59 (30%) <0.001
Bacteraemia, n (%) 78 (5%) 11 (2%) 6 (3%) 0.28
Centre 1, n (%) 109 (6%) 41 (6%) 6 (3%) <0.001
Centre 2, n (%) 102 (6%) 47 (7%) 10 (5%)
Centre 3, n (%) 93 (5%) 38 (5%) 16 (8%)
Centre 4, n (%) 138 (8%) 44 (6%) 34 (17%)
Centre 5, n (%) 89 (5%) 25 (3%) 24 (12%)
Centre 6, n (%) 956 (56%) 441 (62%) 70 (35%)
Centre 7, n (%) 114 (7%) 40 (6%) 25 (13%)
Centre 8, n (%) 97 (6%) 40 (6%) 13 (7%)
Academic status, n (%) 1179 (69%) 522 (73%) 101 (51%) <0.001

3GC, third-generation cephalosporin (ceftriaxone or cefotaxime); DNR, do not resuscitate; ED, emergency department; PSI, pneumonia severity index.
Treatment groups were defined as follows: 3GC, treatment with a 3GC started in the ED and 3GC was administered for at least 5 days without another b-lactam; amox-
icillineclavulanate, treatment with amoxicillineclavulanate started in the ED and administered for at least 5 days without another b-lactam. P values, comparison of 3GC and
amoxicillineclavulanate. As percentages have been rounded, their sum may not reach 100%. Numbers of patients with available data are reported in brackets when necessary.

and amoxicillineclavulanate, respectively. Overall, the numbers of mortality: ORs 2.7 (1.5e4.6) and 3.5 (2.4e5.2), respectively. There
patients treated for at least 1 day with amoxicillineclavulanate and was no significant association between death and treatment with a
cefotaxime/ceftriaxone were 1196 and 591, respectively. The macrolide or with a fluoroquinolone. Fluid therapy in the ED, DNR
median length of antibacterial therapy was 11 (8e13) days. The order in the ED, and PSI classes 4 and 5 were also significantly
median length of treatment with ceftriaxone/cefotaxime and associated with mortality. In-hospital mortality was low in centre
amoxicillineclavulanate was 7 (3e10) and 9 (5e11) days, #3 (2%, 95%CI 0e8%), whereas it ranged from 7% (95%CI 3e15%) to
respectively. 15% (95% CI 10e23%) in the other centres. Hence, centres were
dichotomized between this low-mortality centre (coded G2 in the
Predictors of mortality in the overall population centre variable, 93 patients) and other centres (coded G1 in the
centre variable, 1605 patients).
Among 1698 patients, 178 died, an in-hospital mortality of 10% Seven explanatory variables were entered in the multivariate
model: the three-level treatment group, PSI class 4 or 5, fluid
(9e12%). Mortality among patients treated with amox-
icillineclavulanate, patients treated with ceftriaxone/cefotaxime, therapy in the ED, DNR order in the ED, centre G2, antibacterial
treatment in the preceding 3 months, and treatment with a fluo-
and other patients was 35 out of 716 (5% (3e7%)), 24 out of 198 (12%
(8e18%)) and 119 out of 784 (15% (13e18%)), respectively. Odd ra- roquinolone. In the final model, treatment with a 3GC was signif-
icantly associated with in-hospital mortality (adjusted OR 2.9
tios (ORs) are reported in Table 2. In comparison with patients
treated with amoxicillineclavulanate, patients treated with a 3GC (1.4e5.7)) (Table 3). The Hosmer and Lemeshow goodness-of-fit
test showed no significant difference between observed and
and patients treated with other antibacterial regimens had higher
1174 E. Batard et al. / Clinical Microbiology and Infection 24 (2018) 1171e1176

Table 2
Risk factors for mortality in community-onset pneumonia: univariate analysis.

Characteristic Description Survivors Non-survivors OR (95%CI) p

Cases Number 1520 (90%) 178 (10%) e e


Demographic Age (years), median (IQR) 79 (65e86) 85 (79e90) 1.05 (1.03e1.06) <0.001
Male, n (%) 827 (54%) 103 (58%) 1.2 (0.8e1.6) 0.38
Nursing-home resident, n (%) 348 (1485) (23%) 82 (174) (46%) 2.9 (2.1e4.0) <0.001
Comorbid conditions Neoplastic disease, n (%) 220 (14%) 40 (22%) 1.7 (1.2e2.5) 0.005
Liver disease, n (%) 40 (3%) 4 (2%) 0.9 (0.3e2.1) 0.76
Congestive heart failure, n (%) 249 (16%) 33 (19%) 1.2 (0.8e1.7) 0.46
Cerebrovascular disease, n (%) 213 (14%) 38 (21%) 1.7 (1.1e2.4) 0.01
Renal disease, n (%) 125 (8%) 19 (11%) 1.3 (0.8e2.2) 0.27
History of pneumonia, n (%) 221 (15%) 27 (15%) 1.1 (0.7e1.6) 0.82
Immunocompromised, n (%) 170 (11%) 22 (12%) 1.1 (0.7e1.8) 0.64
Physical examination findings Altered mental status, n (%) 308 (1390) (20%) 85 (155) (48%) 3.6 (2.6e5.0) <0.001
Respiratory rate 30/min, n (%) 352 (1266) (23%) 74 (160) (42%) 2.4 (1.7e3.2) <0.001
Systolic blood pressure <90 mmHg, n (%) 55 (1513) (4%) 15 (176) (8%) 2.5 (1.3e4.3) 0.003
Temperature <35 C or 40 C, n (%) 60 (1508) (4%) 6 (174) (3%) 0.8 (0.3e1.8) 0.71
Pulse 125 bpm, n (%) 115 (1514) (8%) 30 (176) (17%) 2.5 (1.6e3.8) <0.001
Laboratory and radiographic Arterial pH <7.35, n (%) 54 (644) (4%) 26 (90) (15%) 4.6 (2.8e7.6) <0.001
findings Blood urea nitrogen 11 mmol/L, n (%) 367 (1462) (24%) 87 (166) (49%) 3.0 (2.2e4.1) <0.001
Sodium <130 mmol/L, n (%) 73 (1501) (5%) 12 (175) (7%) 1.4 (0.7e2.6) 0.26
Glucose 14 mmol/L, n (%) 51 (1408) (3%) 12 (168) (7%) 2.1 (1.0e3.9) 0.027
Haematocrit <30%, n (%) 95 (1506) (6%) 14 (174) (8%) 1.3 (0.7e2.2) 0.41
Partial pressure of arterial oxygen <8 kPA or oxygen 343 (1495) (23%) 51 (177) (29%) 1.4 (1.0e1.9) 0.07
saturation <90%, n (%)
Pleural effusion, n (%) 84 (6%) 13 (7%) 1.3 (0.7e2.4) 0.34
Pneumonia severity index (PSI) Class 1e3, n (%) 580 (38%) 11 (6%) e e
Class 4e5, n (%) 940 (62%) 167 (94%) 9.4 (5.3e18.5) <0.001
Antibiotic before admission Antibacterial treatment in the preceding 3 months, n (%) 150 (10%) 24 (13%) 1.4 (0.9e2.2) 0.13
Antibacterial therapy before arrival in the ED, n (%) 297 (20%) 34 (19%) 1.0 (0.6e1.4) 0.89
Treatment in the ED Fluid therapy, n (%) 149 (10%) 49 (28%) 3.5 (2.4e5.0) <0.001
DNR order, n (%) 70 (5%) 64 (36%) 11.6 (7.9e17.2) <0.001
In-hospital antibacterial therapy Amoxicillineclavulanatea, n (%) 681 (45%) 35 (20%) 0.3 (0.2e0.4)e <0.001
3GCb, n (%) 174 (11%) 24 (13%) 1.2 (0.7e1.9)e 0.42
Other treatmentc, n (%) 665 (44%) 119 (67%) 1.2 (0.7e1.9)e 0.42
Fluoroquinolone, n (%) 305 (20%) 44 (25%) 1.3 (0.9e1.9) 0.15
Macrolide, n (%) 262 (17%) 28 (16%) 0.9 (0.6e1.4) 0.61
Amoxicillineclavulanatea with neither macrolide or 574 (38%) 26 (15%) 0.3 (0.2e0.4) <0.001
fluoroquinolone, n (%)
Amoxicillineclavulanatea with macrolided, n (%) 48 (3%) 0 (0%) e e
Amoxicillineclavulanatea with fluoroquinoloned, n (%) 68 (4%) 9 (5%) 1.1 (0.5e2.2) 0.72
3GCb with neither macrolide nor fluoroquinolone, n (%) 87 (6%) 12 (7%) 1.2 (0.6e2.1) 0.58
3GCb with macrolided, n (%) 54 (4%) 5 (3%) 0.8 (0.3e1.8) 0.61
3GCb with fluoroquinoloned, n (%) 39 (3%) 7 (4%) 1.6 (0.6e3.3) 0.29
Macrolide alone, n (%) 22 (1%) 0 (0%) e e
Fluoroquinolone alone, n (%) 40 (3%) 0 (0%) e e
Blood culture Bacteraemia, n (%) 68 (4%) 10 (6%) 1.3 (0.6e2.4) 0.49
Centre Centre 1, n (%) 99 (7%) 10 (6%) 0.9 (0.4e1.6) 0.65
Centre 2, n (%) 93 (6%) 9 (5%) 0.8 (0.4e1.6) 0.57
Centre 3, n (%) 91 (6%) 2 (1%) 0.2 (0.0e0.6) 0.017
Centre 4, n (%) 117 (8%) 21 (12%) 1.6 (1.0e2.6) 0.06
Centre 5, n (%) 78 (5%) 11 (6%) 1.2 (0.6e2.2) 0.55
Centre 6, n (%) 850 (56%) 106 (60%) 1.2 (0.8e1.6) 0.36
Centre 7, n (%) 102 (7%) 12 (7%) 1.0 (0.5e1.8) 0.99
Centre 8, n (%) 90 (6%) 7 (4%) 0.7 (0.3e1.3) 0.28
Academic status, n (%) 1051 (69%) 128 (72%) 1.1 (0.8e1.6) 0.45

3GC, third-generation cephalosporin (ceftriaxone or cefotaxime); DNR, do not resuscitate; ED, emergency department; OR, odds ratio.
Treatment groups were defined as follows.
a
Amoxicillineclavulanate treatment started in the ED and was administered for at least 5 days without another b-lactam.
b
3GC treatment started in the ED and was administered for at least 5 days without another b-lactam.
c
The ‘other treatment’ group included patients who did not fulfil the classification criteria of the first two groups.
d
Exposure to macrolides and fluoroquinolones was defined by a treatment duration of 1 day or more.
e
OR compared each group to all the other groups. As percentages have been rounded, their sum may not reach 100%. Numbers of patients with available data are reported in
brackets when necessary.

predicted mortality (p 0.99). The area under the ROC curve of the defined as a treatment for 1 day or more during the hospital stay, it
multivariate model was 0.83 (95%CI 0.80e0.86). Results were un- remained associated with an increased in-hospital mortality in
changed when treatment with a macrolide or with a fluo- multivariate analysis (adjusted OR 1.6 (1.1e2.2)).
roquinolone was added to the model. When the exposures to
macrolides and fluoroquinolones were defined by a treatment Subgroup and sensitivity analyses
duration of 5 days or more, both were retained in the final multi-
variate model (Supplementary file, Table S2). However, the asso- Treatment with a 3GC remained associated with a higher in-
ciation between treatment with a 3GC and mortality remained hospital mortality when the same model was applied to the PSI
unchanged (adjusted OR 3.2, 1.5e6.5). If the exposure to a 3GC was class 5 subgroup of patients (n ¼ 408, aOR 2.5 (1.2e5.4)) or to
E. Batard et al. / Clinical Microbiology and Infection 24 (2018) 1171e1176 1175

Table 3
Risk factors for mortality in community-onset pneumonia: multivariate analysis.

Characteristic Description Adjusted OR (95%CI) p

Antibacterial therapy Amoxicillineclavulanate e e


i.v. 3GC 2.9 (1.4e5.7) 0.002
Other treatment 4.3 (2.7e7.1) <0.001
Centre Centre G1 e e
Centre G2 0.1 (0.0e0.5) 0.01
Pneumonia severity index (PSI) class Class 1e3 e e
Class 4e5 7.8 (4.3e15.7) <0.001
Treatment in the ED DNR order 8.7 (5.2e14.6) <0.001
Fluid therapy 6.3 (2.5e15.1) <0.001
Interactions Fluid therapy  3GC 0.2 (0.1e0.9) 0.036
Fluid therapy  other treatment 0.3 (0.1e0.9) 0.033
DNR order  fluid therapy 0.4 (0.2e1.0) 0.041

3GC, Third-generation cephalosporin (ceftriaxone or cefotaxime); DNR, do-not-resuscitate; ED, emergency department.
Centres were dichotomized between one low-mortality centre (G2) and the others (G1).
Treatment groups were defined as follows: 3GC, treatment with a 3GC started in the ED and administered for at least 5 days without another b-lactam; amox-
icillineclavulanate, treatment with amoxicillineclavulanate started in the ED and administered for at least 5 days without another b-lactam; the ‘other treatment’ group
included patients who did not fulfil the classification criteria of the first two groups. Hosmer and Lemeshow goodness of fit, p 0.99. Area under the ROC curve, 0.83 (95%CI,
0.80e0.86).

patients who had no DNR order in the ED (n ¼ 1564, aOR 2.3 were drawn from only 1161 among 1698 patientsdand some
(1.2e4.5)). Furthermore, results were unchanged when the low- components of the PSI score.
mortality G2 centre was excluded from analysis (Supplementary Second, we did not use multiple imputation to complete missing
file, Table S3). data. We considered missing data to indicate the absence of
severity, in order to assign a PSI class to each patient. Of note, Fine
Comparison of amoxicillineclavulanate and 3GCs by propensity and his colleagues, when they derived the PSI score, could not
score analysis distinguish missing and normal data [25]. Furthermore, our series
mortality rates (2% for patients of the PSI classes 1e3, and 15% for
Eight explanatory variables were included in the model patients of the PSI classes 4 and 5) were similar to mortality rates in
designed to estimate the probability of being treated with a 3GC: the cohort that has been used to derive the PSI score [25]. Hence,
PSI class 4 or 5 (aOR 1.1 (0.8e1.7)), DNR order in the ED (aOR 1.5 with the exception of the estimated prognostic value of bacter-
(0.7e2.9)), fluid loading in the ED (aOR 4.3 (2.6-7.0)), antibacterial aemia, this treatment of missing data probably only slightly biased
therapy in the 3 previous months (aOR 1.8 (1.0-3.1)), antibacterial our results.
therapy started before the arrival in the ED (aOR 3.5 (2.3-5.2)), Third, the level of evidence of the observational design is lower
immune compromise (aOR 2.9 (1.8-4.7)), non-invasive ventilation than for an RCT. However, assuming a 10% in-hospital mortality rate
in the ED (aOR 5.7 (1.2-30.7)), and academic status of the centre and a 3% non-inferiority margin, a trial aiming to show that
(aOR 0.4 (0.3-0.6)). This propensity model allowed matching of 178 amoxicillineclavulanate is non-inferior to ceftriaxone would
patients treated with a 3GC to 178 patients treated with amox- require more than 2000 patients, and is therefore unlikely ever to
icillineclavulanate. All of the explanatory variables of the pro- be conducted.
pensity score model were satisfactorily balanced between the Fourth, numerous patients (49%) received at least two antibac-
amoxicillineclavulanate and 3GC groups, as their SMDs ranged terial agents during their hospital stay. This may have biased the
between <0.001 and 0.107. There was no significant relationship comparison of patients treated with 3GC and with amox-
between treatment with cefotaxime/ceftriaxone and in-hospital icillineclavulanate. We limited this bias by defining exposures to
mortality in this matched sample of patients (OR 1.5 (0.7e3.0)). 3GC and to amoxicillineclavulanate without other b-lactams, and
In other words, the risk difference between matched 3GC- and by adjusting on treatment with a macrolide or a fluoroquinolone.
amoxicillineclavulanate-treated patients was 3.4% (e0.2% to 9.0%). Hence, there is plausibly no survival benefit with ceftriaxone/
cefotaxime over amoxicillineclavulanate when treating
community-onset pneumonia in medical wards. Ceftriaxone has a
Discussion
better activity than amoxicillineclavulanate against Gram-negative
bacteria, particularly against Enterobacteriaceae and Haemophilus
Here, we report the largest series assessing the association be-
influenzae. Conversely, amoxicillineclavulanate may be more
tween mortality and treatment with amoxicillineclavulanate or
effective in aspiration pneumonia that may be unrecognized in
ceftriaxone/cefotaxime in patients hospitalized in medical wards
elderly patients. Further studies are needed, especially conducted
for community-onset pneumonia. Our main finding was that the
in other cohorts, to compare the clinical efficacy and collateral ef-
in-hospital mortality was not lower in patients treated with cef-
fects of ceftriaxone and amoxicillineclavulanate. Of note, our re-
triaxone/cefotaxime in comparison with patients treated with
sults were obtained from patients admitted in medical wards, and
amoxicillineclavulanate, as well in the overall study population as
should not be extrapolated to patients admitted from the ED to
in the subgroup of most severe patients. Our results are consistent
intensive- and intermediate-care units. Indeed, patients in the
with those of previous studies that compared ceftriaxone or cefo-
intensive- or intermediate-care unit should be treated with a 3GC
taxime with amoxicillineclavulanate in community-acquired
according to European guidelines, although this point has been
pneumonia (Supplementary file, Table S4) [22e24].
debated [1,26].
This study has four limitations. First, this was a retrospective
In conclusion, in a series of 1698 patients hospitalized for
study. However, there were no missing data for the response var-
community-onset pneumonia, ceftriaxone/cefotaxime was not
iable (i.e., mortality), and few missing data for the explanatory
associated with a lower mortality than amoxicillineclavulanate.
variables. The latter mainly affected bacteraemiadas blood cultures
1176 E. Batard et al. / Clinical Microbiology and Infection 24 (2018) 1171e1176

Considering that amoxicillineclavulanate may favour resistance to Klebsiella species in hospitalized patients. Antimicrob Agents Chemother
2010;54:2010e6.
3GC in Enterobacteriaceae less than 3GCs, it may be advisable to
[10] Colodner R, Rock W, Chazan B, Keller N, Guy N, Sakran W, et al. Risk factors for
prefer amoxicillineclavulanate to ceftriaxone/cefotaxime in treat- the development of extended-spectrum beta-lactamase-producing bacteria in
ing patients hospitalized in medical wards with community-onset nonhospitalized patients. Eur J Clin Microbiol Infect Dis 2004;23:163e7.
pneumonia. [11] Quale JM, Landman D, Bradford PA, Visalli M, Ravishankar J, Flores C, et al.
Molecular epidemiology of a citywide outbreak of extended-spectrum beta-
lactamase-producing Klebsiella pneumoniae infection. Clin Infect Dis 2002;1:
Transparency declaration 834e41.
[12] Vernaz N, Huttner B, Muscionico D, Salomon JL, Bonnabry P, Lopez-Lozano JM,
et al. Modelling the impact of antibiotic use on antibiotic-resistant Escherichia
The authors declare no conflict of interest. coli using population-based data from a large hospital and its surrounding
community. J Antimicrob Chemother 2011;66:928e35.
[13] Piroth L, Aube H, Doise JM, Vincent-Martin M. Spread of extended-spectrum
Funding beta-lactamase-producing Klebsiella pneumoniae: are beta-lactamase in-
hibitors of therapeutic value? Clin Infect Dis 1998;27:76e80.
None. [14] Lepelletier D, Caroff N, Riochet D, Bizouarn P, Bourdeau A, Le Gallou F, et al.
Risk-factors for gastrointestinal colonisation with resistant Enterobacteri-
aceae among hospitalised patients: a prospective study. Clin Microbiol Infect
Acknowledgments 2006;12:974e9.
[15] Aldeyab MA, Harbarth S, Vernaz N, Kearney MP, Scott MG, Darwish Elhajji FW,
et al. The impact of antibiotic use on the incidence and resistance pattern of
All authors had access to the data. EB and NG are the guarantors extended-spectrum beta-lactamase-producing bacteria in primary and sec-
for the data. ondary healthcare settings. Br J Clin Pharmacol 2012;74:171e9.
[16] Gbaguidi-Haore H, Dumartin C, L’He riteau F, Pefau M, Hocquet D, Rogues A-M,
et al. Antibiotics involved in the occurrence of antibiotic-resistant bacteria: a
Appendix A. Supplementary data nationwide multilevel study suggests differences within antibiotic classes.
J Antimicrob Chemother 2013;68:461e70.
[17] Vibet MA, Roux J, Montassier E, Corvec S, Juvin ME, Ngohou C, et al. Systematic
Supplementary data related to this article can be found at analysis of the relationship between antibiotic use and extended-spectrum
https://doi.org/10.1016/j.cmi.2018.06.021. beta-lactamase resistance in Enterobacteriaceae in a French hospital: a time
series analysis. Eur J Clin Microbiol Infect Dis 2015;34:1957e63.
[18] Marquet A, Vibet M-A, Caillon J, Javaudin F, Chapelet G, Montassier E, et al. Is
References there an association between use of amoxicillineclavulanate and resistance to
third-generation cephalosporins in Klebsiella pneumoniae and Escherichia coli
[1] Woodhead M, Blasi F, Ewig S, Garau J, Huchon G, Ieven M, et al. Guidelines for at the hospital level? Microb Drug Resist 2018. https://doi.org/10.1089/
the management of adult lower respiratory tract infectionsefull version. Clin mdr.2017.0360.
Microbiol Infect 2011;17:E1e59. [19] Weiss E, Zahar JR, Lesprit P, Ruppe E, Leone M, Chastre J, et al. Elaboration of a
[2] SPILF, AFSSAPS. Antibiothe rapie par voie ge
ne
rale dans les infections respi- consensual definition of de-escalation allowing a ranking of beta-lactams. Clin
ratoires basses de l’adulte. Pneumonie aigue communautaire, exacerbations Microbiol Infect 2015;21:649.e1-10.
de bronchopneumopathie chronique obstructive. 2010. http://ansm.sante.fr/ [20] Goffinet N, Lecadet N, Cousin M, Peron C, Hardouin JB, Batard E, et al.
var/ansm_site/storage/original/application/ Increasing use of third-generation cephalosporins for pneumonia in the
b33b6936699f3fefdd075316c40a0734.pdf. [Accessed 10 June 2018]. emergency department: may some prescriptions be avoided? Eur J Clin
[3] Neuman MI, Ting SA, Meydani A, Mansbach JM, Camargo Jr CA. National study Microbiol Infect Dis 2014;33:1095e9.
of antibiotic use in emergency department visits for pneumonia, 1993 through [21] Austin PC. An introduction to propensity score methods for reducing the ef-
2008. Acad Emerg Med 2012;19:562e8. fects of confounding in observational studies. Multivar Behav Res 2011;46:
[4] Blasi F, Garau J, Medina J, Avila M, McBride K, Ostermann H. Current man- 399e424.
agement of patients hospitalized with community-acquired pneumonia [22] Xaba SN, Greeff O, Becker P. Determinants, outcomes and costs of ceftriaxone
across Europe: outcomes from REACH. Respir Res 2013;14:44. v. amoxicillineclavulanate in the treatment of community-acquired pneu-
[5] Batard E, Lecadet N, Goffinet N, Hardouin JB, Lepelletier D, Potel G, et al. High monia at Witbank hospital. Afr Med J 2014;104:187e91.
variability among emergency departments in 3rd-generation cephalosporins [23] Sanchez ME, Gomez J, Gomez Vargas J, Banos V, Ruiz Gomez J, Munoz L, et al.
and fluoroquinolones use for community-acquired pneumonia. Infection Prospective and comparative study between cefuroxime, ceftriaxone and
2015;22:681e9. amoxicillineclavulanic acid in the treatment of community-acquired pneu-
[6] Paterson DL, Playford EG. Should third-generation cephalosporins be the monia. Rev Espan ~ ola Quimioter 1998;11:132e8.
empirical treatment of choice for severe community-acquired pneumonia in [24] Roson B, Carratala J, Tubau F, Dorca J, Linares J, Pallares R, et al. Usefulness of beta-
adults? Med J Aust 1998;6:344e8. lactam therapy for community-acquired pneumonia in the era of drug-resistant
[7] Lee DS, Choe H-S, Lee SJ, Bae WJ, Cho HJ, Yoon BI, et al. Antimicrobial Streptococcus pneumoniae: a randomized study of amoxicillineclavulanate and
susceptibility pattern and epidemiology of female urinary tract infections ceftriaxone. Microb Drug Resist 2001;7:85e96.
in South Korea, 2010e2011. Antimicrob Agents Chemother 2013;57: [25] Fine MJ, Auble TE, Yealy DM, Hanusa BH, Weissfeld LA, Singer DE, et al.
5384e93. A prediction rule to identify low-risk patients with community-acquired
[8] Li D, Chen Y, Zhang W, Zheng S, Zhang Q, Bai C, et al. Risk factors for hospital- pneumonia. N Engl J Med 1997;23:243e50.
acquired bloodstream infections caused by extended-spectrum b-lactamase [26] Hariri G, Tankovic J, Boe €lle P-Y, Dube e V, Leblanc G, Pichereau C, et al. Are
Klebsiella pneumoniae among cancer patients. Ir J Med Sci 2014;183:463e9. third-generation cephalosporins unavoidable for empirical therapy of
[9] Wener KM, Schechner V, Gold HS, Wright SB, Carmeli Y. Treatment with community-acquired pneumonia in adult patients who require ICU admis-
fluoroquinolones or with beta-lactam-beta-lactamase inhibitor combinations sion? A retrospective study. Ann Intensive Care 2017;7:35.
is a risk factor for isolation of extended-spectrum-beta-lactamase-producing

You might also like