Factors Associated With Treatment Failure in Moderately Severe Community-Acquired Pneumonia

Key Points Question What are the risk factors for treatment failure in patients with community-acquired pneumonia (CAP) who reached clinical stability after 3 days of β-lactam treatment? Findings In this secondary analysis of a randomized clinical trial that included 291 adults, only male sex and age were associated with failure in the multivariable analysis. These results were independent of antibiotic treatment duration and biomarker levels. Meaning In this study, among patients with CAP who reached clinical stability after 3 days of antibiotic treatment, male sex and age were associated with higher risk of failure, suggesting that these factors should be taken in account in the treatment of patients with the condition.


Introduction
Up to 5.6 million cases of community-acquired pneumonia (CAP) occur annually in the US, 1 resulting in 600 000 to 800 000 hospitalizations, with the highest incidence rate in older patients. [1][2][3][4] Community-acquired pneumonia is a heterogeneous disease that ranges from a mild, self-limiting disease to a severe infection that causes respiratory failure, shock, and death. 5,6 Treatment failure is the most serious complication. Failure significantly increases the risk of complications, length of hospital stay, and death, especially in patients with severe CAP. 3,[7][8][9][10][11][12][13] The incidence of clinical failure in patients with CAP ranges from 6% to 24% 3,7-12 and can reach up to 31% in patients with severe CAP. 13 Several risk factors for treatment failure have been identified in the literature, such as age, smoking, malnutrition, previous CAP episodes, and comorbidities (chronic pulmonary disease, asthma, and immunosuppression). 14 Reaching clinical stability is associated with a high rate of favorable outcomes. 15 The Pneumonia Short Treatment (PTC) trial was a placebo-controlled randomized clinical trial that studied antibiotic treatment duration (3 vs 8 days of β-lactam treatment) among 310 patients hospitalized with moderately severe CAP who reached clinical stability at day 3 of treatment.
In this secondary analysis, we aim to evaluate the risk factors for treatment failure among this specific population.

Study Design, Sites, and Study Population
We performed a secondary analysis of a double-blind randomized clinical trial (the PTC trial), which included 310 patients with moderately severe CAP in 16 teaching hospitals in France, from December 19, 2013, to February 1, 2018. Data analysis was performed from July 18, 2019, to February 15, 2020. The flow diagram of the patients in the trial is shown in Figure 1. The primary outcome was treatment failure 15 days after first antibiotic intake, defined as a temperature greater than 37.9°C and/or absence of resolution or improvement of respiratory symptoms (coughing frequency or severity, sputum production, dyspnea, or crackles) and/or additional antibiotic treatment for any cause. The study design and main results have been published previously. 16 In brief, patients who met the following inclusion criteria were studied: age of 18 years or older, hospitalized in a non-critical care ward for CAP, treated with β-lactams, chosen at the discretion of the physician in charge to receive amoxicillin-clavulanate (oral or intravenous) or parenteral thirdgeneration cephalosporin (ceftriaxone or cefotaxime), and presented after 72 hours of treatment with a clinical response defined by the presence of stability criteria. 3 Race and ethnicity data are not allowed to be collected in France for research by law; therefore, these data are not available for analysis. Community-acquired pneumonia was defined as the association of at least 1 acute clinical sign compatible with pneumonia (among dyspnea, cough, purulent sputum, or crackles), temperature greater than 38°C, and a pulmonary infiltrate on chest radiography. Stability criteria were defined, according to the Infectious Diseases Society of America, as apyrexia (temperature Յ37.8°C), heart rate less than 100 beats/min, respiratory rate less than 24 breaths/min, arterial oxygen saturation of 90% or higher, systolic blood pressure of 90 mm Hg or higher, and normal

Variables
We performed this secondary analysis among the per-protocol study population, which included all patients randomly assigned to treatment, patients not erroneously included, patients who received their assigned treatment, and patients who received at least 80% of this treatment, except if discontinuation was attributable to worsening of their condition. Those who withdrew consent after more than 1 dose of study treatment and those lost to follow-up, except if they received additional treatment since day 3, were excluded.
The variables included demographic characteristics, clinical and radiological data, and results of usual blood tests from the first day of β-lactam treatment (day 0). Disease severity within the first 24 hours after diagnosis, determined with the PSI, 18 antibiotic treatment duration (3 or 8 days), and pneumonia-related symptoms scored using the CAP score 19 were also included in the secondary analysis (eAppendix in Supplement 1). (The CAP score, which is a clinical score used to quantify subjective CAP symptoms, has been described by Moussaoui et al. 19

Statistical Analysis
Continuous variables are presented as means (SDs) and as tabular descriptions for qualitative characteristics. Normality tests were first performed, and normal (and lognormal) distribution was found for all data. We used χ 2 tests to compare the distributions of categorical variables, whereas 2-tailed, unpaired t tests were used to compare the distributions of quantitative continuous variables. All reported P values were based on 2-sided tests, and P Յ .05 was considered statistically significant.
To identify risk factors associated with failure at day 15, a univariate analysis by logistic regression was performed, using demographic and medical characteristics as well as all clinical, biological, and radiological data from day 0 of antibiotic treatment. A multivariable analysis by logistic regression was then performed using all variables from the univariate analysis that had a P Յ .20, except for variables such as PSI score and urea nitrogen level to avoid multicollinearity with other variables from the regression model (eg, age, sex, and creatinine clearance). Odds ratios (ORs) were calculated from the univariate and multivariable analysis to quantify association with failure at day 15 with 95% CIs. An OR greater than 1.00 was considered to be associated with failure. Analyses were performed with the use of R software, version 3.6.1 (R Foundation for Statistical Computing).

Results
Among the 310 patients included in the PTC trail, the per-protocol analysis at day 15 (Figure 2).
The factors significantly associated with treatment failure at day 15 in univariate analysis ( and platelet count at day 0 (OR, 1.00; 95% CI, 1.00-1.00) had a P < .20 and therefore were not statistically significant in the univariate analysis but were included in the multivariable regression model. When PSI score and urea level at day 0 were excluded to avoid multicollinearity, the factors significantly associated with treatment failure at day 15 in multivariable analysis (

Discussion
In this secondary analysis of a randomized clinical trial of patients with CAP who reached clinical stability after 3 days of antibiotic treatment, only male sex and age were associated with higher risk of failure. Our study population is similar to the usual population described in the literature concerning patients hospitalized with CAP, with the exception of patients with immunosuppression, who were excluded from our trial. Indeed, the median age of the population with CAP usually described is between 70 and 74 years, with most being male and having a high number of comorbidities, such as diabetes, chronic obstructive pulmonary disease, chronic heart failure, and a PSI score of III to IV. [20][21][22]  Abbreviations: CAP, community-acquired pneumonia; PSI, Pneumonia Severity Index.
SI conversion: To convert creatinine clearance to milliliters per second per square meters, multiply by 0.0167; urea nitrogen to millimoles per liter, multiply by 0.357; sodium to millimoles per liter, multiply by 1; glucose to millimoles per liter, multiply by 0.0555; hematocrit to a proportion of 1.0, multiply by 0.01; white blood cells and neutrophils to ×10 9 per liter, multiply by 0.001; platelets to ×10 9 per liter, multiply by 1; and C-reactive protein to milligrams per liter, multiply by 10.
a Data are presented as mean (SD) unless otherwise indicated.
b Data statistically significant. Although approximately half of patients with moderately severe CAP reach clinical stability within 3 days, 15,23 which is associated with a shorter hospital stay and a better prognosis, 22 the originality of our study is to focus on those patients and observe their clinical outcome. Our study found that risks factors for treatment failure among patients hospitalized for CAP who reached clinical stability at day 3, in univariate analysis, were age and sex, which are well-known risk factors associated with failure in the literature and may be the most influential characteristics 24 ; PSI score, which has been used to evaluate the lethality rate 18 ; chronic lung disease; and kidney failure, which is often associated with failure and therefore could be a marker of severity rather than a proper risk factor. In the multivariable analysis, only male sex and age were significantly associated with treatment failure.

JAMA Network Open | Infectious Diseases
Risks factors for failure have been explored in the literature. Despite heterogeneity of failure definition, age and comorbid conditions have also been associated with a higher risk of failure in previous works. 9,12,25 In a systematic review of the literature, 14   pulmonary disease, asthma, functional impairment, poor dental health, immunosuppressive therapy, oral corticosteroids, and treatment with gastric acid-suppressive drugs. Some of these factors could be corrected, which would reduce morbidity and mortality among adult patients with CAP, particularly among the older patients. Regarding nonmodifiable criteria, age and chronic pulmonary disease were also identified in our study.
In the current study, only a few of these risk factors were significant, possibly because only patients with CAP who reached clinical stability at day 3 were included. As previously reported, reaching clinical stability is associated with a high rate of favorable outcome. 15  varying from 6% to 9%. 8,9 The failure rate in our trial seems higher than previously reported, especially among patients who reached early clinical stability probably because of the stringent definition of cure in our trial. Indeed, persistence of clinical symptoms (eg, cough) without worsening or additional antibiotic was considered as failure. In fact, our description of failure indicated that the main causes for classification as failure were the persistence or worsening of clinical symptoms. Few patients were classified as having treatment failure for additional antibiotic treatment or death. In addition, the main persistent or worsened symptoms of patients at day 15 were purulent sputum, cough, and dyspnea but not fever. However, these symptoms are usually reported as persistent among patients with pneumonia without indicating treatment failure. 26 In our study, all patients had similar evolutions of CAP score from day 0 to day 3. The CAP score at day 0 was not associated with failure. However, patients classified as having treatment failure at day 15 had lower CAP scores at day 8 and day 15 for each respiratory symptom. Therefore, this finding is consistent with the study's definition of failure, which included persistence or worsening of respiratory symptoms.
Finally, following biomarker levels was not associated with failure in our study. Indeed, C-reactive protein and procalcitonin levels, which were collected when available, did not seem informative among patients who were stable at day 3. Another study 27 also found that neither procalcitonin nor leukocyte count was associated with cure. Procalcitonin is an interesting tool to shorten treatment duration, as demonstrated by 2 meta-analyses. 28,29 Nevertheless, a recent trial 28 that compared procalcitonin levels and clinical evolution among patients with suspected lower respiratory tract infection found no association with antibiotic prescriptions. Finally, the recent Infectious Diseases Society of America and American Thoracic Society guidelines state that, despite reducing duration of antibiotic therapy, the use of procalcitonin level monitoring does not lead to overall lower durations of treatment compared with recommended durations. 29 This finding reinforces the importance of clinical examination on biological data.

Limitations
This study has limitations. Definition of CAP cases and outcomes varies widely among clinical trials.
The CAP diagnostic criteria within randomized clinical trials are heterogeneous, which could hinder the validity of their results. 30 Furthermore, a wide variety of failure definitions for CAP has been used, depending on trials. The US Food and Drug Administration has suggested a definition supported by clinical response, which is difficult to apply to current practice because of its complexity. 31 However, our study population included only patients who reached stability, which is associated with a