Severe atypical pneumonia in critically ill patients: a retrospective multicenter study

Background Chlamydophila pneumoniae (CP) and Mycoplasma pneumoniae (MP) patients could require intensive care unit (ICU) admission for acute respiratory failure. Methods Adults admitted between 2000 and 2015 to 20 French ICUs with proven atypical pneumonia were retrospectively described. Patients with MP were compared to Streptococcus pneumoniae (SP) pneumonia patients admitted to ICUs. Results A total of 104 patients were included, 71 men and 33 women, with a median age of 56 [44–67] years. MP was the causative agent for 76 (73%) patients and CP for 28 (27%) patients. Co-infection was documented for 18 patients (viruses for 8 [47%] patients). Median number of involved quadrants on chest X-ray was 2 [1–4], with alveolar opacities (n = 61, 75%), interstitial opacities (n = 32, 40%). Extra-pulmonary manifestations were present in 34 (33%) patients. Mechanical ventilation was required for 75 (72%) patients and vasopressors for 41 (39%) patients. ICU length of stay was 16.5 [9.5–30.5] days, and 11 (11%) patients died in the ICU. Compared with SP patients, MP patients had more extensive interstitial pneumonia, fewer pleural effusion, and a lower mortality rate [6 (8%) vs. 17 (22%), p = 0.013]. According MCA analysis, some characteristics at admission could discriminate MP and SP. MP was more often associated with hemolytic anemia, abdominal manifestations, and extensive chest radiograph abnormalities. SP-P was associated with shock, confusion, focal crackles, and focal consolidation. Conclusion In this descriptive study of atypical bacterial pneumonia requiring ICU admission, mortality was 11%. The comparison with SP pneumonia identified clinical, laboratory, and radiographic features that may suggest MP or CP pneumonia. Electronic supplementary material The online version of this article (10.1186/s13613-018-0429-z) contains supplementary material, which is available to authorized users.


Open Access
*Correspondence: sandrine.valade@aphp.fr 1 AP-HP, Medical ICU, Hôpital Saint-Louis, 1 Avenue Claude Vellefaux, 75010 Paris, France Full list of author information is available at the end of the article In previous non-ICU studies, compared to bacterial pneumonia, AP was associated with younger age and fewer comorbidities, a lower risk of severe respiratory failure, and better outcome [4,6,13,14,18]. For patients admitted to ICU, studies remained rare.
The main objective of the study was to describe AP in patients admitted to ICU. Our secondary objective was to compare the diagnostic strategy and outcomes between Mycoplasma pneumoniae-related pneumonia (MP) and Streptococcus pneumoniae-related pneumonia (SP) admitted to ICU.

Patients with atypical pneumonia (AP)
This is a retrospective chart review of adults admitted to 20 ICUs in France with a diagnosis of AP over the 16-year period from 2000 to 2015 (Additional file 1: Figure S1). Inclusion criteria were pneumonia defined with sepsis and a new pulmonary infiltrate on the chest radiograph and either a positive specific polymerase chain reaction (PCR) test for MP or CP on respiratory specimens (noninvasive samples or bronchoalveolar lavage) or blood serologic tests suggesting acute MP or CP infection (elevated specific IgM or fourfold increase in IgG level between two time points or elevated anti-MP IgG combined with presence of cold agglutinins) [19].

Data collection
Clinical and laboratory data at ICU admission were collected, as well as organ failure during ICU stay. The SAPS II score [20] was used to assess severity at ICU admission. We also collected extra-pulmonary symptoms; arthritis was defined as new inflammation with one or more joints, myocarditis with cardiac dysfunction and troponin elevation and cutaneous involvement with the onset of skin rash. Bacterial and/or viral co-infections at diagnosis were recorded.

Patients with Streptococcus pneumoniae pneumonia (SP-P)
Patients with MP-AP were compared to a group of consecutive patients with proven SP-P admitted to one of the study ICUs (Saint Louis Hospital, Paris) during the same period. SP-P was diagnosed based on sepsis with a new pulmonary infiltrate and identification of SP in at least one microbiological specimen (blood culture, respiratory specimen, or urinary antigen with no alternative diagnosis).

Statistical analysis
Categorical variables were described as n (%) and quantitative variables as median [25th-75th percentiles]. We first described the features in the patients with AP at ICU admission. Then, we conducted univariate analyses with a nonparametric test to compare the groups with MP-AP and SP-P. Finally, multiple correspondence analysis (MCA) was performed to identify the dimensions associated with the parameters at ICU admission (HIV, symptom duration, shock, confusion, diarrhea, physical chest findings, chest radiograph abnormalities, bilirubin level, and hemolytic anemia) and the causative organism, using the FactoMineR library in the R software platform. MCA is an extension of simple correspondence analysis designed to analyze relations among categorical variables. The aim is to redefine the principal dimensions or axes of the space in a way that captures the highest possible percentage of the inertia (which can be likened to the explained variance).
All tests were two-tailed. p values < 0.05 were considered significant. All statistical analyses were carried out using the R 2.13.1 statistical platform (http://www.Rproje ct.org).
The causative organism was MP in 76 (73%) patients and CP in 28 (27%) patients and was identified by serological testing (positive IgM or elevated IgG) in 71 patients, positive PCR on respiratory samples in 33 patients (18 on bronchoalveolar lavage, 10 on naso-pharyngeal aspirate, 2 on tracheal aspirate and 4 on nasal swab) and by both diagnostic methods in 5 patients. None of the collected variables differed between patients diagnosed with PCR, serology or both (Additional file 4: Table S3). Co-infection was found in 18 (20%) patients and was related to viruses (n = 9; influenza, rhinovirus, respiratory syncytial virus, coronavirus) or bacteria (n = 6; Haemophilus influenzae, Proteus mirabilis, Staphylococcus aureus, Serratia marcescens) or Pneumocystis jirovecii (n = 3). None of MP patients had co-infection with CP or SP.
The first-line antibiotics were active on MP and CP in 62 (60%) patients. Time from ICU admission to antibiotic initiation was 1 [0-4] day. Combination therapy was used in 61 (59%) patients and consisted to a third-generation cephalosporin (C3G) and a macrolide in 24 (39%) patients, a C3G and a quinolone in 13 (21%) patients, another betalactam and a macrolide in 16 (26%) patients, another betalactam and a quinolone in 6 (9%) patients, or another antibiotic and a macrolide in 2 (3%) patients. Antibiotics was adapted according to microbiology results with a macrolide (n = 72), a quinolone (n = 24) or a cycline (n = 3).

Outcomes of atypical pneumonia (AP)
Eleven (11%) patients died in the ICU. ICU stay length was 16.5 [9.5-30.5] days. Persistent hypoxemia was present at ICU discharge in 60 (58%) patients. By univariate analysis, factors associated with mortality were age ≥ 65 years (p = 0.033), signs of respiratory distress (p = 0.017), and interstitial opacities on the chest  Signs of consolidation and decreased breath sounds were more common in SP-P than in MP-AP (30% vs. 9% and 38% vs. 17%, respectively). MP-AP involved 4 quadrants on chest X-ray (26% vs. 8%, p = 0.013) but less frequently pleural space (5% vs. 11%, p = 0.007). The bilirubin level was higher in the patients with SP-P (15 [9.2-24.5] μmol/L vs. 8.4 [5.8-13] µmol/L, p = 0.0006). MP-AP was associated with the use of mechanical ventilation (66% vs. 50%, p = 0.049). ICU length of stay (LOS) seemed prolonged in case of MP-AP regardless of the ICU outcome (median LOS 37 vs. 5 days and 15 vs. 5 days, respectively, in patients who died in the ICU and in patients who were discharged alive). However, 28-day mortality was lower in the MP-AP group (5% vs. 20%, p = 0.005). Figure 2 shows the MCA results for the clinical and radiological characteristics at admission. Several characteristics discriminated between MP and CP. MP was more often associated with hemolytic anemia, abdominal manifestations and extensive chest radiograph abnormalities. SP-P was associated with shock, confusion, focal crackles, and focal consolidation.

Discussion
This multicenter study is the largest one analyzing 104 AP patients admitted to ICU. Extra-pulmonary symptoms were seen for one-third of patients, corresponding to data on previous study for patients not admitted to ICU [21]. However, AP in non-ICU patients was described as mild [6], whereas a substantial proportion of our patients had severe acute pneumonia, with shock at ICU admission for 10% of patients and mechanical ventilation required for 72% of patients including 45% of patients with ARDS.
In previous studies, patients with MP-AP were younger and had fewer comorbidities, lower respiratory disease severity and better outcomes [4,6,13,14,18]. In our study, with ICU patients, age was similar for patients with MP-AP and SP-P.
Previous studies also compared clinical and radiological features according to the causative organism of pneumonia [8,15,18]. In a Japanese cohort, among patients with pneumonia and audible crackles, these were more often heard only in late inspiration in patients with AP and throughout inspiration in patients with other bacteria [22]. In our study, patients with MP-AP had no specific clinical findings, except signs of consolidation which were associated with SP-P. On radiological findings, compared to SP-P, MP-AP was more often responsible for ground-glass opacification, centrilobular nodules, bronchial wall thickening, and diffuse radiological abnormalities [1,15,18]. In our study, extensive interstitial pneumonia was more common in MP-AP than in SP-P.
The Japanese Respiratory Society published guidelines for identifying MP-AP [17] and established a scoring system based on six parameters: age < 60 years, minor or no comorbidities, stubborn cough, abnormal chest auscultation, the absence of sputum and of an etiological agent identifiable by rapid diagnostic testing, and peripheral white blood cell count < 10,000/μL. A score ≥ 4 indicates a high probability of MP-AP (sensitivity, 88.7%; and specificity, 77.5%). Another scoring system performed well in separating patients with pneumonia into three groups: pyogenic bacteria; MP, CP, or virus; and unknown agent [23]. Nevertheless, neither scoring system had been assessed in ICU patients. In our study, MCA provided insights into differences between MP-AP and SP-P. Hemolytic anemia, diffuse chest radiograph abnormalities, and interstitial opacities were associated with MP-AP. On the contrary, HIV infection, shock, neurological symptoms, gastrointestinal symptoms, signs of consolidation, shorter symptom duration, higher bilirubin level, and radiological alveolar opacities were strongly linked to SP-P.
Compared to patients with SP-P, those with MP-AP more often required mechanical ventilation and spent more time in the ICU yet had a lower risk of death. This lower mortality may be ascribable to the smaller number of MP-AP patients with extra-pulmonary organ failure (shock, neurological manifestations, acute renal failure) and to the lower SAPS II severity score in the MP-AP group (32 [22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41] vs. 42 [30-55], p = 0.005). Interestingly, intracellular pathogens are underdiagnosed like viruses, but under-covered despite the availability of therapeutic agents. These findings are in line with these from Menendez et al. [24] who reported a lack of antibiotic compliance in patients with CAP. Our descriptive data may be useful to help clinicians to discriminate SP-related pneumonia and MP-related pneumonia, even if a double antibiotherapy active against atypical pathogens is recommended in severe patients.
This study had several limitations. First, the study design was retrospective and patients were included within a 16-year period. ICU management may have changed over this period. ICU admissions criteria could be different according to the center and the year of admission. Atypical pneumonia remains rare, and the main objective of the study was to describe AP in the most severe patients. However, the study assessed mostly the clinical and radiological characteristics at admission which would be unlikely to change between the centers.
Secondarily, SP-P patients were included from only one single center, whereas AP patients were included from several centers. The main objective of the study was to describe patients at ICU admission. Although admission rules would be different between the centers, the clinical presentation would not be affected. Thirdly, only patients with proven AP based on positive microbiologic samples were included. Half of the patients with MP-AP had their diagnosis based on serological testing. More recently only PCR was used to diagnose Mycoplasma pneumoniae infection. Positivity of IgM anti-MP is considered as the gold standard, and PCR sensitivity is equal [25]. Although some of the patients had serological tests with fourfold increase in IgG level between two time points, we believe that we included only proven MP-AP patients. Although different diagnostic tests were used within the study period and among the centers, those tests were enough sensitive and specific to include real MP-AP pneumonia.
Fourth, we did not include patients with Legionella pneumophila pneumonia, a more frequent atypical pneumonia. Although Legionella pneumophila pneumonia was associated with higher risk of ICU admission comparing to MP-AP and CP-AP, our goal was to focus on AP that is usually non-severe and only occasionally leads to ICU admission. Moreover, several studies analyzed Legionella pneumophila pneumonia. Similarly, we did not include more rare etiology of pneumonia as Q fever.