Predicting Mycoplasma pneumoniae and Chlamydophila pneumoniae in community-acquired pneumonia (CAP) pneumonia: epidemiological study of respiratory tract infection using multiplex PCR assays.

Community-acquired pneumonia (CAP) is a common illness that can lead to mortality. β-lactams are ineffective against atypical pathogen including Mycoplasma pneumoniae. We used molecular examinations to develop a decision tree to predict atypical pathogens with CAP and to examine the prevalence of macrolide resistance in Mycoplasma pneumoniae. We conducted a prospective observational study of patients aged ≥ 18 years who had fever and respiratory symptoms and were diagnosed with CAP in one of two community hospitals between December 2016 and October 2018. We assessed combinations of clinical variables that best predicted atypical pathogens with CAP by classification and regression tree (CART) analysis. Pneumonia was defined as respiratory symptoms and new infiltration recognized on chest X-ray or chest computed tomography. We analyzed 47 patients (21 females, 44.7%, mean age: 47.6 years). Atypical pathogens were detected in 15 patients (31.9%; 12 Mycoplasma pneumoniae, 3 Chlamydophila pneumoniae). Ten patients carried macrolide resistant Mycoplasma pneumoniae (macrolide resistant rate 83.3%). CART analysis suggested that factors associated with presence of atypical pathogens were absence of crackles, age < 45 years, and LD ≥ 183 U/L (sensitivity 86.7% [59.5, 98.3], specificity 96.9% [83.8, 99.9]). ur simple clinical decision rules can be used to identify primary care patients with CAP that are at risk for atypical pathogens. Further research is needed to validate its usefulness in various populations.Trial registration Clinical Trial (UMIN trial ID: UMIN000035346).

Introduction

Community-acquired pneumonia (CAP) can lead to mortality [1-3]. Atypical pathogens, such as Mycoplasma pneumoniae, Chlamydophila pneumoniae, and Legionella pneumophila, are recognized as important causes of CAP [4].

Mycoplasma pneumoniae is one of the most common pathogens in CAP [5]. β-lactams, which are commonly-used antimicrobials, are ineffective against it [6]. According to the Cochrane review in 2012, no clinical symptoms or signs are especially useful for reliable diagnosis of CAP with atypical pathogens including Mycoplasma pneumoniae [7]. Diagnostic scoring criteria for consideration of atypical pathogen infections among adult patients with pneumonia were recently published by the Japanese Respiratory Society (JRS), and are now widely used in Japan [8]. Unreliable testing methods were used in the validation study, however, so there has been a need for a more accurate diagnostic method. Where it is available, the molecular method has become an option as a reference test to identify multiple respiratory pathogens. To rule out the diagnosis of atypical pathogens, more reliable criteria are needed. We developed a decision tree to address this gap. To improve the diagnosis of atypical pathogens in patients with pneumonia, we employed several molecular methods.

This study was conducted to develop a decision tree to predict atypical pathogens with CAP confirmed by molecular examinations.

Materials and methods

Subjects

Based on studies of fever in the elderly and on a study of influenza, we recruited patients who were febrile (1 °C higher than their baseline body temperature, or > 37 °C), and who were coughing for at least 3 days [9-11]. They were aged ≥ 18 years and were diagnosed with upper respiratory tract infection (URTI) in one of two community hospitals between December 2016 and October 2018. This study focuses on patients with community-acquired pneumonia and was conducted as a part of our prospective observational research investigating the characteristics of atypical pathogen infections [12, 13].

The study sites were the Tone Chuo Hospital (TCH, 253 beds) and the Akashi Medical Center (AMC, 382 beds), both local medical support centers located in Japan with emergency medical care centers and primary care practices. Excluded from this study were patients without informed consent, those with unstable physical conditions (e.g. shock, coma or impaired consciousness), those for whom sample collections were unable to be performed safely, those with history of multiple exacerbations of chronic pulmonary disease, apparent history or presence of dysphagia, presence of obstructive pneumonia, lung abscess, empyema, healthcare-associated pneumonia, or hospital-onset pneumonia referred from other facilities, tuberculosis, nontuberculous mycobacterium lung infections, lung mycosis, sinusitis, or tonsillitis, and patients with a recent history of fever or cough lasting more than 21 days. The patients who took antibiotics at home were not excluded from the study to promote generalizability.

Outcome measures

The primary outcome was CAP with atypical pathogens. Pneumonia was defined as respiratory symptoms and new infiltration that could be recognized on chest X-ray or chest computed tomography [14]. Early in the course of infection, chest CT can sometimes aid in the detection of CAP when chest radiographies are normal [15, 16]. All images were reviewed by a board-certified pulmonary physician (N.I.) for the determination of the final diagnosis. Nasopharyngeal or pharyngeal samples were obtained from all patients at the time of enrollment. Detection of atypical pathogens was made using FilmArray system (Biomérieux, USA) and the FilmArray Respiratory Panel tests for a comprehensive panel of 20 respiratory viruses and bacteria [17]. Analyses of macrolide resistance were performed by GENECUBE Mycoplasma system (TOYOBO, Co., Ltd., Osaka, Japan) [18], because it uses pharyngeal samples and has a higher M. pneumoniae detection rate than nasopharyngeal samples used in FilmArray system [19]. We collected demographic and clinical data on the age, gender, visiting month, comorbidities, history of close contact with confirmed atypical pathogen infections, history of preceding antimicrobial use, history of signs and symptoms (rhinorrhea, sputum, severe cough, sore throat, myalgia, arthralgia, diarrhea, and rash), duration of symptoms at the time of clinical visits, findings of chest auscultation, laboratory findings (white blood cell [WBC] count and C-reactive protein [CRP] levels), CURB-65 score, A-DROP score, and presence of pneumonia [20, 21]. Severe cough was defined as cough with vomiting, or that disturbed sleep, or was persistent [22]. If sputum was available, a quantitative culture was obtained. We used the IFCC-recommended method for lactate dehydrogenase (LD) measurement to reduce fluctuation. If necessary, the physician performed antigen testing (influenza antigen testing, pneumococcal urinary antigen testing, legionella urinary antigen testing, Mycoplasma pneumoniae antigen testing), or loop-mediated isothermal amplification method of sputum sample for the detection of Legionella pneumophila.

The study design was registered as a University Hospital Medical Information Network Clinical Trials Registry (UMIN-CTR) Clinical Trial (UMIN trial ID: UMIN000035346) on 22 December 2018 (UMIN-CTR URL: http://www.umin.ac.jp/ctr/index.htm). This study was approved by the Akashi Medical Center Research Ethics Committee.

Statistical analyses

To determine the best prediction model for atypical pathogens in CAP patients, we performed classification and regression tree (CART) analysis [23]. In CART analysis, we classified prognostic groups according to the interaction between variables and we decided the cutoff point in each variable. A receiver-operating characteristics (ROC) curve was used to evaluate the sensitivity, specificity, and correct diagnosis rate of the scores for atypical pathogens by JRS guidelines, with the area under the curve (AUC) indicating its discriminatory ability. For analysis of patient characteristics, we used Fisher’s exact test for categorical variables and performed Student’s t test for continuous variables. We also evaluated the utility of the scores for atypical pathogens by the published JRS guidelines in our study population [8]. Scores were determined by the following items: (i) age < 60 years old; (ii) absence of, or only minor underlying diseases; (iii) stubborn cough; (iv) negative or scant chest auscultatory findings; (v) no sputum, or no identified etiological agent by rapid diagnosis; and (vi) white blood cell count < 10,000/μL. The JRS scoring criteria without laboratory tests consisted of items i–v, and a score ≥ 3 were considered to be indicative of an atypical pathogen pneumonia. The scoring criteria with laboratory tests consisted of items i–vi, and a score ≥ 4 was considered indicative of an atypical pathogen infection. All statistical analyses were performed using JMP Pro 11.2.1 software program (SAS Institute Inc., Cary, NC, USA).

Results

Figure 1 shows the flow of participants, 51 patients were assessed for eligibility. We excluded patients with chronic symptoms (n = 3) and one patient without data on outcome measures (n = 1). The final study population was 47 patients, including 21 females (44.7%). The mean age of the patients was 47.6 (SD 20.1) years old. Besides chest X-ray, chest computer tomography (CT) scan was performed for 19 patients. The most frequent chest CT finding was consolidation, which was found in 15 patients. Four patients were diagnosed with CAP solely by chest CT scan findings. Comorbidities were as follows: chronic heart failure (n = 1, 2.1%), chronic kidney disease (n = 2, 4.3%), chronic liver disease (n = 3, 6.4%), and diabetes mellitus (n = 7, 14.9%). Two patients were immobile. Among all patients, 40 patients (85.1%) reported sputum, 31 patients (66.0%) reported malaise, 24 patients (51.1%) presented with headache, and 24 patients (51.1%) presented with heat sensation. Mean CURB-65 score was 0.4 (SD 0.7) and mean A-DROP score was 0.3 (SD 0.6) (Table 1).

Fig. 1

Flowchart of patient enrollment and analysis

Table 1

Characteristics of study patients, and patients with and without AP

	All patients	AP positive	AP negative	p value
n	47	15	32
Age (years)	47.6 [20.1]	35.0 [7.3]	53.5 [21.5]	< 0.01
Female	21 (44.7)	8 (53.3)	13 (40.6)	0.41
Asthma	1 (2.1)	0 (0.0)	1 (3.1)	1
Chronic pulmonary disease	0 (0.0)	0 (0.0)	0 (0.0)	NA
Chronic heart failure	1 (2.1)	0 (0.0)	1 (3.1)	1
Chronic kidney disease	2 (4.3)	0 (0.0)	2 (6.3)	0.56
Chronic liver disease	3 (6.4)	0 (0.0)	3 (9.4)	0.54
Central nervous disease	0 (0.0)	0 (0.0)	0 (0.0)	NA
Diabetes mellitus	7 (14.9)	1 (6.7)	6 (18.8)	0.4
Immunocompromised status	0 (0.0)	0 (0.0)	0 (0.0)	NA
Immobilization	2 (4.3)	0 (0.0)	2 (6.3)	0.56
Season (August–December)	18 (38.3)	7 (46.7)	11 (34.4)	0.42
Close contact	4 (8.5)	3 (20.0)	1 (3.1)	0.09
Preceding antimicrobial use	16 (34.0)	7 (46.7)	9 (28.1)	0.21
Macrolides, quinolones or tetracyclines use	4 (8.5)	1 (6.7)	3 (9.4)	1
Days after onset of illness	7.8 [3.4]	7.1 [2.9]	8.2 [3.5]	0.29
Rhinorrhea/nasal congestion	21 (44.7)	9 (60.0)	12 (37.5)	0.15
Sputum	40 (85.1)	14 (93.3)	26 (81.3)	0.4
Severe cough	19 (40.4)	7 (46.7)	12 (37.5)	0.55
Sore throat	22 (46.8)	6 (40.0)	16 (50.0)	0.52
Dyspnea	14 (29.8)	5 (33.3)	9 (28.1)	0.74
Myalgia/arthralgia	21 (44.7)	7 (46.7)	14 (43.8)	0.85
Headache	24 (51.1)	10 (66.7)	14 (43.8)	0.14
Malaise	31 (66.0)	10 (66.7)	21 (65.6)	0.94
Heat sensation	24 (51.1)	7 (46.7)	17 (53.1)	0.68
Chill	22 (46.8)	8 (53.3)	14 (43.8)	0.54
Diarrhea	5 (10.6)	2 (13.3)	3 (9.4)	1
Emesis	6 (12.8)	3 (20.0)	3 (9.4)	1
Systolic blood pressure, mmHg	124.7 [20.3]	125.5 [13.0]	124.4 [22.9]	0.87
Diastolic blood pressure, mmHg	74.7 [15.0]	80.4 [10.1]	72.2 [16.2]	0.09
Pulse rate, bpm	93.4 [14.6]	99.1 [13.7]	90.9 [14.5]	0.08
BT, ℃	37.5 [0.8]	37.5 [0.6]	37.5 [0.9]	0.97
Respiratory rate	17.1 [4.2]	15.9 [4.0]	17.7 [4.2]	0.18
SpO2, %	96.1 [2.1]	96.7 [1.8]	95.8 [2.3]	0.22
Crackles	15 (31.9)	0 (0.0)	15 (46.9)	< 0.01
Decreased breath sounds	6 (12.8)	0 (0.0)	6 (18.8)	0.16
Rash	1 (2.1)	0 (0.0)	1 (3.1)	1
Tonsil swollen	2 (4.3)	1 (6.7)	1 (3.1)	1
Tonsil white pus	0 (0.0)	0 (0.0)	0 (0.0)	NA
Cervical lymphadenopathy	1 (2.1)	0 (0.0)	1 (3.1)	1
WBC, /mm3	9210 [3122]	8282 [1944]	9540 [3411]	0.26
CRP, mg/dL	8.9 [14.6]	6.6 [6.5]	9.2 [7.5]	0.64
LD, U/L	232.9[110.9]	211.5 [52.2]	240.8[125.7]	0.46
BUN, mg/dL	13.4 [9.2]	10.7 [3.9]	14.3 [10.4]	0.27
CURB-65	0.4 [0.7]	0.0 [0.0]	0.6 [0.7]	< 0.01
A-DROP	0.3 [0.6]	0.0 [0.0]	0.4 [0.7]	< 0.01
Admission on the day	18 (38.3)	1 (2.1)	17 (53.1)	< 0.01

BT: Body temperature, SpO2: saturation of percutaneous oxygen, AP: Atypical pathogens, NA: Not applicable

Categorical data are presented as numbers (proportion, %). Continuous data are presented as mean values [standard deviation]

Eighteen patients were admitted on the day of hospital visit. Patients with atypical pathogens were younger than the patients without CAP. Crackles were not found in any patients with atypical pathogen. CURB-65 score, A-DROP score and admission rate were low in patients with atypical pathogen.

Atypical pathogens were found in 15 patients (32%), which included 12 patients with Mycoplasma pneumoniae, including 10 with macrolide resistance, and three patients with Chlamydophila pneumoniae. Macrolide resistance rate among patients with Mycoplasma pneumoniae was 83% (Table 2).

Table 2

Microbiological characteristics of the study patients

n	Total
Atypical pathogens	15 (31.9)
Mycoplasma pneumoniae	12 (25.5)
Macrolide resistant Mycoplasma pneumoniae	10 (21.3)
Chlamydophila pneumoniae	2 (4.3)
Chlamydophila pneumoniae + human Rhinovirus	1 (2.1)
Streptococcus pneumoniae†	2 (4.3)
Viruses	5 (10.6)
Parainfluenza virus	2 (4.3)
Human rhinovirus	1 (2.1)
Human metapneumovirus	1 (2.1)
Human coronavirus OC43	1 (2.1)

Macrolide resistance rate: 83.3%

†Diagnosis was made from the result of sputum cultures

Among these 15 patients, one patient had both atypical pathogen and viral infection (C. pneumoniae with human rhinovirus). Viral infections without accompanying atypical pathogen infections were found in five patients (10.6%). Adenovirus, Bordetella pertussis and Influenza were not detected. Legionella pneumophila was found in a patient’s sputum by loop-mediated isothermal amplification method. Streptococcus pneumoniae were yielded from the sputum cultures of two patients.

Figure 2 shows the decision tree for the presence of atypical pathogens. Among patients with no crackles, those < 45 years of age and those with LD > 183 U/L, 13 out of 14 patients had atypical pathogens. Patients with LD > 183 numbered fourteen in the “all patients” group and just one in the “negative patients” group. The decision tree discriminated atypical pathogens with sensitivity 86.7% (95% CI 0.60–0.98), specificity 96.9% (95% CI 0.84–1.00) and correct diagnosis rate was 93.6% (95% CI 68.0–100%) (Tables 3, 4 and 5).

Fig. 2

Decision tree for the presence of atypical pathogens. Among patients with no crackles, those < 45 years and those with LD > 183 U/L, 13 out of 14 patients had atypical pathogens

Table 3

Our decision tree criteria (n = 47)

	Atypical pathogens positive	Atypical pathogens negative	p value
Predicted	13 (92.9%)	1 (7.1%)	< 0.001
Not predicted	2 (6.1%)	31 (93.9%)

Sensitivity 86.7% [59.5, 98.3]

Specificity 96.9% [83.8, 99.9]

Correct diagnosis rate 93.6% [68.0, 100]

Table 4

JRS criteria without laboratory tests (n = 47)

	Atypical pathogens positive	Atypical pathogens negative	p value
Score ≥ 3	15 (50.0%)	15 (50.0%)	< 0.001
Score < 3	0 (0.0%)	17 (100.0%)

Sensitivity 100% [69.8, 100]

Specificity 53.1% [34.7, 70.9]

Correct diagnosis rate 68.1% [52.9, 80.9]

Table 5

JRS criteria with laboratory tests (n = 42)

	Atypical pathogens positive	Atypical pathogens negative	p value
Score ≥ 4	11 (57.9%)	8 (42.1%)	< 0.001
Score < 4	0 (0.0%)	23 (100.0%)

Missing values for “laboratory tests” (n = 5)

Sensitivity 100% [61.5, 100]

Specificity 74.2% [55.4, 88.1]

Correct diagnosis rate 81.0% [65.9, 91.4]

Using the JRS atypical pathogen diagnostic scoring criteria, 30 (63.8%; 30/47) met the score (≥ 3) for the criteria without laboratory tests, and 19 (45.2%; 19/42) met the score (≥ 4) for the criteria with laboratory tests. The JRS criteria without laboratory tests discriminated atypical pathogens with AUROCC of 0.79, sensitivity 100% (95% CI 69.8–100%), specificity of 53.1% (95% CI 34.7–70.9%) and correct diagnosis rate was 68.1% (95% CI 52.9–80.9%). The JRS criteria with laboratory tests discriminated atypical pathogens with AUROCC of 0.87, sensitivity 100% (95% CI 61.5–100%), specificity of 74.2% (95% CI 55.4–88.1%), and correct diagnosis rate was 81.0% (95% CI 65.9–91.4%) (Tables 3, 4 and 5, Fig. 3).

Fig. 3

ROC analysis of decision tree to differentiate the presence of atypical pathogens based on the Japanese guidelines. The decision tree discriminated atypical pathogens with ROC area of 0.87, sensitivity 100%, and specificity 74.2% for the criteria with laboratory tests and ROC area of 0.79, sensitivity 100%, and specificity 53.1% for the criteria without laboratory tests. ROC Receiver-operating curve

Discussion

In summary, to identify primary care patients with CAP that may be at risk for atypical pathogens, our decision tree uses three items: absence of crackles, age < 45 years and LD > 183 U/L. The clinical decision rules can identify primary care patients with CAP at risk for atypical pathogens with high yield (sensitivity 86.7%, specificity 96.9%). It is necessary to compare the diagnostic performance of the JRS criteria in the current study with that of the JRS criteria in previous studies (sensitivity 70.4% and specificity 91.8% in the original study, and sensitivity 77.0% and specificity 93.0% in the validation study) [8, 24]. Ishida et al. validated JRS scoring criteria retrospectively and included patients with Mycoplasma pneumoniae pneumonia, Chlamydophila pneumoniae pneumonia, pneumococcal pneumonia, and Hemophilus influenzae pneumonia. They omitted patients with viral pneumonia, which accounts for approximately 10% of the patients in our study population [8]. We performed validation of JRS scoring criteria of our study participants and the sensitivity and specificity for atypical pathogen infections were 100% and 74.2%, respectively, for the criteria with laboratory tests. Higher detection rate of the pathogens, including viruses by molecular methods would explain the higher sensitivity and lower specificity in our study participants. JRS scoring criteria may have the potential to be utilized to rule out the atypical pathogens with CAP.

Higher detection rate of the pathogens is also required to maintain the diagnostic accuracy in clinical prediction rules because reliable data could not be obtained by using a standard method that is imperfect [25]. Lui et al. assessed CAP-hospitalized patients in a prospective observational study using cultures, antigen testing and paired serology [4]. They could not provide a cutoff point with reasonable sensitivity and specificity to discriminate patients with pneumonia caused by atypical pathogens from patients with bacterial pneumonia. Causal organisms were identified in only 39.2% of their patients [4]. In the present prospective study, 23 out of 47 patients (48.9%) were positive for atypical pathogens. We used a molecular method to identify the pathogens of CAP, a more sensitive method for detection of pathogens than conventional methods. Jain et al. reported that a pathogen was detected in only 38% of patients among adults with radiographic evidence of pneumonia in a prospective population-based surveillance study, although the study did not address any clinical prediction rules [26]. Their study used culture, serologic testing, and antigen detection combined with molecular testing [26]. Most of their specimens, except for blood cultures, were taken after the administration of antimicrobials. In contrast, 34% of participants reported preceding antimicrobial use, which might explain the high-pathogen detection rate in our study.

One of the items used in the present study, age < 45 years, is consistent with previous reports on a clinical prediction rule for atypical pathogens with CAP. [4, 27] Lui et al. developed a prediction rule to discriminate CAP caused by atypical pathogens composed of age < 65 years, female gender, fever ≥ 38.0 °C, respiratory rate < 25/min, pulse rate < 100/min, serum sodium > 130 mmol/L, leucocyte count < 11,000/µL and Hb < 11 g/dL (sensitivity 54.0% and specificity 80.0%) [4]. Their study was designed for hospitalized patients, and a majority of atypical pathogen infections were elderly patients (63.4%) with comorbidities (41.8%) [4]. Older patients are at risk of early mortality, and therefore require hospitalization [28]. In our study, patients were younger and had fewer comorbidities than those in Lui’s study, so our prediction model might be better suited to primary care settings, including outpatients. The prediction model to discriminate CAP caused by Mycoplasma pneumoniae reported by Liu et al. included the characteristics of being < 45 years of age and not coexisting diseases (sensitivity 54.9%, specificity 58%) [27]. The study precluded chlamydophilial infections. It also lacked data on LD, which played a role in the items for discriminating atypical pathogens with CAP in our study.

Macrolide resistance in Mycoplasma pneumoniae being an emerging worldwide problem is also of great importance [29]. Patients with macrolide-resistant Mycoplasma pneumonia have presented prolonged fever and cough with high prevalence of extrapulmonary complications, sometimes resulting in life-threatening infection [30-32]. Mutation analysis with molecular methods can reliably determine the presence of macrolide resistance [30, 33]. Among CAP patients, the reported macrolide resistance rate has been reported as 88.3% in China, 70.3% in Korea, 49.4% in Japan, 20% in Italy, 10% in the United States and 3.1% in Germany [34-39]. In the present study, macrolide resistance rate was as high as 83.3% among atypical pathogens with CAP. Regional differences in macrolide resistance rate have also been reported in Japan, ranging in prevalence between 50 and 93% [40]. Akashi et al. reported that preceding macrolide use was a risk factor for macrolide resistance [33], although this was uncommon (< 10%) among our patients. The high resistance rates in our study might be associated with regional factors, such as previous excessive use of macrolides and lack of tight control of antimicrobial drug prescriptions. Further adequately sized studies should aim to determine the reason for the high macrolide resistance rate among patients with CAP.

Several limitations associated with the present study warrant mention. First, participants were recruited from just two institutions and a modest number of patients, so validation in future studies is required. Second, we did not include Legionella pneumophila in the respiratory panel tests we used, and one patient with Legionella pneumophila could not therefore be included in the atypical pathogen group. Third, our study excluded patients with critical conditions (shock, coma or impaired consciousness) and some of these patients might have had higher likelihood of pneumonia due to typical pathogens such as Streptococcus pneumoniae. Fourth, we used upper respiratory tract samples for detection of pathogens instead of lower respiratory tract samples (e.g., sputum and bronchial lavage fluid), but collecting and testing of upper respiratory tract samples is a feasible way to increase overall testing rate in office-based settings. Fifth, the decision tree is based on a nonobjective clinical criterion (the presence or absence of crackles). The Japanese Respiratory Society guidelines also use chest auscultatory findings. Moreover, crackles have been reported to have fair to moderate inter-observer agreement (Fleiss’ kappa/intraclass correlation coefficient = 0.4–0.6) to diagnose CAP [41]. Our very simple three-item clinical decision criteria can predict atypical pathogens with CAP, and we suggest it may be used easily in the clinical practices, especially in primary care.

Conclusions

This is the first prospective multicenter study to develop a decision tree to predict atypical pathogens with CAP confirmed by a molecular method. After wider validation in larger studies, our simple clinical decision rules could be useful in identifying primary care patients with CAP that are at risk for atypical pathogens.