- Top of page
- SUMMARY AT A GLANCE
Background and objective: Agents such as Mycoplasma pneumoniae, Chlamydophila pneumoniae and Legionella pneumophila are recognized as important causes of community-acquired pneumonia (CAP) worldwide. This study examined the role of these ‘atypical pathogens’ (AP) among adult hospitalized patients with CAP.
Methods: A prospective, observational study of consecutive adult CAP (clinico-radiological diagnosis) patients hospitalized during 2004–2005 was conducted. Causal organisms were determined using cultures, antigen testing and paired serology. Clinical/laboratory/radiological variables and outcomes were compared between different aetiologies, and a clinical prediction rule for AP was constructed.
Results: There were 1193 patients studied (mean age 70.8 ± 18.0 years, men 59.3%). Causal organisms were identified in 468 (39.2%) patients: ‘bacterial’ (48.7%), ‘viral’ (26.9%), ‘AP’ (28.6%). The AP infections comprised Mycoplasma or Chlamydophila pneumoniae (97.8%) and co-infection with bacteria/virus (30.6%). The majority of AP infections involved elderly patients (63.4%) with comorbidities (41.8%), and more than one-third of patients were classified as ‘intermediate’ or ‘high’ risk CAP on presentation (pneumonia severity index IV–V (35.1%); CURB-65 2–5 (42.5%)). Patients with AP infections had disease severities and outcomes similar to patients with CAP due to other organisms (oxygen therapy 29.1% vs 29.8%; non-invasive ventilation 3.7% vs 3.3%; admission to the intensive care unit 4.5% vs 2.7%; length of hospitalization 6 day vs 7 day; 30-day mortality: 2.2% vs 6.0%; overall P > 0.05). 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 × 109/L and Hb < 11 g/dL were features associated with AP infection, but the derived prediction rule failed to reliably discriminate CAP caused by AP from bacterial CAP (area under the curve 0.75).
Conclusions: M. pneumoniae and C. pneumoniae as single/co-pathogens are important causes of severe pneumonia among older adults. No reliable clinical indicators exist, so empirical antibiotic coverage for hospitalized CAP patients may need to be considered.
- Top of page
- SUMMARY AT A GLANCE
Agents such as Mycoplasma pneumoniae, Chlamydophila pneumoniae and Legionella pneumophila are recognized as important causes of community-acquired pneumonia (CAP) worldwide.1,2 These organisms, frequently referred as ‘atypical pathogens’ (AP), do not respond to beta-lactam therapy alone. While M. pneumoniae and C. pneumoniae are often implicated as the causes of mild CAP in young ambulatory patients,3,4 their roles among the older and sicker hospitalized patients remain uncertain, particularly in the Asia–Pacific region.3,5 In certain regions including Hong Kong, empirical β-lactam mono-therapy is recommended, while the use of a macrolide is optional.6 However, recent studies have suggested that treatment of AP may be associated with faster clinical recovery and lower mortality.2
This study reports the epidemiology, disease severity and clinical outcomes of patients with AP infection in a large cohort of adults hospitalized with CAP. The study investigated whether there are any useful clinical, laboratory and radiological indicators of AP infection, and developed a score-based clinical prediction rule to assist clinicians to decide on empirical antibiotic coverage for AP.
- Top of page
- SUMMARY AT A GLANCE
The study analysed 1193 patients who completed the protocol. Patients were mostly elderly (73.4% were >65 years of age), 49.2% had an underlying comorbid illness and at presentation their pneumonia was frequently classified as ‘intermediate’ and ‘high’ risk (PSI class IV–V in 47.3%; CURB-65 score 2–5 in 50.8%) (Table 1).7 Aetiologic diagnoses were established in 468 (39.2%) patients (‘aetiology uncertain’, n = 725). Among these, 134 (28.6%) patients had AP infections, including M. pneumoniae (n = 78), C. pneumoniae (n = 55), L. pneumophila (n = 1) and C. burnetii (n = 2). Two patients had dual mycoplasma/chlamydophila infections. Co-infections were observed in 41 (30.6%) of the 134 patients with AP infections, with concomitant pathogens being bacterial (n = 31), viral (n = 8) or another AP (n = 2). Mycoplasma or Chlamydophila pneumoniae (28.0%), influenza virus (21.8%), S. pneumoniae (21.6%) and H. influenzae (13.2%) together were responsible for pneumonia in 85% of CAP patients in whom the infectious agent was identified.
Table 1. Clinical characteristics and aetiologic diagnoses of community-acquired pneumonia (CAP) in 1193 patients
|Clinical characteristics||Whole cohort n = 1193||Aetiologies established n = 468|
|Age (mean ± SD)||70.8 ± 18.0||69.6 ± 18.6|
|Nursing home resident||22.6||17.7|
|PSI class IV-V||47.3||42.1|
|CURB-65 score 2–5||50.8||50.0|
|Supplemental O2 therapy†||29.0||29.6|
|LOS in hospital (median, IQR)||7 (5–10)||7 (5–10)|
|Aetiological agents, by patient (%)|
|Atypical pathogens, AP (n = 134)‡||11.2||28.6|
| Mycoplasma pneumoniae (n = 78)||6.5||16.7|
| Chlamydophila pneumoniae (n = 55)||4.6||11.8|
| Coxiella burnettii (n = 2)||0.2||0.4|
| Legionella pneumophila (n = 1)||0.1||0.2|
| Mixed AP and bacteria (n = 31)||2.6||6.6|
| Mixed AP and virus (n = 8)||0.7||1.7|
|Bacterial pathogens (n = 228)||19.1||48.7|
| Streptococcus pneumoniae (n = 101)||8.5||21.6|
| Haemophilus influenzae (n = 62)||5.2||13.2|
| Others (n = 82)||6.9||17.5|
|Viral pathogens (n = 126)||10.6||26.9|
| Influenza virus (n = 102)||8.5||21.8|
| Others (n = 24)||2.0||5.1|
|Mycobacterium. Tuberculosis (n = 38)||3.2||8.1|
Of the 134 patients with AP infections, the majority (63.4%) were aged >65 years (with >16% being nursing home residents) and 35.1% and 42.5% were classified at presentation as ‘intermediate’ and ‘high’ risks according to the PSI (class IV–V) and CURB-65 (score 2–5) rules respectively. Compared with patients with other causes of pneumonia (Table 2), patients with AP infection had comparable disease severity and clinical outcomes, including the need for oxygen therapy, non-invasive ventilation, ICU admission and total LOS (P > 0.05). The 30-day mortality tended to be lower in these patients (n = 3, 2.2%), but this reduction was not statistically significant (P = 0.09).
Table 2. Clinical characteristics and outcomes of 134 patients infected with atypical pathogens, compared with patients with other confirmed aetiologies of community-acquired pneumonia, and statistical significance of the difference
|Characteristics||Atypical pathogen† (n = 134)||Other aetiologies (n = 334)||P value|
|Age (mean ± SD)||66.4 ± 20.6||70.9 ± 17.7||0.027|
|Nursing home residents||16.4||18.3||0.637|
|PSI class IV-V||35.1||44.9||0.051|
|CURB-65 score 2–5||42.5||53.0||0.041|
|Supplemental O2 therapy||29.1||29.8||0.878|
|LOS in hospital, median (IQR)||6 (5–10)||7 (5–10)§||0.877|
Patient demographics, clinical, routine laboratory and radiological variables were compared between patients with AP (n = 95) and bacterial (n = 183) mono-infections (Table 3). Patients with mixed AP and bacterial/viral infections were excluded from these comparisons. When compared with bacterial pneumonia, age <65 years (OR 2.5, 95% CI: 1.4–4.6; P = 0.003), female gender (OR 1.9, 95% CI: 1.1–3.4; P = 0.024), fever ≥38.0°C (OR 2.7, 95% CI: 1.5–4.9; P = 0.001), respiratory rate <25/min (OR 1.9, 95% CI: 1.0–3.6; P = 0.039), pulse rate <100/min (OR 1.8, 95% CI: 1.0–3.3; P = 0.047), absence of hyponatraemia with serum sodium >130 mmol/L (OR 12.5, 95% CI: 1.5–100.9; P = 0.018), leucocyte count <11 × 109/L (OR 2.0, 95% CI: 1.1–3.5; P = 0.021) and Hb < 11 g/dL (OR 2.2, 95% CI: 1.1–4.5; P = 0.024) at presentation were independent indicators of AP infection as shown by multivariate analysis. However, a score-based prediction rule derived from these variables could not reliably discriminate patients with pneumonia caused by AP from patients with bacterial pneumonia (AUC = 0.75, 95% CI: 0.69–0.81) (Fig. 1). No specific cut-off value could be identified that provided reasonable sensitivity or specificity. For instance, to obtain a specificity of 80%, the cut-off value provided a sensitivity of only 54%. When compared with viral pneumonia, only respiratory rate <25/min (OR 2.7, 95% CI: 1.3–5.9; P = 0.009) and absence of runny nose (OR 2.5, 95% CI: 1.1–5.7; P = 0.025), but no laboratory variables, were independently associated with AP infection. Comparison of patients with AP mono-infection to patients with bacterial or viral pneumonia, respectively, showed no significant differences in PSI class, CURB-65 score, requirement for oxygen therapy, non-invasive ventilation and ICU care, and hospital LOS even when mixed infections were excluded from these comparisons (all P > 0.05) (see Table 2 footnotes). The only exception was a lower all-cause 30-day mortality (AP 1.1% vs bacterial 6.6%; P = 0.04) in patients with AP mono-infection compared with patients with bacterial infection.
Table 3. Clinical, laboratory and radiological features of patients with atypical pneumonia (AP) at presentation compared with patients with bacterial pneumonia
|Clinical variables||AP pneumonia† (n = 95)||Bacterial pneumonia (n = 183)||P value|
|Age >65 years||57.9||74.9||0.004|
|Nursing home resident||20.0||19.1||NS|
|Onset >3 days before admission||43.2||36.1||NS|
|Systolic BP, mm Hg (mean ± SD)||140.1 ± 24.1||144.0 ± 31.3||NS|
|Tachycardia (pulse rate ≥100)||48.4||60.1||0.063|
|Tachypnoea (resp. rate ≥25)||24.2||42.6||0.002|
|Glasgow Coma score <15||9.5||12.0||NS|
|Radiological changes (%)|| || || |
| Airspace consolidation||83.7||82.6||NS|
| Reticular or ground-glass||16.3||17.4||NS|
| Pleural effusion||19.6||11.3||NS|
|Laboratory variables (mean ± SD)|| || || |
| Sodium (mmol/L)||136.3 ± 3.4||134.9 ± 5.3||0.007|
| Urea (mmol/L)||6.9 ± 4.6||7.7 ± 4.0||NS|
| Hb (g/dL)||11.9 ± 1.9||12.6 ± 1.7||0.004|
| Platelet (×109/L)||248 ± 91||238 ± 107||NS|
| WCC (×109/L)||11.6 ± 5.7||13.9 ± 6.5||0.006|
| Bilirubin (µmol/L)||15.7 ± 13.6||16.6 ± 17.0||NS|
| Alanine aminotransferase (IU/L)||43.9 ± 54.0||36.9 ± 51.2||NS|
Figure 1. Receiver operator characteristic (ROC) curve for discriminating pneumonia caused by atypical pathogens from bacterial pneumonia, using a score-based clinical prediction rule. Area under ROC curve (AUC) = 0.75; 95% CI: 0.69–0.81. In general, the discriminatory ability of AUC can be interpreted as follows: AUC ≥ 0.90 as ‘excellent’, 0.80 to <0.90 ‘good’, 0.70 to <0.80 ‘fair’ and <0.70 ‘poor’. Intersection of the two lines indicates the corresponding sensitivity (54%) and specificity (80%) to diagnose pneumonia due to atypical pathogens (refer to text).
Download figure to PowerPoint
- Top of page
- SUMMARY AT A GLANCE
This is one of the largest Asian studies to report on the clinical significance and outcomes of adults hospitalized with CAP due to AP. M. pneumoniae and C. pneumoniae, as single or co-pathogens, can cause severe diseases in older patients, resulting in hospitalization. Our results concur with data from other Asia–Pacific studies, which showed that mycoplasma and chlamydophila are responsible for 9–12% and 6–13% of CAP respectively,2,14,23–26 whereas L. pneumophila infection remains uncommon (∼5%).2,23,24,26
In contrast to some early studies which described mild or ‘walking’ pneumonia due to APs in children and younger adults,3,4 our results and more recent reports have shown that these organisms are important causes of CAP in elderly community-dwellers and nursing home residents.27–31 Very often these patients have comorbid medical conditions; and many of them would be sick enough to warrant hospitalization.2,5,23,31–34 Some may even require intensive care because of severe sepsis and/or respiratory failure.8,23,33–35 In fact, based on the clinical prediction rules of PSI and CURB-65, >35–40% of the patients with AP infection in our cohort were classified as ‘intermediate’ and ‘high’ risk on presentation. About 30% of these patients had required oxygen therapy, >8% received ventilatory support (non-invasive ventilation in ward/mechanical ventilation in ICU), and the total hospital LOS was 5–10 days. These outcomes were generally comparable with CAP patients with aetiologies other than AP.28,32,34,36 Although there was a lower overall mortality in patients with AP infection, the reduction was statistically insignificant. Co-infection (with a bacteria/virus), which occurred commonly (>30%), might have contributed to the overall disease severity and outcome.5,9,24,27,33,34,36 We observed that two out of the three deaths in the AP infection group were patients who had a concomitant bacterial pathogen isolated. Either as a sole or a co-pathogen, our results suggested that AP infection can be associated with severe CAP in older adults, and their treatment may need to be considered. This suggestion is supported by a recent analysis which showed that treating hospitalized CAP patients with dual β-lactam and macrolide therapies is associated with decreased time to clinical stability, decreased LOS and reduced mortality.2
Laboratory diagnosis of acute M. pneumoniae and C. pneumoniae infections remains difficult.9,37 Therefore we attempted to develop a ‘clinical prediction rule’, based on the results of routine assessments, to assist clinicians in identifying AP infections and to indicate the requirement for treatment other than a beta-lactam (such as a macrolide). In other studies/regions, prediction rules for AP infection in adult CAP have provided sensitivities ranging from 48% to 77%, and specificities 89–93%.5,38,39 Consistent with other reports, younger age, high fever, normal WCC, mild anaemia, absence of significant hyponatraemia and lack of tachypnoea/tachycardia were found to be indicators of AP infection.5,36,38 Despite its ‘fair’ discriminatory ability (AUC 0.75), our prediction rule did not seem to provide a cut-off point with reasonable sensitivity and specificity for AP infection. It is possible that the heterogeneous clinical features of the various ‘atypical pathogens’, and their different relative compositions in each studied cohort (e.g. L. pneumophila) limited the generalizability of these prediction rules.5,39 Further studies to include certain surrogate markers may be warranted.5
Because reliable clinical indictors are lacking, our data support the latest recommendations from the Infectious Diseases Society of America and the American Thoracic Society to use an empirical antibiotic regime that covers APs in all hospitalized CAP patients in our region. In addition to an anti-pneumococcal β-lactam, a macrolide (or doxycycline) should be considered; respiratory fluoroquinolone is another option. Recent analyses (which included studies from the Asia–Pacific region) show that lack of specific treatment for AP may be associated with poorer clinical outcomes.2,14,40,41 However, well-conducted randomized controlled trials (e.g. β-lactam vs β-lactam + macrolide) are necessary to resolve this issue,2,42 and the diagnosis of viral (influenza) infections in the remaining 20–30% of hospitalized CAP patients may help to reduce antibiotic prescriptions.43,44
Our study is limited by the fact that acute infections caused by M. pneumoniae and C. pneumoniae were diagnosed based on serological assays, and not culture or PCR;1,9,37 and urinary antigen tests for legionella were performed in only 200 patients.26 Cultures, although specific, are insensitive; and PCR assays for these infections have not been standardized/validated.1,9,37 To avoid exaggeration, we have adopted the more stringent criteria for serological diagnosis; patients with results indicating a ‘probable’/‘presumptive’ aetiology (e.g. a single, moderately high titre) were classified as ‘aetiology uncertain’ and not included in the detailed analysis.5,14,15,23 The sensitivities and specificities of the serology assays used to detect M. pneumoniae infection (complement fixation; IgM and IgG EIA, PLATELIA) have been shown to be reasonably accurate when performed on adult acute/convalescent-phase sera samples collected 2 weeks apart (AUC 0.87–0.94; PCR assay as the ‘gold standard’).18,19,26 Although the serological diagnosis of acute C. pneumoniae was not based on the MIF test, the SeroCP-EIA (Savyon)(using preserved specific LPS of C. pneumoniae as antigens), when evaluated against the MIF ‘gold standard’, gave a sensitivity, specificity and an overall agreement of >90% (n = 240).20,21 Since the MIF test lacks standardization (e.g. use of different strains), and studies have shown poor inter-observer consistency on the interpretation of endpoints, a well-evaluated commercial EIA method was chosen for this study.14,26,37 Finally, it is possible that patients who had died early during admission would not have had the convalescent-phase sera available for study. Together with the more stringent diagnostic criteria adopted, the prevalence and severity of APs in adult CAP might have been underestimated.
In conclusion, M. pneumoniae and C. pneumoniae as single/co-pathogens are important causes of severe pneumonia in older adults. As reliable clinical indicators are lacking, their empirical coverage among hospitalized CAP patients may need to be considered.