Etiology of community-acquired pneumonia and diagnostic yields of microbiological methods: a 3-year prospective study in Norway

Community-acquired pneumonia (CAP) is a common disease and a significant cause of morbidity and mortality worldwide [1-3]. Differences in epidemiology of pathogens make the knowledge of local etiology crucial for the appropriate choice of empirical antimicrobial treatment, which has a major impact on the prognosis of the patient [4-7]. An important rationale for microbial testing is to enable pathogen-directed therapy, and thus avoid unnecessary antibiotic use [8]. In clinical practice causative pathogens often remain unknown, and the majority of patients are treated empirically [8]. Moreover, published data on etiology of CAP in Norway [9], as well as in several other countries, are hampered by old data, limited samples and detection methods confined to bacterial culture.

The use of diagnostic methods based on polymerase chain reaction (PCR) have resulted in increased detection of many bacterial and viral pathogens associated with CAP [8]. One of the advantages of PCR is its ability to identify respiratory pathogens also after initiation of antibiotic therapy [10]. Some recent studies have shown that a high microbial yield can be achieved when real-time PCR assays are combined with conventional diagnostic methods like bacterial cultures, urinary antigen assays and serology [11-16]. Such extensive approaches have led to recognition of viruses as important causes of CAP and that coinfections are more common than previously thought [17,18].

The aim of this study was to describe etiology of CAP in adults admitted to a general hospital in South-Eastern Norway using conventional methods and real-time PCR assays for the detection of respiratory pathogens. We also investigated the diagnostic yields of the methods applied, especially the potential benefit of PCR.

This study is the first to describe etiology of CAP in hospitalized adults in Norway using conventional methods and PCR. The main results were: Firstly, etiology was established in the majority of patients, especially when samples were complete and no antibiotics were prescribed prior to hospital admission. Furthermore, PCR improved diagnostic yield, particularly in diagnosing viral infections. Secondly, S. pneumoniae and respiratory viruses (mainly influenza viruses and rhinovirus) were most commonly detected, usually with copathogens. Viruses were detected more frequently during winter and spring, which was also the time period they occurred most commonly in combination with bacteria. Thirdly, OP swabs gave more positive results than NP swabs for qPCR detection of S. pneumoniae; and serology gave more positive results than real-time PCR for detection of B. pertussis and influenza viruses.

Establishing a microbiological diagnosis for patients with CAP is challenging. As is the case in any study of samples not obtained through invasive techniques and with quantitation, the link between presence and causal is still quite unclear. However, in this study we applied a vast array of available tests to determine the presence of known pathogens, including those only identified by real-time PCR, following recommendations for interpretation applicable to our region. The diagnostic yield of 63% is consistent with results of other studies, although reports vary considerably from 39% to 76% [11-16,27-31]. These variations may be attributable to differences in the epidemiology of pathogens, study population, diagnostic methods and available patient specimens. The high diagnostic yield that was achieved in the complete sampling group underscores that etiology, nowadays, can be established in the majority of CAP patients if multiple techniques are applied and sample collection is optimized.

The prevalence of S. pneumoniae, other “typical” bacteria and Legionella pneumophila, which almost exclusively occurred in “imported” cases, are in line with previous studies from Northern-Europe [7,11] and USA [32]. C. pneumoniae was expectedly uncommon [7,11]. The relatively high frequency of B. pertussis is probably explained by an ongoing epidemic [33], and the low frequency of M. pneumoniae due to timing; the study was conducted between two 5–7 years epidemic cycles [34].

The potential role of viruses in causing pneumonia, both as the sole pathogen as well as in association with bacteria, is debated. Furthermore, the validity of upper-airway samples to establish an etiologic diagnosis is controversial. However, the impacts of viruses per se on adult CAP are increasingly being recognized [35-38]. In line with other studies [11,28,29], we identified viruses by all methods in 34% of all cases and in 41% of cases with complete sampling. In contrast, conventional methods identified only 14% of all cases, which confirms the increased sensitivity of PCR reported by others [11,16,39]. As expected, the occurrence of adult CAP with viral findings varied considerably during the year, peaking in March at ~70%, and was significantly higher during winter and spring compared to summer and fall. Influenza viruses were most frequently detected; 14% by serology compared to 6% by PCR, the difference between the two methods being statistically significant—and no more than 2 cases of influenza A (H1N1) virus infection were detected by PCR—illustrating the substantial value of serology in epidemiological studies. However, serology is generally not recommended for clinical decision making due to low specificity, which in our study may have led to an overestimation of numbers since a probable diagnosis was not only based on seroconversion but for some patients on high titer in single sample. Interestingly there was no increase in CAP admissions due to influenza viruses during the 2009 H1N1 pandemic, indicating that influenza viruses are major pathogens regularly occurring in patients hospitalized with CAP irrespective of the H1N1 pandemic. The low diagnostic yield of influenza viruses by PCR technique also underscores the clinical problem with false negative PCR influenza tests, and may suggest that sampling from both the upper and lower respiratory tract should be performed, at least in severely affected individuals [40]. Another explanation may be that the PCR influenza test is more likely to be negative later in the course of a secondary pneumonia to influenza virus infection leading to hospitalization.

Coinfections have previously been reported to account for 4% to 30% of adult CAP [28,30]. In our study, more than one pathogen was found in 26% of all patients and in 34% of patients with complete sampling. In agreement with recent studies, S. pneumoniae was the most common bacterial agent associated with coinfections, and a copathogen was found in more than one-half of the cases with the majority of pathogens being viral agents [11,12,14,41], usually influenza viruses or rhinovirus. One of our important observations, applying comprehensive and complete diagnostics, was an exceptionally high detection rate of viral–bacterial coinfections in the course of the winter season, peaking at 60% in January and February. This number is considerably higher than the 39% recently reported [40], indicating that seasonal activity patterns of respiratory pathogens also impacts on the occurrence of adult CAP with viral–bacterial coinfections, which appear to be the dominant cause of pneumonia during those times.

Good-quality sputum culture is often difficult to obtain and reported yields are varying [1]. In our study, sputum samples were obtained from 62% of patients; of those obtained were 56% judged as inadequate for culture, and of those cultured revealed 21% a positive result. NP swab cultures were obtained from 99% of patients; only 13% of these revealed a positive result. Our concordance analyses indicate, however, that the two methods are complementary.

PCR improved the yield by 8–15% when combined with conventional methods, proving that PCR is a valuable tool in addition to conventional methods. However, since serological tests have little clinical impact in the acute situation, our findings underscore the need for further improvements. Our results, using flocked swabs, indicate quite equal effectiveness of NP and OP sampling for real-time PCR detection of viruses and the concordance rate between the two specimens was good (~90%). These results are in contrast to those from two previous reports [29,42] that observed superiority of NP sampling, one of the studies using flocked swabs [42]. To our knowledge, no study comparing NP and OP sampling for qPCR detection of S. pneumoniae in adult CAP has yet been published: In line with the findings of Principi et al. [26], we found that OP swabs were significantly more effective than NP swabs for qPCR detection of S. pneumoniae. However, it is not known whether these might actually represent false-positives as, according to our analyses, OP swabs showed poorer specificity (90.2%) compared to NP swabs (96.8%) (Additional file 1). Furthermore, NP swabs were not assessed by direct fluorescent antigen for specimen adequacy prior to testing. However, excellent specimen adequacy of NP flocked swabs has recently been reported [43]. Nevertheless, due to the lack of diagnostic gold standards, accuracy of such tests is difficult to assess and more research is needed to confirm this finding.

Certain limitations concern data collection and analyses. Firstly, the requirement of rectal temperature >38.0°C may exclude predominantly elderly patients with CAP, who remain afebrile, and hence contribute to explain the low number of nursing home residents in the study. Secondly, some patients who met the inclusion criteria were not enrolled in the study and the complete microbial sampling could not be applied in every patient. We have, however, no reason to suspect these factors to induce systematic selection bias. Thirdly, not all sputum and NP culture samples were taken before administration of antibiotics, which influences the validity of the microbial etiology patterns found and the comparison of diagnostic tests. Fourthly, specificity issues have been raised regarding the ply gene used, which has been identified in Streptococcus mitis. However, we used quantitative technique, and any DNA amount of S. mitis would probably have been low, at least in NP samples. Furthermore, we did not include a control group to determine the prevalence of adult pneumococcal nasopharyngeal carriage. The frequency of carriage in Norway hardly differs appreciably from the 4% reported from Sweden [44]. Also, we used a C_q cut-off value of 10⁵ cfu/mL to distinguish colonization from true infection [25]. Because we did not include control subjects we were unable to investigate the significance of the proportion of certain viruses, in particular rhinovirus, among patients with CAP. Finally, we included 48 (18%) patients with immunosuppression (i.e., primary or acquired immunodeficiency, active malignancy, immunosuppressive drugs) that may have an impact on the etiological results. However, this proportion of immunocompromized patients in our study does not exceed 25% that probably would have hampered comparison of etiological results with other studies from which such patients have been excluded [7]. Moreover, inclusion of these patients may also be a strength of the study as they better reflect the population being referred to this local hospital. These limitations have been considered in the analyses and presentation of results.

In conclusion, it was possible to establish etiology in 4 out of 5 CAP patients with the aid of PCR, which was particularly useful in diagnosing viral infections. The main pathogens identified were S. pneumoniae and viruses, most often influenza viruses and rhinovirus, usually detected with other agents. Viruses were more frequently detected during winter and spring, which was also the time period they occurred most commonly in combination with bacteria. Further research is needed to assess whether OP is superior to NP swabbing for qPCR detection of S. pneumonia in adults with CAP.