Introduction
Community-acquired pneumonia (CAP) as the most common cause of death due to infection in adults brings a great clinical and economic burden to the health care systems worldwide [1,2]. The definition of the etiology directly affects the decision concerning the clinical treatment of patients with CAP, which will finally influence the outcome for patients with CAP [3,4]. The bacterial causes of CAP have been well characterized, with Streptococcus pneumoniae being the most important bacterial pathogen [5,6]. In contrast to bacterial agents, the involvement of respiratory viral infections in adult patients with CAP has yet not been well defined.
With the development of virological diagnostic techniques, an increasing number of studies evaluating the incidence of respiratory viral infections in adult patients with CAP indicates the important contributions of respiratory viruses to CAP [7,8]. However, up to now, no synthetic data has clearly demonstrated the incidence of overall respiratory viral infections in adult patients with CAP and the discrepancy in the contributions to the pathogenesis of CAP across respiratory viral species.
The aim of this meta-analysis is to determine the incidence of viral infections in adult patients with CAP and demonstrate the incidence discrepancy among respiratory viral species according to previous epidemiology researches based on PCR-related techniques.
Methods
Search Strategy
In July 2013, we conducted a systematic search utilizing PubMed and Embase search engines for citations published from January 1, 2000 to July 19, 2013. The initial search was undertaken using free-word, keyword and MeSH searches for ‘community-acquired pneumonia', ‘virus', ‘adult' and ‘PCR'. Two researchers (X.W. and Q.W.) independently searched the titles and abstracts identified initially on screen for the selection of the potential studies. We also searched the reference lists of the included studies. When needed, we contacted the authors to obtain missing or additional information. We settled any discrepancies through discussion with a third researcher (M.W.).
Inclusion/Exclusion Criteria
Studies had to meet the following criteria for inclusion: (1) patients were ≥16 years old; (2) studies clearly stated the definition of CAP; (3) studies provided the incidence of respiratory viral infections in CAP patients; (4) respiratory viruses were detected by highly sensitive techniques (PCR and real-time PCR); (5) at least 90% of the samples collected were used for viral detection; (6) the full paper was available in the English literature.
Studies were excluded if they focused on patients in an immunosuppressed state, such as in the case of patients with solid organ or bone marrow transplantations, AIDS or receiving chemotherapy or taking other immunosuppressive drugs etc.
Quality Assessment and Data Extraction
Two reviewers independently assessed the quality of all included studies using score quality criteria based on the principles of the quality assessment of observational studies (as shown in online suppl. table 1; see www.karger.com/doi/10.1159/000369561 for all online suppl. material). The relevant trail characteristics were independently extracted by 2 reviewers (X.W. and Q.W.), and any discrepancies were solved with complete agreement through rechecking the source papers and discussion. Data extraction for each study was done according to a standardized form designed by us to capture and record all relevant information required for analysis. For each study, 2 reviewers extracted the data independently, cross-checked with each other and resolved disagreements through discussion. For all included studies, the following information was recorded: author, year of publication, country, number of patients, number of respiratory virus species detected, viral detection method, specimen collected for viral detection, number of patients with viral infections and number of patients with mixed infections by viruses and other pathogens.
Statistical Analysis
According to the expected heterogeneity across studies, the combined incidence of respiratory viral infections with 95% was calculated with a DerSimonian-Laird random-effects model [9]. Statistical heterogeneity was tested with Cochran's χ2 and quantified with the I2 statistic (30-60% for moderate heterogeneity; 50-90% for substantial heterogeneity; 75-100% for considerable heterogeneity) [10]. In order to reduce the heterogeneity across studies and perform further analysis, a subgroup analysis was required. We further investigated potential sources of heterogeneity by metaregression analysis. Factors examined individually in univariable models were the sample type used for viral detection (by comparing a lower respiratory tract specimen with an upper respiratory tract specimen), time span of patient enrollment, geographical region, number of viral species detected and number of viral detection methods. We did all analyses in Stata (version 12.1, StataCorp LP, College Station, Tex., USA) with the commands ‘metan' (for random-effects meta-analysis), ‘metareg' (for metaregression) and ‘metainf' (for sensitivity analysis).
Results
A total of 768 records were returned during our searches. After initial screening and removal of duplicates, we reviewed 76 papers in full text. Finally, 23 studies (n = 6,404) published between February 2001 and July 19, 2013 were included as eligible reports for our further analysis (fig. 1, table 1). The quality score of those 23 reports is listed in table 1. Ten of the 23 reports achieved a score ≥7, and the other 13 reports scored between 4 and 6. Of the 23 reports, 2 were conducted in adult outpatients with CAP [11,12], 17 in adult inpatients with CAP [7,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28] and 4 in both outpatients and inpatients with CAP [29,30,31,32]. Nine reports were carried out in Europe (4 in Spain, 3 in the Netherlands, 1 in the UK and 1 in Sweden) [7,14,17,18,20,24,28,29,30], 5 in Southeast Asia (2 in China, 1 in Japan, 1 in Korea and 1 in Vietnam) [11,12,15,19,32], 3 in Australia [22,26,27], 4 in America (2 in the USA, 1 in Canada and 1 in Chile) [13,16,25,31] and 2 in Israel [21,23]. Moreover, respiratory viral detection in 12 studies was solely based on PRC/real-time PCR [7,11,12,13,14,15,16,17,18,19,23,27,29], and viral detection in the other studies was based on a combination of PCR/real-time PCR and conventional methods [20,21,22,24,25,26,28,30,31,32].
Incidence estimates of respiratory viruses in patients with CAP ranged from 8.6 to 56.2% (fig. 2); heterogeneity was considerable (χ2 = 269.9, p < 0.0001, I2 = 91.8%). The random-effects combined incidence was 22.4% (95% CI = 19.0-25.7, I2 = 91.8%). Considering the high level of heterogeneity, we further conducted a subgroup analysis for the separate calculation of combined incidences according to the types of patients (inpatients, outpatients and mixed patients), time span of patient enrollment (1 year vs. >1 year), economic level of the countries (developed vs. developing countries), PCR methods (routine PCR vs. real-time PCR) or according to the region where each report was conducted (Europe, Southeast Asia, Australia, America and the Middle East), as shown in table 2. The combined incidence of respiratory viruses was 10.2% (95% CI = 7.3-13.1, I2 = 0%) in outpatients [11,12], 22.7% (95% = 19.0-26.4, I2 = 90.8%) in inpatients [7,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,31,32] and 27.7% (95% = 16.4-39.0, I2 = 94.7%) in mixed patients [29,30]. In the subgroup analysis according to regions, the combined incidence was highest in the Middle East (32%, 95% CI = 27-38, I2 = 0%) [21,23] and lowest in Southeast Asia (16.6%, 95% CI = 10.5-22.8, I2 = 85.1%) [11,12,15,19,32]. The combined incidence was 24.3% (95% CI = 20.3-28.3, I2 = 92.7%) and 15.8% (95% CI = 10.4-21.3, I2 = 84.2%) in developed countries and developing countries, respectively. In the subgroup analysis stratified by the time span of the studies, the combined incidence of respiratory viruses in the studies covering 1 year and >1 year was 23.6% (95% CI = 22.1-25.1, I2 = 92.6%) and 24.3% (95% CI = 22.5-26.2, I2 = 96.9%), respectively. In addition, considering that real-time PCR is more sensitive than routine PCR, which might lead to the significant heterogeneity in our study, we further performed a subgroup analysis stratified by PCR methods (routine PCR vs. real-time PCR). The combined incidence of respiratory viruses was 20.9% (95% CI = 17.1-24.6, I2 = 89.2%) in studies using routine PCR and 24.7% (95% CI = 18.1-31.3, I2 = 94.3%) in studies using real-time PCR.
Table 2
Combined incidence estimates of viral infections in adult patients with CAP by geographical region
Considering that the subgroup analysis cannot explain all sources of heterogeneity among the incidence of respiratory viruses in patients with CAP, we further carried out a univariate metaregression analysis (table 3) and found that the viral incidence was higher in studies in which lower respiratory tract specimens were collected for viral detection (p = 0.048) than in those in which only upper respiratory tract specimens were collected for viral detection. Additionally, we did a sensitivity analysis and found that the study by Cilloniz et al. [20] might have contributed to the heterogeneity. However, the recombined incidence of respiratory viruses after omission of that study was 23.0% (95% CI = 19.7-26.3, I2 = 90.5%), which was similar to the combined incidence before omission.
Table 3
Univariate metaregression for the incidence of viral infections and viral infections mixed with other pathogens in adult patients with CAP
Of the 23 reports, 18 studies [7,12,14,15,16,17,18,19,20,21,22,24,26,27,28,29,30,31,32] provided data for the calculation of the coinfection incidences of respiratory viruses with other pathogens (n = 5,058). As shown in figure 3, the incidence estimates of coinfections ranged from 3 to 28%; heterogeneity was considerable (χ2 = 200.9, p < 0.0001, I2 = 91.5%). The combined overall incidence of coinfections of respiratory viruses with other pathogens (mainly bacteria) was 12.4% (95% CI = 9.7-15.0, I2 = 91.5%). Due to the high level of heterogeneity, the subgroup analysis was performed according to the patient types, economic level of the countries and regions where the studies were performed, as mentioned above. The combined incidence of coinfections was 12.9% (95% CI = 9.7-16.1, I2 = 91.3%) in inpatients with CAP [7,14,15,16,17,18,19,21,22,24,26,27,28] and 13.7% (95% CI = 7.1-20.3, I2 = 90.6%) in mixed patients with CAP [29,30,31,32]. Since only 1 study provided the incidence of coinfections in outpatients with CAP [12], a combined estimate was not performed. In the subgroup analysis stratified by the time span of the studies, the combined incidence of respiratory viruses in the studies covering 1 year and >1 year was 12.5% (95% CI = 9.2-15.7, I2 = 91.3%) and 12.4% (95% CI = 9.7-15.0, I2 = 94.1%), respectively. Additionally, the combined incidence of coinfections was 12.3% (95% CI = 9.3-15.3, I2 = 91.1%) and 12.4% (95% CI = 9.7-15.0, I2 = 93.6%) in developed countries and developing countries, respectively. Moreover, the combined incidence of coinfections was similar between Europe [7,14,17,18,24,28,29,30], Southeast Asia [12,15,19,32] and Australia [22,26,27] (12.9, 11.4 and 10.6%, respectively) as shown in table 4. In the individual variable metaregression analysis, none of the factors we explored further was significantly associated with heterogeneity (table 3). Moreover, in the sensitivity analysis, none of the studies were identified as contributing to the heterogeneity.
Table 4
Combined incidence estimates of viral infections mixed with other pathogens in adult patients with CAP by geographical region
Fig. 3
The combined incidence of viral infections mixed with other pathogens in adult patients with CAP.
Taking into account the discrepancy of incidence among respiratory virus species, we extracted the incidence of the 7 most common respiratory viral species for an individual combined incidence estimate. As shown in figure 4 and table 5, the 3 most frequently detected viruses were the influenza virus (8.9%, 95% CI = 7-11, I2 = 79.7%), rhinovirus (6.0%, 95% CI = 4-8, I2 = 87.4%) and coronavirus (4.7%, 95% CI = 3-7, I2 = 77.4%). However, the respiratory syncytial virus, metapneumovirus and adenovirus were less commonly detected.
Fig. 4
Discrepancies of the combined incidence across the common respiratory viral species. RSV = Respiratory syncytial virus.
Discussion
Our meta-analysis of the incidence of respiratory viral infections in adult patients with CAP identified 23 studies of 6,404 individuals. At present, there is a lack of quantitative syntheses of studies regarding the incidence of respiratory viral infections in adult patients with CAP worldwide. Our main finding is that respiratory viruses contribute to approximately 1/5 infections in adult patients presenting with CAP, highlighting their role in the pathogenesis of CAP. Furthermore, the influenza virus is the most frequent respiratory viral etiology in adult patients with CAP.
The incidence of viral infections in adult outpatients with CAP is lower than that in adult inpatients (10 vs. 22%) as demonstrated in the subgroup analysis, implying that viral infections could aggravate the condition of patients presenting with pneumonia. Moreover, the combined incidence of viral infections mixed with other pathogens (mainly bacteria) is 12% in our study, constituting more than half of the total viral infections in adult patients with CAP. Evidence from several studies illustrated that viral infections made hosts susceptible to secondary bacterial infections because mucosal barriers in the respiratory tracts were impaired during previous viral infections [33,34], which led to the bacterial invasions becoming more facile. There is evidence as well, although sparse, suggesting that mixed viral-bacterial infections would generate a severer inflammatory status and clinical presentation than individual bacterial or viral infections [24,26,35]. Severe fatal pneumonia can be caused by concomitant infection with the influenza virus and Staphylococcus aureus in both children and adults [36,37,38]. Moreover, in 1 pneumonia study focused on adults, rhinovirus-pneumococcal coinfections correlated with severe pneumonia and raised mortality [24,26,35]. Therefore, our results would be helpful for clinicians to better recognize the burden of viral infections and mixed viral-bacterial infections on adult patients presenting with CAP.
The influenza virus is the most frequent viral pathogen detected in adult patients with CAP as our study demonstrates, which is inconsistent with the fact that the respiratory syncytial virus is the leading viral pathogen in pediatric patients with CAP [39,40,41,42]. Influenza vaccination has been previously proven to be efficient in preventing influenza viral infections especially during a pandemic influenza period [43]. Furthermore, several studies which enrolled patients with a chronic lung disease indicated that an influenza vaccination significantly reduced the incidence of pneumonia or acute exacerbations in those patients [44]. Therefore, a regular and routine influenza vaccination of patients with chronic lung disease would be needed to lower the burden of pneumonia caused by the influenza virus. Moreover, newly developed anti-influenza virus drugs (neuraminidase inhibitors, such as oseltamivir and zanamivir), which have established roles in the early treatment of influenza infections through reducing the median time to the resolution of the symptoms and the risk of pneumonia [45], provided an efficient weapon for clinicians to manage patients with viral pneumonia caused by the influenza virus.
Respiratory viruses have been realized as a common etiology in patients with CAP, especially children with CAP. However, conventional virological diagnostic techniques such as serology, immunofluorescence and culture have been underestimating the contribution of viruses to the pathogenesis of CAP [46,47], especially in adult patients, due to their low sensitivity and narrow application. Owing to the development of more rapid and sensitive detection methods (PCR and real-time PCR), the viral diagnosis is largely improved, particularly through the combined application of these methods with conventional detection techniques [7]. Lots of previous studies evaluating the role of respiratory viruses in adult patients with CAP have been reported; nevertheless, there are discrepancies in the detection methods used for viral diagnosis across those studies. Therefore, we only included studies applying sensitive detection techniques (PCR and real-time PCR) for viral diagnosis in our further meta-analysis to reduce possible heterogeneity along with obtaining quantitative syntheses of original data.
Nevertheless, significant heterogeneity was still found during the meta-analysis and subgroup analysis. Our following heterogeneity analyses generated 1 potentially important finding, namely that lower respiratory tract specimens were associated with higher overall viral incidence. This may be because the molecular detection of the virus in lower respiratory tract specimens was more sensitive than the one in upper respiratory tract specimens. The potential implication of this result is that the presence of viral nucleic acids, detected by PCR of lower respiratory tract samples, implies stronger evidence of the infection by virus. However, no relation was identified between the incidence of coinfection and the lower respiratory tract specimens, which might be explainable. As we could speculate, discrepancies of both bacterial diagnostic and viral detection would lead to a significant heterogeneity of the incidence of coinfection. Caution is necessary when combined incidence estimates are used due to the significant heterogeneity, and further surveys are required for the standardization of the specimen collection, specimen quality, PCR primers and PCR reaction conditions during respiratory viral etiology diagnosis.
Our study had several limitations. First, we excluded many studies in the non-English literature and thus lost the raw data of those reports. Second, studies included in our meta-analysis were predominantly performed in high-income and middle-income regions. Therefore, fundamental incidence data for low-income countries are missing. Third, due to the lack of associated original data we were unable to explore the relationship between clinical severity of pneumonia and causative viral pathogens.
Thus, our study suggests that approximately one fifth of adult patients with CAP are infected by respiratory viruses, more than half of those viral infections are characterized as mixed infections with other pathogens, and influenza virus is the leading viral pathogen in adult patients with CAP, although there is significant heterogeneity during the quantitative synthesis of the raw data. Further surveys are required to establish a standardization of specimen management and detection processes for viral diagnostics.
Financial Disclosure and Conflicts of Interest
The authors have no support or funding to report. They declare that there are no competing interests.

Get Permission