Viral and Atypical Bacterial Detection in Young Nepalese Children Hospitalized with Severe Pneumonia

ABSTRACT Respiratory viruses cause a substantial proportion of respiratory tract infections in children but are underrecognized as a cause of severe pneumonia hospitalization in low-income settings. We employed 22 real-time PCR assays and retrospectively reanalyzed 610 nasopharyngeal aspirate specimens from children aged 2 to 35 months with severe pneumonia (WHO definition) admitted to Kanti Childrens’ Hospital in Kathmandu, Nepal, from January 2006 through June 2008. Previously, ≥1 of 7 viruses had been detected by multiplex reverse transcription-PCR in 30% (188/627) of cases. Reanalyzing the stored specimens, we detected ≥1 pathogens, including 18 respiratory viruses and 3 atypical bacteria, in 98.7% (602/610) of cases. Rhinovirus (RV) and respiratory syncytial virus (RSV) were the most common, detected in 318 (52.1%) and 299 (49%) cases, respectively, followed by adenovirus (AdV) (10.6%), human metapneumovirus (hMPV) (9.7%), parainfluenza virus type 3 (8.4%), and enterovirus (7.7%). The remaining pathogens were each detected in less than 5%. Mycoplasma pneumoniae was most common among the atypical bacteria (3.7%). Codetections were observed in 53.3% of cases. Single-virus detection was more common for hMPV (46%) and RSV (41%) than for RV (22%) and AdV (6%). The mean cycle threshold value for detection of each pathogen tended to be lower in single-pathogen detections than in codetections. This finding was significant for RSV, RV, and AdV. RSV outbreaks occurred at the end of the monsoon or during winter. An expanded diagnostic PCR panel substantially increased the detection of respiratory viruses in young Nepalese children hospitalized with severe pneumonia. IMPORTANCE Respiratory viruses are an important cause of respiratory tract infections in children but are underrecognized as a cause of pneumonia hospitalization in low-income settings. Previously, we detected at least one of seven respiratory viruses by PCR in 30% of young Nepalese children hospitalized with severe pneumonia over a period of 36 months. Using updated PCR assays detecting 21 different viruses and atypical bacteria, we reanalyzed 610 stored upper-respiratory specimens from these children. Respiratory viruses were detected in nearly all children hospitalized for pneumonia. RSV and rhinovirus were the predominant pathogens detected. Detection of two or more pathogens was observed in more than 50% of the pneumonia cases. Single-virus detection was more common for human metapneumovirus and RSV than for rhinovirus and adenovirus. The concentration of virus was higher (low cycle threshold [CT] value) for single detected pathogens, hinting at a high viral load as a marker of clinical significance.

all children were still breastfed. About a quarter of children were wasted (Z score of ,22 for weight-for-length/height) ( Table 1). Seventy percent of participants lived within the Kathmandu Valley (28).
Clinical characteristics. About half of the cases presented with fever (axillary temperature of .37.5°C), 82% had wheezing, and 91% had crepitations on auscultation. In addition to lower chest wall indrawing, half of the children also presented with additional severity signs, such as nasal flaring, grunting, or head nodding, and 19% had at least one of four danger signs (Table 1). Sixty-two percent were hypoxic (O 2 saturation of ,90%). Among the 584 who had their plasma C-reactive protein (CRP) concentration measured, the median (interquartile range [IQR]) CRP concentration was 20.1 (7.1 to 46.8) mg/liter; 175 (30%) had CRP concentrations of .40 mg/liter, and 79 (13.5%) had CRP concentrations of .80 mg/liter (Table 1). Radiographic pneumonia, defined as endpoint consolidation on chest X-ray, was detected in 166 of 459 (36.2%) chest radiographs available for interpretation. Selected clinical features by pathogen are shown in Table S1 in the supplemental material.
Respiratory pathogen detection. We reanalyzed 610 of the previously collected specimens, employing 22 PCR assays. The expanded panel included the seven respiratory viruses in the original study and an additional 14 pathogens. The retesting resulted in a pathogen being detected in 602 of 610 (98.7%) cases, compared to 188 of 627 (30%) cases in the on-location analyses in Nepal (Table 2).
Analyses for RSV, hMPV, parainfluenza virus types 1 to 3 (PIV-1 to -3), and IA and influenza B virus (IB) had already been performed on this population. The updated PCR assays resulted in 284 additional detections of these seven viruses (194 versus 478), while the 14 new PCR assays yielded 611 new detections. The updated assays failed to detect one case each of PIV-3 and hMPV, two cases of PIV-1, three cases of IA, four cases of RSV, and six cases of IB.
We observed that the mean cycle threshold (C T ) values were lower for the specimens that were positive both in the previous analyses and in the reanalysis compared to the mean C T values of specimens that were positive in the reanalysis only, with mean differences in C T values of 1.8 for RSV (P = 0.005), 3.4 for hMPV (P = 0.09), and 5.0 for PIV-3 (P = 0.003).
Rhinovirus and RSV were the most frequent pathogens in the study, detected in 318 (52.1%) and 299 (49%) cases, respectively, followed by AdV (10.6%), hMPV (9.7%), PIV-3 (8.4%), and enterovirus (7.7%). Each of the other pathogens was detected in less than 5% of cases. Among the atypical bacteria, Mycoplasma pneumoniae was most common, identified in 22 cases (3.7%). A single pathogen was detected in 45.3% of the study specimens (n = 276), two pathogen in 38% (n = 232), three pathogen in 12% (n = 73), and four or more pathogen in 3.4% (n = 21). RSV, hMPV, IA and IB were detected as single pathogen more often than RV, PIV-3, AdV, and BoV, which were more often present in codetections (Table 3). Enterovirus, parechovirus (PeV), and CoV-NL63 were not detected alone/as single pathogen in any of the specimens. Neither were PIV-2, influenza C virus, CoV 229E, and CoV-HKU1, but they were each detected in ,10 cases. Single detection was seen for CoV-OC43 in three cases. Among the atypical bacteria, only 4 of 22 specimens with M. pneumoniae were single detections, and none of the 7 C. pneumoniae or 2 B. pertussis isolates were detected alone.
For RSV, hMPV, IA, RV, AdV, and BoV, the mean C T value tended to be lower in single-pathogen detection than in codetection. This finding was significant for RSV, RV, and AdV (Table 4). The mean C T value for RSV was lower than that of RV both in the single-detection group and in the codetection group. Moreover, the mean difference in C T values between single detections and codetections was three times greater for RV than for RSV, 4.1 versus 1.3, respectively, reflecting a greater variability of RV C T values overall. In addition, the mean C T value of RV was higher when codetected with RSV than when not (31.8 versus 30.0). Viral seasonal distribution. During the 30-month study period, RSV epidemics occurred at the end of the monsoon (September to October) or during winter (November to February). The first RSV outbreak peaked in September (2006) and was over by December, while the following year, there was a smaller peak in October but the outbreak was prolonged; it reached its maximum peak in January and did not end until April (2007) (Fig. 1A). We also observed RSV activity at the beginning of the study that peaked in February, possibly representing the tail end of a larger epidemic starting before the study period.
The hMPV activity occurred in epidemics similar to those of RSV but with lower magnitudes. The first hMPV outbreak lasted from October 2006 through April 2007, with two peaks in December and March. The second outbreak occurred from August to December 2007 with a peak in October. Neither RSV nor hMPV was detected in significant numbers between epidemics, in contrast to RV and AdV, which were detected throughout the study, AdV evenly throughout the year and RV with various epidemic peaks (Fig. 1B). RV outbreaks peaked in autumn (September in both years), winter (December 2007), and spring (March 2006), and the two major RV peaks coincided with or substantially overlapped the RSV outbreaks.

DISCUSSION
Retrospective reanalysis of previously collected NPA specimens with an expanded real-time PCR panel including 21 respiratory pathogens resulted in 895 additional detections, of which 284 were new detections of the previously analyzed viruses and 611 were detections of the 14 newly added microbial targets.
Retesting for the original seven viruses with updated assays increased the number of detections substantially, from 194 in 627 cases to 478 in 610 cases. Using real-time multiplex PCR methods, the detections of hMPV increased more than six times, detections of RSV increased three times, and detections of PIV-3 increased two times, whereas IA, IB, PIV-1, and PIV-2 were detected in similar numbers as in the previous analyses, suggesting that the Hexaplex plus assay lacked sensitivity, especially with regard to RSV, hMPV, and PIV-3. Specimens that tested positive in reanalysis only had higher mean C T values, supporting this notion. However, detailed calculation of the limits of detection (LODs) for the qualitative Hexaplex plus assays and semiquantitative real-time assays was not performed at the time of local validation. Direct comparison of the LODs of the old and new methods was not possible, as the Hexaplex plus assay was no longer in use in the routine diagnostic laboratory at the time of reanalysis.
After reanalysis with the expanded respiratory panel, we detected respiratory viruses in 98% of pneumonia cases. This detection rate is higher than those reported by two comparable studies (29,30). These comprehensive multicenter pneumonia etiology studies in low-and middle-income countries (LMICs) reported detection of one or more viruses by PCR in specimens from the upper respiratory tract in 89% and 78% of hospitalized children ,5 years of age with severe pneumonia. Both studies used multiplex real-time PCR and tested for a range of respiratory viruses similar to that in the current study.
The higher proportion of viral detections in the current study was mainly due to RSV and RV, which were the most frequently detected by far, both present in about half of the cases. This is a larger proportion than in the two above-mentioned studies, which detected RSV in 20 to 26.5% and RV in 23 to 25% of pneumonia cases (29,30). These discrepancies may be explained by differences in age distribution, case definition, study timing, and test sensitivity. Notably, both studies demonstrated substantial variation in the detection of pathogens between countries.
The incidence of viral infection in children hospitalized with community-acquired pneumonia (CAP) is significantly higher in those ,2 years of age than in those between 2 and 4 years of age (10,31,32). This is even more pronounced in the first year of life, when RSV plays a pivotal role (29,30). Our study population consisted of children ,3 years old, with 82.5% aged 2 to 11 months, which is a higher proportion than in the two LMIC studies cited above.
The pneumonia case definition (WHO Integrated Management of Childhood Illnesses [IMCI] guidelines) used in this study was designed for high sensitivity in a lowresource setting and will consequently identify most episodes of ALRI, including bronchiolitis and pneumonia (33). RSV is known to be the most common cause of acute bronchiolitis in young children, followed by RV, and some studies of acute bronchiolitis have detected RSV in up to 70 to 80% of cases and RV in up to 34% (34)(35)(36)(37).
The magnitudes of RSV and RV outbreaks depend on a complex interplay between susceptibility in the population, climatic factors, and circulating viral subtypes, among other factors (38)(39)(40)(41), and may vary widely between studies.
In this study, we emphasized optimal sensitivity for RV detection, including all the newly discovered rhinovirus species C (RV-C) strains. This increased the RV test sensitivity considerably.
The proportions of other pathogens detected in the current study were in line with the other two studies (29,30), except that we detected bocavirus less frequently (4.4% versus 9.2% and 13.1%) and M. pneumoniae somewhat more frequently (3.7% versus 1.1% and 1.5%.).
Codetections occurred in 53.3% of cases in the current study. Multiple detections are common in respiratory specimens of young children with acute respiratory infection (ARI). In epidemiological studies based on molecular detection methods, at least two viruses or atypical bacteria have been reported in 12 to 40% of cases (10,12,15,(42)(43)(44). Codetections vary across age groups and are more common in younger children, who usually are more susceptible to viral infections and have higher viral loads in their respiratory secretions (45,46). Moreover, RVs are shed for longer time periods; after the onset of symptomatic respiratory infection, rhinovirus RNA may take up to 5 to 6 weeks to disappear from the nasal secretions of children (47). Seasonal circulation patterns will also influence codetections, as well as overall detections, explaining the differences between studies.
We observed 133 codetections of RSV and RV. RV demonstrated an endemic seasonal pattern with detections throughout the year, and RV peaks coincided/overlapped with RSV outbreaks (Fig. 1). One could speculate that RV might be a "bystander" in some children with RSV, as RSV detections during RSV outbreaks would be the most likely cause for the hospital admission (48). In addition, RSV was more often detected as a single infection than was RV, and the mean C T value for RSV was significantly lower than that of RV in codetections.
Respiratory syncytial virus, hMPV, IA, and IB were detected as single pathogen more often than RV, PIV-3, AdV, CoV, and BoV, which were frequently identified in codetections. Some viruses were not detected alone in any of the specimens, including enterovirus (EV), PeV, and CoV-NL63. It has been shown that AdV, RV, EV, and BoV can be detected for longer periods in the airways, due to asymptomatic carriage, persistent shedding after infection (47,49,50), or reactivation of latent infection (51,52), which may explain why these viruses are more often found in codetections. Unexpectedly, PIV-3, a virus well known to cause ALRI in young children (53), was also more frequently found in combination with other pathogens. Studies of CoV in children with ARI have reported that nearly half to two-thirds of CoV detections were codetections, including in Nepalese children (21,54). Similar findings have been reported for AdV in Norwegian children with ARI (55).
The frequency of codetections and the fact that many of the pathogens in question are also commonly detected in asymptomatic children highlights the need for a casecontrol design in epidemiological studies to quantify the etiological contributions of individual pathogens in pneumonia (60). A review and meta-analysis of 23 case-control studies published from 1990 to 2014 estimated that there was strong evidence of causality for RSV and less strong evidence for RV in children ,5 years with ALRI (48). This was confirmed in the two more recent case-control studies cited above (29,30). The meta-analysis also found strong evidence of a causal role for influenza virus, hMPV, and PIV, whereas no difference between cases and controls was seen for the detection of AdV, BoV, or CoV (48). This is mainly in line with the tendency seen in our findings regarding single and codetected pathogens.
M. pneumoniae was detected alone in only 4 of the 22 cases that tested positive for this pathogen. Others have reported a relative infrequency of codetections between M. pneumoniae and respiratory viruses (32). Using the Biofire FilmArray, Zheng and coworkers reported that 26/80 (33%) respiratory specimens from symptomatic children had a viral pathogen detected along with M. pneumoniae (61). Using a case-control design, Spuesens et al. demonstrated that M. pneumoniae was present in the upper respiratory tract of 21% of asymptomatic children aged 3 months to 16 years (62), suggesting that asymptomatic carriage of this atypical bacterium is not uncommon and could explain its involvement in codetections.
Seasonality. We have proposed previously that RSV activity follows a biennial rhythm in the Kathmandu Valley, with alternating early and late onset of epidemics, similar to what has been observed in Europe (28). In southern Nepal, Perchetti et al.
reported hMPV infections to occur during fall and winter months with detections starting in September or October and in two of three seasons extending into March/April (63), similar to our findings. The larger RV outbreaks peaked in September both years, and a smaller peak was seen during winter or spring. Others have previously described RV infections appearing year-round with peaks in early autumn and spring (38,64,65), a pattern that has been associated with the circulation of different RV types, creating simultaneous or successive epidemics (66).
Limitations. The main limitation of our study is that it was embedded in a clinical trial of zinc as an adjunct treatment for severe pneumonia (67) and, therefore, was subject to the eligibility criteria of the main trial. The major consequence of this was that not all children with severe pneumonia admitted to the Kanti Children's Hospital were included during the study period, consequently making our findings somewhat less generalizable. A history of recurrent wheezing, nonconsent, heart disease, or other severe conditions were the main reasons for exclusion (67).
Pneumococcal conjugate vaccine (PCV) has been reported to reduce the risk of virusassociated pneumonia (6). Vaccines against Hib and S. pneumoniae were introduced in the routine infant immunization program in Nepal in 2009 and 2015, respectively, i.e., after the collection of specimens for this study. The implementation of these vaccines may have altered the local distribution of pathogens in severe childhood pneumonia.
The enrollment of the study participants was conducted over 30 months. Despite being longer than most studies, it may still be too short to capture pathogens whose epidemiology is greater than two-and-a-half-year cycles.
Conclusion. In summary, using a broad respiratory PCR panel, we were able to detect a respiratory pathogen in almost all children 2 to 35 months old with severe pneumonia. The study was conducted in a large, well-defined population of young, hospitalized children in a subtropical-to-temperate low-income country. Further studies, preferably with a case-control design and quantitative detection methods, are warranted to elucidate the local epidemiology of respiratory viruses in the era of routine use of PCV and the Hib vaccine to inform future strategies for prevention and treatment of pneumonia in low-income settings.

MATERIALS AND METHODS
Study subjects, case definition, and specimen collection and handling. The original study was conducted between 1 January 2006 and 30 June 2008 at Kanti Children's Hospital in Kathmandu, a tertiary referral hospital for the whole of Nepal. Children aged 2 to 35 months presenting to the hospital were screened for eligibility. Severe pneumonia was defined as cough or difficult breathing combined with lower chest wall indrawing (LCI), according to the Integrated Management of Childhood Illnesses (IMCI) algorithm for acute respiratory infections (68). Moreover, children with wheezing were given up to 3 doses of nebulized salbutamol 15 min apart and were not found eligible if LCI disappeared on reassessment (69). Additional details of participant selection and other study procedures can be accessed in a previous publication (28).
Study physicians collected NPA specimens from all children upon enrollment in the study. Immediately following collection, the NPA specimen was divided into three aliquots and stored at 270°C. While one aliquot was analyzed on location in Nepal, the remaining two aliquots were later shipped to Norway on dry ice.
Previous PCR methods and results. A total of 627 children with severe pneumonia included in the original study from whom NPA specimens were collected at presentation had a valid PCR result (70). Each specimen was tested for RSV, influenza A and B viruses, PIV-1, -2, and -3, and hMPV using a commercially available multiplex reverse transcription-PCR assay (Hexaplex Plus, Prodesse, Inc., Waukesha, WI). We identified $1 respiratory virus in 188 of the 627 cases (30%), whereas in 439 cases (70%), no virus was detected (28). Of the 627 previously collected NPA specimens stored at 270°C, 610 were available for reanalysis.
PCR methods for the reanalysis. At the Department of Microbiology and Infection Control, University Hospital of North Norway, nucleic acids were extracted automatically (NucliSens easyMAG [bioMérieux] or Chemagic STAR [Hamilton]). Simultaneous triplex or duplex real-time reverse transcription-PCR analyses were performed for PIV-1, -2, and -3, influenza B virus and respiratory syncytial virus, human metapneumovirus and human coronavirus NL63, influenza A virus and human coronavirus OC43, and rhinovirus and human coronavirus 229E using previously described primers, probes, and cycling conditions (22). In-house real-time PCR analyses were performed for adenovirus, Mycoplasma pneumoniae, Chlamydia pneumoniae, and Bordetella pertussis using TaqMan probes (Roche Diagnostics, Basel, Switzerland). All amplifications were carried out in the ABI 7500 realtime PCR system (Applied Biosystems). Analyses were performed during 2014.
To test for additional viruses, nucleic acids were extracted automatically (MagNA Pure 96; Roche Diagnostics, Basel, Switzerland) at the Department of Microbiology, Lillehammer Hospital. All 610 lysates were tested with PCRs for additional viruses, viz., human coronavirus HKU1, human bocavirus 1, enterovirus, parechovirus, influenza C virus, PIV-4, and rhinovirus species A to C at the Department of Medical Microbiology, St. Olav's Hospital, Trondheim, Norway. The PCRs were in-house, TaqMan real-time assays (Table 5) (71). Amplifications were performed on the CFX96 real-time system (Bio-Rad). Analyses were performed during 2017 to 2018.
The PCR results were considered positive for all sigmoidal curves and cycle thresholds (C T ) of ,40 cycles. All positive RV-A to -C results were evaluated for cross-reactivity due to EV infections by use of parallel EV PCR testing. Samples with weak amplification curves for RV and a positive EV PCR were regarded as RV negative.
Statistical methods. We analyzed the data using Stata/MP 12.0 for Macintosh (Stata Corporation, College Station, TX). The 95% confidence intervals for proportions were calculated using the "ci" command. To compare the mean C T values in single versus codetections for each pathogen, we used the two-sample t test. Statistical significance was defined as a P value of ,0.05.
Ethical issues. The study and biobank had ethical clearance from the Nepal Health Research Council, Kathmandu, and the Western Regional Committee for Medical and Health Research Ethics, Norway (REC West 129.03). The implementation of the project was in agreement with international ethical principles for medical research involving human subjects as stated in the latest version of the Helsinki Declaration.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.3 MB.