The Etiology of Pneumonia in HIV-uninfected South African Children

Supplemental Digital Content is available in the text.

ened susceptibility to invasive pneumococcal disease and respiratory virus-associated pneumonia during the first 6 months of life in HIV-exposed uninfected compared with HIV-unexposed infants. 6 Furthermore, HIV-exposed infants may be at increased risk for disease due to opportunistic pathogens, such as Pneumocystis jirovecii and human cytomegalovirus. [6][7][8][9] Previous studies of pneumonia etiology in HIV-uninfected South African children under 5 years of age, conducted in the era before Haemophilus influenzae type b (Hib) and pneumococcal polysaccharide-protein conjugate vaccines were incorporated into the National Expanded Program on Immunization, in 1999 and 2009, respectively, indicated that respiratory viruses, pneumococcus and Hib were the predominant organisms leading to severe pneumonia requiring hospitalization. 10,11 These studies did not, however, systematically evaluate for differences in etiology of pneumonia between HIV-exposed uninfected and HIV-unexposed children.
In the era of molecular diagnostic techniques for the identification of respiratory viral infections, and postintroduction of Hib and pneumococcal conjugate vaccine (PCV), respiratory viruses have been identified in 78% of South African children hospitalized with lower respiratory tract infection. 12 This study evaluated the pathogen profiles for pneumonia etiology among hospitalized HIV-uninfected, including HIVexposed, children in Soweto, South Africa, as part of the multicenter Pneumonia Etiology Research for Child Health (PERCH) study. A companion paper, 13 details the pneumonia etiology profiles in HIV-infected children enrolled at the South African PERCH site.

Location
Enrollment into the PERCH study in South Africa occurred at Chris Hani Baragwanath Academic Hospital, which is a public sector health care facility, and was the only hospital serving the majority of Soweto's population at the time of the study. Soweto is located 1600 meters above sea level, has a summer rainfall pattern, and an autumn/winter respiratory illness season which lasts from March through August. 14,15 Further detail regarding the study catchment area is included in Supplemental Digital Content 1, http:// links.lww.com/INF/D829. Pneumonia contributed 11.7% of under 5 deaths in South African children in 2012/2013 with a national average case fatality rate of 3.8%. 16 The antenatal HIV seroprevalence rate in Gauteng Province in 2012/2013 was 34.0%, and the mother-to-child vertical transmission rate was 2.2% [95% confidence interval (CI), 1.3%-3.1%]. 5 Health care is provided free of charge to all children under 6 years of age attending public sector health facilities in South Africa. 17

Participants
Eligibility and exclusion criteria for PERCH case and control selection have been described. 18 In this analysis, we included HIV-uninfected children between the ages of 1 and 59 months, hospitalized with signs of WHO-defined severe/very severe pneumonia (cases). 19 HIV-uninfected controls were frequencymatched to cases according to age-stratification (1-5, 6-11, 12-23 and 24-59 months), and all participants were resident in the study catchment area. Community control selection procedures are described in Supplemental Digital Content 1, http://links.lww. com/INF/D829.

Clinical Procedures
Enrollment occurred through active surveillance in the hospital pediatric admissions ward for potentially eligible cases. Once consented and enrolled, cases were evaluated at baseline, 24 and 48 hours, and again on the day of hospital discharge, to monitor changes in clinical status. Cases were evaluated again after hospital discharge, at least 30 days subsequent to the date of enrollment. All community controls were assessed once, on the day on which they presented to the research clinic at Chris Hani Baragwanath Academic Hospital.

Specimen Collection and Laboratory Methods
Standardized specimen collection and laboratory procedures were followed in cases and controls, as previously described, [20][21][22][23] and are detailed in Supplemental Digital Content 1, http://links. lww.com/INF/D829. A blood culture and a chest radiograph was obtained from cases as part of the routine diagnostic workup. 24 Study-specific specimens obtained from cases and controls included nasopharyngeal, oropharyngeal (NP/OP) swabs for multiplex real-time polymerase chain reaction (PCR) to detect 33 respiratory pathogens (Fast Track Diagnostics Respiratory Pathogens 33 test (FTD-33), Fast Track Diagnostics, Sliema, Malta), whole blood for detection of pneumococcal autolysin (lytA) by PCR, 25 and serum for antibiotic activity. 26 Age-appropriate HIV testing was also done, with consent, in all participants. HIV-exposure status was determined as detailed in Supplemental Digital Content 1, http://links.lww.com/INF/D829.

Statistical Analysis
Descriptive analyses of clinical and laboratory measures, reporting percentages in subgroups (stratified by case/control and HIV-exposure status, as well as by disease severity and radiologic findings amongst cases) were undertaken, and proportions were compared using logistic regression, adjusting for age (in months). Medians and interquartile range were used to describe continuous data. In instances where numerous comparisons were done, P values were adjusted using the Benjamini-Hochberg method. 27 Twosided P values < 0.05 were considered to be statistically significant.
Predefined organism-specific thresholds for defining high load that best distinguished between cases and controls for NP/ OP swab FTD-33 PCR, and lytA on whole blood PCR from the PERCH foundational analyses have been published, 25,[28][29][30] and were applied to cytomegalovirus, H. influenzae, P. jirovecii and Streptococcus pneumoniae in the current analysis. Conditional logistic regression of respiratory pathogen prevalence in the upper respiratory tracts (for all tested potential pathogens) and whole blood (for pneumococcus only) of cases compared with controls, adjusting for age (in months) and all other pathogens, was used to derive the adjusted odds of each organism being associated with case status. This was integrated into analyses of etiology in addition to other specimens and laboratory tests from cases and controls, which was undertaken using a Bayesian approach through the PERCH Integrated Analysis (PIA) that also accounted for sensitivity and specificity of all measurements. 31 The overarching PERCH paper reported on the etiology of pneumonia in HIV-uninfected children with radiologically confirmed pneumonia at each of the study sites. 34 The current paper focuses on the South African cohort and presents the pneumonia etiology estimates for HIV-uninfected children, stratified by HIV-exposure status, age and pneumonia severity. The methodology in the current analysis differs from the overarching PERCH paper through its use of a higher sensitivity prior estimate (20%-50%) for Mycobacterium tuberculosis (Mtb) than was used for the all-site combined analysis (in which the sensitivity prior for Mtb was 10%-30%). The higher sensitivity prior estimate adopted in the South African site-specific analysis was chosen because more intensive screening for Mtb was conducted at the site, 23 and because the site has a high burden of tuberculosis as outlined in Supplemental Digital Content 1, http://links. lww.com/INF/D829.

Ethical Considerations
The Human Research Ethics Committee of the University of the Witwatersrand (M101129) and the Institutional Review Board of the Johns Hopkins Bloomberg School of Public Health approved the study. Parents or legal guardians of all cases and controls provided written consent for participation in PERCH.
There were 20 in-hospital deaths among HIV-uninfected children, with the case fatality rate being similar in HIV-exposed (8/298, 2.7%) and HIV-unexposed children (

Microbiologic Results in HIV-uninfected Cases
Clinically significant pathogens were isolated in 17 (2.1%) of 802 blood cultures submitted in the cases, with Gram-negative organisms (n = 10, 58.8%) predominating over Gram-positive species (n = 7, 41.2%). The single most common blood culture pathogen was, however, Staphylococcus aureus (n = 5; 29.4% of all significant isolates) (Supplemental Digital Content 6, http:// links.lww.com/INF/D834). Nine (52.9%) of the 17 clinically significant blood culture isolates were in children with radiologically confirmed pneumonia, 5 (29.4%) in children with normal chest radiographs and 3 (17.6%) in children with uninterpretable chest radiographs. Bacteremia in children with radiologically confirmed pneumonia was present in 3.0% (5/164) of HIV-exposed children, and in 1.6% (4/245) of those that were HIV-unexposed, P = 0.494. Detailed descriptions of children with Gram-negative bacteremia, those with microbiologically confirmed pneumococcal infection (n = 4) and microbiological results of children that underwent lung aspiration are presented in

Comparison of NP/OP FTD-33 and Whole Blood Pneumococcal PCR Results Between HIVuninfected Cases and Controls, Stratified by HIVexposure Status
In both HIV-exposed and HIV-unexposed children, the presence of any respiratory viral organism detected by PCR on NP/OP swab was significantly associated with case-status (Tables 2 and 3). Furthermore, in HIV-exposed children, RSV [adjusted odds ratio (aOR), 19.03], influenza A (aOR, 5.10), and nontype b H. influenzae (aOR, 1.72) on NP/OP PCR testing and high load pneumococcus detected in whole blood by PCR (aOR 3.34) were associated with radiologically confirmed pneumonia case-status; Table 2. Organisms associated with case-status in HIV-unexposed children with radiologically confirmed pneumonia compared with controls were generally similar but more organisms were significantly associated because sample size was larger ( Table 3). The additional organisms included parainfluenza virus 1 (aOR, 19.15), influenza B (aOR, 9.07), parainfluenza virus 3 (aOR, 6.90), Bordetella pertussis (aOR, 6.85), and high-density P. jirovecii (aOR, 2.32); Table 3.
Among the eight HIV-exposed children who died in hospital, high-density NP/OP carriage of P. jirovecii and HMPV were significantly higher (both 2/8 cases each) compared to control carriage (

PERCH Integrated Analysis Determination of Pathogen Etiology Fraction in HIV-uninfected South African Children, Stratified by HIV-exposure Status
Bayesian analytic outputs identified RSV as being the most important contributor to severe/very severe pneumonia in HIVexposed (EF 31.6%) and HIV-unexposed (EF 36.4%) children hospitalized with radiologically confirmed pneumonia (Fig. 1). Furthermore, Mtb contributed substantially to the EF of radiologically confirmed pneumonia in HIV-exposed (EF 11.6%) and HIVunexposed (EF 8.3%) children (Fig. 1). Overall, respiratory viral pathogens contributed a larger combined EF in HIV-exposed and HIV-unexposed children than did bacteria (Fig. 1).

Age-and Pneumonia Severity-stratified Analysis for Etiologic Fraction Estimation in HIV-exposed and HIV-unexposed Children
The "top 10" pathogens associated with radiologically confirmed pneumonia in HIV-exposed and HIV-unexposed children, stratified by age group, are shown in Table 4. RSV and Mtb were the first and second ranked pathogens associated with radiologically confirmed pneumonia in children <12 months of age in HIVexposed (EF 38.2% and 10.4%) and HIV-unexposed (EF 46.5% and 9.7%) children (Table 4). Pneumococcus and Mtb were the first and second ranked pathogens (EF 16.5% and 16.3%) associated with radiologically confirmed pneumonia in HIV-exposed children ≥12 months of age (Supplemental Digital Content 11, http://links. lww.com/INF/D839 and Table 4). The "top 10" organisms contributed over 69% of the EF in each age group (Table 4), and the contribution of the "not otherwise specified" category to pneumonia etiology was 2.5% or less ( When stratifying by pneumonia severity, RSV remained the top-ranked pathogen in HIV-uninfected children overall, HIV-exposed and HIV-unexposed children (Supplemental Digital Content 13, http://links.lww.com/INF/D841). Furthermore, in the severity-stratified analysis, Mtb was consistently implicated within the "top 10" ranked organisms regardless *Conditional odds ratio derived by logistic regression, adjusting age (in months) and presence of all other pathogens: 2 analyses were combined in the output of this Table: the first with no threshold applied for human cytomegalovirus, H. influenzae, P. jirovecii, and S. pneumoniae, and the second with threshold density cutoffs (as noted below) applied to these pathogens. The first analysis output was used to report the adjusted conditional odds for cytomegalovirus, H. influenzae, P. jirovecii, and S. pneumoniae with no threshold density cutoff applied. The second analysis output was used to report the adjusted conditional odds for all pathogens named in the Table. †Cutoff density threshold which best distinguished between cases and controls, derived by receiver operating characteristic analysis using leave-one-out cross-validation. ‡Cutoff density for H. influenzae (nontype b, and type b) on NP/OP swabs: 5.9 log 10 copies/mL. §Cutoff density for S. pneumoniae on NP/OP swabs: 6.9 log 10 copies/mL. ¶Vaccine-type or nonvaccine-type pneumococcus amongst children with high density NP/OP pneumococcal carriage. ║Cutoff density for S. pneumoniae in whole blood specimens: 2.2 log 10 copies/mL. **Cutoff density for P. jirovecii on NP/OP swabs: 4.0 log 10 copies/mL. † †Cutoff density for human cytomegalovirus on NP/OP swabs: 4.9 log 10 copies/mL. CI indicates confidence interval; CXR+, radiologically confirmed pneumonia; N/E, no estimate; NP/OP, Nasopharyngeal/oropharyngeal. Bolded values indicates statistically significant results, in which 95% confidence intervals do not include zero. of HIV-exposure status. P. jirovecii ranked fifth in association with severe pneumonia in HIV-exposed children (EF 3.5%; 95% CrI, 0.0%-9.9%) but did not feature in the "top 10" organisms associated with pneumonia in HIV-unexposed children (Supplemental Digital Content 13, http://links.lww.com/ INF/D841). *Conditional odds ratio derived by logistic regression, adjusting age (in months) and presence of all other pathogens: 2 analyses were combined in the output of this Table: the first with no threshold applied for human cytomegalovirus, H. influenzae, P. jirovecii, and S. pneumoniae, and the second with threshold density cutoffs (as noted below) applied to these pathogens. The first analysis output was used to report the adjusted conditional odds for cytomegalovirus, H. influenzae, P. jirovecii, and S. pneumoniae with no threshold density cutoff applied. The second analysis output was used to report the adjusted conditional odds for all pathogens named in the Table. †Cutoff density threshold which best distinguished between cases and controls, derived by receiver operating characteristic analysis using leave-one-out cross-validation. ‡Cutoff density for H. influenzae (nontype b, and type b) on NP/OP swabs: 5.9 log 10 copies/mL. §Cutoff density for S. pneumoniae on NP/OP swabs: 6.9 log 10 copies/mL. ¶Vaccine-type pneumococcus amongst children with high density NP/OP pneumococcal carriage. ║Cutoff density for S. pneumoniae in whole blood specimens: 2.2 log 10 copies/mL. **Cutoff density for P. jirovecii on NP/OP swabs: 4.0 log 10 copies/mL. † †Cutoff density for human cytomegalovirus on NP/OP swabs: 4.9 log 10 copies/mL. CI indicates confidence interval; CXR+, radiologically confirmed pneumonia; N/E, no estimate; NP/OP, Nasopharyngeal/oropharyngeal. Bolded values indicates statistically significant results, in which 95% confidence intervals do not include zero.

Sensitivity Analysis in the PERCH Integrated Analysis Outputs for Mtb in the South African PERCH Cohort
Sensitivity analyses which adopted a lower (10%-30%) sensitivity prior for Mtb culture from clinical specimens resulted in higher EFs attributable to Mtb in the South African HIV-uninfected cohort, including an EF of 19.5% (95% CrI, 6.1%-36.4%) in HIV-exposed children >12 months of age (compare Table 4

DISCUSSION
The South African site contributed 25% (435/1769) of HIVuninfected cases with radiologically confirmed pneumonia enrolled in the multi-site PERCH study. 34 The results of the analyses presented here give further insight into the etiology of childhood pneumonia in a low-middle income sub-Saharan African setting with established vaccination programmes against Hib and pneumococcus as well as a high HIV-and tuberculosis burden. In summary analysis, illness severity and age-stratified analyses (amongst infants), RSV was implicated as the leading pathogen associated with radiologically confirmed pneumonia in HIV-uninfected South African children, as also observed across all other PERCH sites. 34 RSV has also been identified as the leading cause of communityacquired pneumonia (CAP) in other recent studies. 6,35,36 Furthermore, PIA outputs for HIV-uninfected children consistently highlighted the prominent role of Mtb in CAP etiology in the South Africa African HIV-uninfected children, despite high Bacillus Calmette-Guérin coverage in the cohort. Mtb featured within in the "top 10" organisms associated with childhood severe/very severe pneumonia at all PERCH sites, 34   922 per 100,000 to 849 per 100,000. 37 Marked improvements in integration of HIV and tuberculosis services in South Africa, with expedited access onto antiretroviral treatment for HIV-infected persons, has driven this decline in tuberculosis incidence. 38 Documented significant reductions in childhood microbiologically confirmed pulmonary tuberculosis in South Africa have been noted in HIV-infected and -uninfected children, also reflecting up-scaled access to antiretroviral treatment for HIV-infected individuals. 39,40 The cumulative incidence (per 100,000 population) of culture-confirmed pulmonary tuberculosis in HIV-uninfected Sowetan children under 5 years of age in 2012, the midpoint of PERCH enrollment activities at the South African site, was 27.5 (95% CI, 18.8-38.8), 40 which is similar to the current WHO estimate for microbiologically confirmed childhood tuberculosis in South Africa (39; 95% CI, . 41 When taking into consideration the imperfect sensitivity (20%-50% in the current analyses) of cultureconfirmation of disease caused by Mtb, these incidence rates likely represent a conservative estimate of the role of Mtb in acute severe/ very severe pneumonia in our setting. HIV-exposed status appears to be an important determinant of Mtb-associated CAP in South Africa, as evidenced by the PIA outputs which ranked the organism first amongst HIV-exposed children ≥12 months of age (  42,43 High rates of exposure to infectious tuberculosis cases perpetuate the burden of latent infection and disease in their close contacts, which is concerning considering that coverage of isoniazid preventive therapy for at-risk children in South Africa is suboptimal, [44][45][46] despite compelling evidence of its effectiveness in preventing tuberculosis. 47 Relative immune paresis of HIV-exposed infants, compared to those that are HIV-unexposed, 48 may explain the contribution of P. jirovecii and vaccine-serotype pneumococci to case-status in HIVexposed children in this study. Although P. jirovecii contributed similar EFs (2.4% vs. 1.9%) to pneumonia etiology in HIV-exposed and HIV-unexposed children in the South African PERCH cohort, it was associated with disease in the youngest children (regardless of HIV-exposure status) and featured among the "top 10" pathogens associated with severe pneumonia in HIV-exposed infants in severity-stratified analyses. P. jirovecii was also significantly associated with case-status among children dying in-hospital. Heightened risk for disease associated with vaccine preventable infections amongst HIV-exposed children was evidenced by the contribution of vaccine-serotype pneumococci amongst the "top 10" pneumonia pathogens in HIV-exposed children at the South African PERCH site. In contrast, nonvaccine serotype pneumococci were implicated among the "top 10" pathogens in HIV-unexposed children. Highdensity vaccine-type pneumococcal carriage prevalence in cases with radiologically-confirmed pneumonia were 6.1% (10/164) and 3.3% (8/245) in HIV-exposed and HIV-unexposed children, respectively. Certain studies have demonstrated less robust responses to PCV in HIV-exposed compared to HIV-unexposed children. 49 The wider spectrum of nonviral pathogens (B. pertussis, H. influenzae and P. jirovecii) that were statistically associated with case-status in the conditional logistic regression analyses of NP/OP carriage in HIV-unexposed compared with HIV-exposed children could represent a type-I sampling error, incurred by smaller numbers of HIV-exposed than HIV-unexposed children in this study. However, some pathogens were more prevalent amongst HIVunexposed than amongst HIV-exposed cases with radiologicallyconfirmed pneumonia, such as parainfluenza 1 (3.3% vs. 0.6%) and parainfluenza 3 (7.8% vs. 3.7%). Cohen et al 6       www.pidj.com | S67 higher incidence rates of lower respiratory tract illness associated with adenovirus, enterovirus, HMPV, human rhinovirus, parainfluenza virus 1 and RSV in HIV-exposed compared with HIV-unexposed South African children, albeit in a cohort of infants with respiratory illness requiring hospitalization but not confined to WHO severe/very severe pneumonia as was the case in PERCH. Viral pathogens were associated with pneumonia etiology in 49.8% (95% CrI, 39.4%-60.6%) HIV-exposed children and 57.9% (95% CrI, 48.8%-67.5%) HIV-unexposed children with radiologically confirmed pneumonia, which contrasts with the 26.8% (95% CrI, 15.7%-38.2%) viral etiologic estimate in HIV-infected South African children. 13 This may have implications for empiric therapy of CAP in HIV-uninfected children residing in regions with widespread vaccine coverage against Hib and pneumococcus if the clinical scenario permits, so as to limit the use of empiric antibiotic therapy through clinical guidelines either encouraging avoidance of antibiotic therapy or promoting short-course, narrowspectrum antibiotic therapy. Few respiratory viral pathogens are preventable through vaccination, and a limited armamentarium of antiviral agents are available to treat severe disease attributable to these pathogens. Uptake of influenza vaccination in sub-Saharan Africa is limited, and vaccine effectiveness is unpredictable. 50 A fresh realization of the contribution that RSV makes to the global pediatric pneumonia burden has invigorated efforts to develop safe and efficacious RSV vaccines. 51 A strategy of vaccinating pregnant women against influenza has been shown to impact considerably on the burden of all-cause pneumonia hospitalizations among their infants. 52 A recent trial of antenatal RSV vaccine administration has shown efficacy in preventing severe RSV-associated pneumonia in infants. 53

CONCLUSIONS
RSV contributes substantially to the burden of severe/very severe pneumonia requiring hospitalization in HIV-uninfected children in South Africa. A safe, effective vaccine against RSV would be anticipated to impact substantially on the burden of pneumonia hospitalization amongst young children in this setting. Furthermore, Mtb was prominently associated with HIVuninfected case-status in all age, severity, and HIV-exposure groups. Efforts must be made to strengthen tuberculosis programmes in South Africa, including a renewed emphasis on the importance of isoniazid preventive therapy in child contacts of infectious tuberculosis cases, as well as active case finding so as to identify undiagnosed close contacts of newly diagnosed tuberculosis patients.