Aetiology of childhood pneumonia in a well vaccinated South African birth cohort: A nested case-control study of the Drakenstein Child Health Study

Background: Pneumonia is a leading cause of mortality and morbidity in children globally. The cause of pneumonia after introduction of the 13-valent pneumococcal conjugate vaccine (PCV13) has not been well studied in low-income and middle-income countries, and most data are from cross-sectional studies of children admitted to hospital. We aimed to longitudinally investigate the incidence and causes of childhood pneumonia in a South African birth cohort. Methods: We did a nested case-control study of children in the Drakenstein Child Health Study who developed pneumonia from May 29, 2012, to Dec 1, 2014. Children received immunisations including acellular pertussis vaccine and PCV13. A nested subgroup had nasopharyngeal swabs collected every 2 weeks throughout infancy. We identified pneumonia episodes and collected blood, nasopharyngeal swabs, and induced sputum specimens. We used multiplex real-time PCR to detect pathogens in nasopharyngeal swabs and induced sputum of pneumonia cases and in nasopharyngeal swabs of age-matched and site-matched controls. To show associations between organisms and pneumonia we used conditional logistic regression; results are presented as odds ratios (ORs) with 95% CIs. Findings: 314 pneumonia cases occurred (incidence of 0·27 episodes per child-year, 95% CI 0·24-0·31; median age 5 months [IQR 3-9]) in 967 children during 1145 child-years of follow-up. 60 (21%) cases of pneumonia were severe (incidence 0·05 episodes per child-year [95% CI 0·04-0·07]) with a case fatality ratio of 1% (three deaths). A median of five organisms (IQR 4-6) were detected in cases and controls with nasopharyngeal swabs, and a median of six organisms (4-7) recorded in induced sputum (p=0·48 compared with nasopharyngeal swabs). Bordetella pertussis (OR 11·08, 95% CI 1·33-92·54), respiratory syncytial virus (8·05, 4·21-15·38), or influenza virus (4·13, 2·06-8·26) were most strongly associated with pneumonia; bocavirus, adenovirus, parainfluenza virus, Haemophilus influenzae, and cytomegalovirus were also associated with pneumonia. In cases, testing of induced sputum in addition to nasopharyngeal swabs provided incremental yield for detection of B pertussis and several viruses. Interpretation: Pneumonia remains common in this highly vaccinated population. Respiratory syncytial virus was the most frequently detected pathogen associated with pneumonia; influenza virus and B pertussis were also strongly associated with pneumonia. Testing of induced sputum increases the yield for detection of several organisms. New vaccines and strategies are needed to address the burden of childhood pneumonia. Funding: Bill & Melinda Gates Foundation, Medical Research Council South Africa, National Research Foundation South Africa, National Institute of Health, and H3Africa. We did a nested case-control study of children included in the Drakenstein Child Health Study12 who developed pneumonia from May 29, 2012, to Dec 1, 2014. The Drakenstein Child Health Study12 was undertaken at two public, primary health-care clinics located about 2 km apart in Paarl, a periurban area in South Africa. One clinic (TC Newman) served a mixed-race population and the second clinic (Mbekweni) a black African population. Pregnant women aged 18 years or older, at 20–28 weeks’ gestation, attending one of the two clinics for antenatal care, and remaining in the area for at least 1 year were enrolled. Ethics approval was obtained from the University of Cape Town Faculty of Health Sciences Research Ethics Committee, and the Provincial Research committee approved the study. Mothers provided written informed consent at enrolment and provided consent after the first year. All births occurred at Paarl hospital. Follow-up of children was done from May 29, 2012, through early childhood13 and paralleled routine child health visits at 6, 10, and 14 weeks and 6, 9, 12, 18, 30, and 42 months. An additional study visit was done at 6–10 weeks at Paarl hospital. All children were given primary health care and immunisations at the two clinics including four doses of a five vaccine combination (diphtheria, tetanus, acellular pertussis, H influenzae type b, and inactivated polio vaccine) at 6, 10, and 14 weeks and 18 months; the measles vaccine at 9 months and 18 months; and the PCV13 at 6 weeks, 14 weeks, and 9 months. Continuous pneumonia surveillance was implemented at all local clinics and at Paarl hospital. Mothers were given a mobile phone number if they needed to contact the study team at any time. Mothers were counselled regarding key respiratory symptoms and advised to attend or contact study staff whenever a child developed cough or difficulty breathing. Primary health nurses and study staff were trained to recognise WHO-defined pneumonia or severe pneumonia.14 Study staff reviewed patient records at catchment clinics (Phola Park, Thokoza; JJ du Preez, Paarl; Klein Nederburg, Paarl, South Africa) and Paarl hospital, and performed surveillance for any missed pneumonia episode. All admissions to hospital were at Paarl hospital, the only hospital serving this population. Children were followed throughout their duration in hospital, and at 2 days and 6 weeks after discharge or after an ambulatory episode. Longitudinal measurement of risk factors (nutrition, environment, vaccinations received, and child and maternal factors) was done at study visits and at case presentation. Infant anthropometry and maternal smoking or passive smoke exposure were measured by urine cotinine longitudinally and at case presentation.15 Nasopharyngeal swabs were collected every 2 weeks for the first year in a subgroup (intensive cohort); enrolment in the intensive cohort was at the participant’s discretion. All children had nasopharyngeal swabs taken every 6 months. A chest radiograph was done in infants with pneumonia who had been admitted to hospital. Laboratory staff were masked to case-control status. Cases were any episode of pneumonia, irrespective of severity, excluding congenital pneumonia (defined as presentation before postnatal discharge). Controls were incidence-density matched to cases (1–2:1) by birth date (to within 2 weeks), age of presentation (to within 2 weeks), and site of enrolment. By design, controls could be sampled more than once, but this occurred only infrequently. We did separate analyses of cases compared with asymptomatic controls and with controls with symptoms of upper respiratory tract infection (ie, cough, runny or blocked nose, or sore throat). For every case, two nasopharyngeal swabs (FLOQSwabs, Copan Diagnostics, Murrieta, CA, USA) and an induced sputum specimen were obtained.10 The first nasopharyngeal swab taken was immediately transferred into nucleic acid preservation medium (PrimeStore, Longhorn Vaccines and Diagnostics, San Antonio, TX, USA), the second swab was placed into 1 mL of skimmed milk-tryptone-glucose-glycerol (STGG) transport medium. Swabs were transported on ice to the laboratory and frozen at −80°C for batch testing. The swab in STGG was cultured for bacteria; total nucleic acid was extracted from the swab in nucleic acid preservation medium with mechanical lysis on a Tissuelyzer LT (Qiagen, Hilden, Germany) followed by extraction with the QIAsymphony Virus/Bacteria Mini Kit (Qiagen, Hilden, Germany). We did quantitative, multiplex, real-time PCR (qPCR) with FTDResp33 (Fast-Track Diagnostics, Esch-sur-Alzet, Luxembourg) to identify up to 33 potential organisms of respiratory viruses (influenza A, B, and C; parainfluenza 1, 2, 3, and 4; coronaviruses NL63, 229E, OC43, HKU1; human metapneumoviruses A and B; rhinovirus; respiratory syncytial viruses A and B; adenovirus; enterovirus; parechovirus; bocavirus; and cytomegalovirus), fungi (Pneumocystis jirovecii), and bacteria (Mycoplasma pneumoniae, Chlamydophila pneumoniae, Streptococcus pneumoniae, H influenzae type b, Staphylococcus aureus, Moraxella catarrhalis, Bordetella pertussis, Klebsiella pneumoniae, Legionella species, salmonella species, and H influenzae). K pneumoniae and Legionella spp were omitted from this analysis because of difficulties with assay specificity. Standard curves were derived with plasmid standards supplied by the manufacturer for each organism. Every induced sputum specimen was transported to the laboratory on ice and split into two aliquots. The first underwent nucleic acid extraction and testing with FTDResp33, as previously stated; the second was cultured for bacteria. A blood culture for bacteria was obtained in cases admitted to hospital. For control children, two nasopharyngeal swabs were collected at each visit and stored for later retrieval and processing as described for case swabs. Laboratory staff were masked to case-control status. Weight-for-age and height-for-age Z scores were derived from WHO child growth standards.16 Socioeconomic status comprised a composite of asset ownership, household income, employment, and education.13 We compared data from cases and controls with conditional logistic regression. Dependent variables of interest were organisms from FTDResp33, analysed as binary (present or absent) and continuous (log copies or specimen) values. Model building examined potential confounding factors identified a priori from demographic and clinical measures of child’s sex, in-utero HIV exposure, maternal age, maternal smoking, and socioeconomic status; because no clear confounding factors were consistently identified based on appreciable changes in the point estimate for pathogen–pneumonia associations, models presented account for matching factors only. Because children might have had more than one episode of pneumonia (and thus participate in more than one case-control set), we also examined mixed-effects models with children and case-control sets as random or fixed effects, and also conditional logistic regression models restricted to first case episodes. We based co-organism analysis on matched case-control pairs, investigating the presence or absence of two organisms at a time. Of 16 possible response patterns, seven involved co-occurrence, with the less than one, one, or more than one pattern of co-occurrence in both cases and controls not adding information that differentiated cases from controls. The observed frequencies of the remaining six response patterns were compared with their expected values based on the hypothesis of random co-occurrence with a Pearson χ2 test. Multiple p values for different organism pairs were corrected with use of the Benjamini-Hochberg correction for false discovery rate. We did subsidiary analyses to stratify case-control comparisons on pneumonia severity and control children’s symptoms. Throughout, regression diagnostics followed standard procedures, and all statistical tests were two-sided at α=0·05. We used Stata (version 13.0) and R (version 3.2.2) for data analyses. The sponsors of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. All authors had full access to all the data and had final responsibility for the decision to submit for publication.