Risk factors for Bordetella pertussis disease in hospitalized children

Introduction Despite a resurgence of disease, risk factors for pertussis in children in low and middle-income countries are poorly understood. This study aimed to investigate risk factors for pertussis disease in African children hospitalized with severe LRTI. Methods A prospective study of children hospitalized with severe LRTI in Cape Town, South Africa was conducted over a one-year period. Nasopharyngeal and induced sputum samples from child and nasopharyngeal sample from caregiver were tested for Bordetella pertussis using PCR (IS481+/hIS1001). History and clinical details were documented. Results 460 children with a median age of 8 (IQR 4–18) months were enrolled. B. pertussis infection was confirmed in 32 (7.0%). The adjusted risk of confirmed pertussis was significantly increased if infants were younger than two months [aRR 2.37 (95% CI 1.03–5.42]), HIV exposed but uninfected (aRR 3.53 [95% CI 1.04–12.01]) or HIV infected (aRR 4.35 [95% CI 1.24–15.29]). Mild (aRR 2.27 [95% CI 1.01–5.09]) or moderate (aRR 2.70 [95% CI 1.13–6.45]) under-nutrition in the children were also associated with higher risk. The highest adjusted risk occurred in children whose caregivers had B. pertussis detected from nasopharyngeal swabs (aRR 13.82 [95% CI 7.76–24.62]). Completion of the primary vaccine schedule (three or more doses) was protective (aRR 0.28 [95% CI 0.10–0.75]). Conclusions HIV exposure or infection, undernutrition as well as detection of maternal nasal B. pertussis were associated with increased risk of pertussis in African children, especially in young infants. Completed primary vaccination was protective. There is an urgent need to improve primary pertussis vaccine coverage in low and middle-income countries. Pertussis vaccination of pregnant women, especially those with HIV infection should be prioritized. Children less than 13 years of age admitted over a one-year period (07 September 2012 to 06 September 2013) for LRTI to the Red Cross War Memorial Children’s Hospital (RCH), Cape Town, South Africa, were prospectively screened for enrolment. The hospital provides services for children up to 12 years of age. Children with WHO-defined severe pneumonia (age specific tachypnoea or/and lower chest indrawing requiring hospitalization) or apnea were eligible to be included. A child was included if the legal guardian gave written consent and the child had not been in contact with a health care facility in the previous 48 hours to two weeks prior to screening for enrolment. Enrolment was limited to the first four qualifying children per working weekday. History of symptoms of the presenting illness and information on current socio-demographic factors including type of housing, access to amenities such as tap water, electricity and toilet facilities, was taken from the caregiver. The mother’s level of education was recorded. Socio-economic status was categorized into quartiles on the basis of a validated weighted composite score used elsewhere that included asset ownership, employment and education [18]. The use of household biofuels, presence of smokers in the household and the number of people sharing the bedroom with the child were established. Information on day-care attendance was also collected. The mother’s HIV status (and that of the primary caregiver if this was not the mother) was established. If the mother or caregiver was HIV infected the latest available CD4 count was recorded and used in the staging of HIV disease according to the Centre for Disease Control (CDC) classification [19]. History was taken on the presence and duration of recent primary caregiver respiratory symptoms as well as presence and numbers of other household members with similar symptoms. The vaccination status of each child was verified using the national standardized immunization handheld record, the Road to Health Card (RTHC); specifically, the date and type of each vaccine was copied from the record. All children who missed any vaccinations that should have been received for age, where referred to an immunization catch-up program. The weight of each child, as measured on admission, was used to evaluate nutritional status using World Health Organization (WHO) weight for age Z scores (WAZ). Mild under-nutrition was defined as ≤ -1 WAZ > -2, moderate under-nutrition ≤ -2 WAZ > -3 and severe under-nutrition WAZ ≤ -3 [20]. Each child was screened for HIV infection using an ELISA test (Architect HIV Ag/Ab Combo, Abbott Diagnostics, Wiesbaden). The diagnosis of HIV infection was made if both the ELISA and an HIV PCR test (COBAS AmpliPrep/COBAS Taqman HIV-1, Roche Molecular Diagnostics, Pleasanton, CA) were positive in children younger than 18 months. A positive ELISA was confirmed with a different ELISA test (Enzygnost Anti-HIV 1/2 Plus, Siemens/Dade Behring, Erlangen) in children older than 18 months to diagnose HIV infection. Children younger than 18 months who were ELISA positive, but PCR negative were classified as HIV exposed uninfected while older children were classified as HIV exposed uninfected if the mother was HIV infected at the time of the pregnancy but the child tested HIV negative. Caregivers who did not know their HIV status were counselled and offered HIV testing. Children or caregivers with HIV who were not accessing appropriate treatment were referred to public health facilities for further follow up and treatment of HIV. Methods employed in the collection of respiratory specimens have been published [21]. Briefly, nasopharyngeal (NP) specimens from caregivers as well as paired NP and induced sputum (IS) specimens from children were tested by PCR for B. pertussis. The NP specimen was taken with a flocked nylon swab which was transported in a nucleic acid preservation medium (PrimeStore® MTM, Longhorn Vaccines and Diagnostics, San Antonio, TX). The IS specimen was collected after the NP was taken from each child. All specimens were frozen at minus 80°C until they were thawed for batched molecular diagnostic testing. A commercially validated duplex real-time PCR assay targeting the insertion sequence IS481 for Bordetella and IS1001 for Bordetella parapertussis (LightMix® Bordetella pertussis and parapertussis Kit, TIB MolBiol, Berlin, Germany) was used for screening all the respiratory specimens [22, 23]. All specimens testing positive for IS481 were further tested for the presence of insertion site hIS1001 in order to exclude Bordetella holmesii (IS481 +, IS1001-, hIS1001 +) before the diagnosis of B. pertussis infection was made [24]. Positive cases were offered macrolide treatment and the same offered to household contacts for prophylaxis. The study was approved by the Human Research Ethics Committee of the Faculty of Health Sciences of the University of Cape Town; Reference: 371/2011. Written informed consent was sought and received from the legal guardian for the participation of both the child and the guardian/caregiver in the study. The study aimed to investigate risk factors for pertussis as a secondary outcome and was thus not specifically powered to achieve this secondary outcome. The study sample was determined to attain a 3% precision above and below a point estimate risk of 5% for the primary outcome (prevalence of pertussis). Categorical data are presented as percentages with 95% confidence intervals (CI). All continuous data are summarized as medians with interquartile ranges (IQR). A χ2 test assessed strength of association between two categorical variables with a two-tailed cut-off significance set at p<0.05. A causal model employing the current understanding of respiratory disease processes and pathogenesis of pertussis was constructed using a directed acyclic graph (DAG) to identify variables in the model required for minimal sufficient adjustment sets for estimating total independent effects of each assessed risk factor [25]. To adjust for potential confounders for each risk factor as identified by the DAG, a generalized linear Poisson regression model with robust error variance was used to estimate adjusted relative risks (aRR) and their 95% level of confidence in a multivariable analysis. For all analyses, Stata Statistical Software Release 13 (StataCorp LP, College Station, TX) was employed.