Incidence of severe and nonsevere pertussis among HIV-exposed and-unexposed zambian infants through 14weeks of age: Results from the southern Africa mother infant pertussis study (samips), a longitudinal birth cohort study

Background. Maternal vaccination with tetanus, reduced-dose diphtheria, and acellular pertussis vaccine (Tdap) could be an effective way of mitigating the high residual burden of infant morbidity and mortality caused by Bordetella pertussis. To better inform such interventions, we conducted a burden-of-disease study to determine the incidence of severe and nonsevere pertussis among a population of Zambian infants. Methods. Mother-infant pairs were enrolled at 1 week of life, and then seen at 2-to 3-week intervals through 14 weeks of age. At each visit, nasopharyngeal (NP) swabs were obtained from both, and symptoms were catalogued. Using polymerase chain reaction (PCR) to identify cases, and a severity scoring system to triage these into severe/nonsevere, we calculated disease incidence using person-time at risk as the denominator. Results. From a population of 1981 infants, we identified 10 with clinical pertussis, for an overall incidence of 2.4 cases (95% confidence interval [CI], 1.2-4.2) per 1000 infant-months and a cumulative incidence of 5.2 cases (95% CI, 2.6-9.0) per 1000 infants. Nine of 10 cases occurred within a 3-month window (May-July 2015), with highest incidence between birth and 6 weeks of age (3.5 cases per 1000 infant-months), concentrated among infants prior to vaccination or among those who had only received 1 dose of Diphtheria Tetanus whole cell Pertussis (DTwP). Maternal human immunodeficiency virus (HIV) modestly increased the risk of infant pertussis (risk ratio, 1.8 [95% CI, .5-6.9]). Only 1 of 10 infant cases qualified as having severe pertussis. The rest presented with the mild and nonspecific symptoms of cough, coryza, and/or tachypnea. Notably, cough durations were long, exceeding 30 days in several cases, with PCRs repeatedly positive over time. Conclusions. Pertussis is circulating freely among this population of Zambian infants but rarely presents with the classical symptoms of paroxysmal cough, whooping, apnea, and cyanosis. Maternal HIV appears to increase the risk, while lack of effective exposure to DTwP increased the risk. To provide methodological consistency, the Zambia site harmonized its screening case definition, nasopharyngeal (NP) swabbing procedures, polymerase chain reaction (PCR) testing processes, and pertussis severity scoring systems with the Pakistan and South African sites (see Omer et al and Nunes et al in this supplement). SAMIPS was a longitudinal birth cohort. The SAMIPS study population consisted of mother–infant pairs from the Chawama compound, a large informal periurban slum located to the southwest of Lusaka’s city center. Chawama compound measures roughly 30 km2 and is home to a population of approximately 142 000 persons. The Chawama Primary Health Clinic (PHC) is the only government-supported clinic in Chawama compound and the predominant source for all medical care in this community. Housing our study at the Chawama PHC put us at the nexus of nearly all primary care in the compound. The institutional review boards at Boston Medical Center and Excellence in Research Ethics and Science Converge in Lusaka jointly provided ethical oversight. All mothers provided written informed consent, with consent forms presented in English and the 2 dominant vernacular languages spoken in Lusaka: Bemba and Nyanja. As this was an observational study, it did not need to be registered at ClinicalTrials.gov. Enrollment was limited to mothers who signed consent and agreed to all procedures, anticipated remaining in the Chawama catchment area for the next 3 months, and granted us access to records documenting their HIV status. To minimize mortality-related attrition in this longitudinal cohort, infants were limited to those who were deemed healthy, were not premature (<37 weeks’ gestation) or underweight (<2800 g), did not result from a complicated pregnancy or delivery, and were within their first 10 days of life. Mother–infant pairs were enrolled at the first postpartum scheduled well-child visit. Baseline data were collected on the mother and infant, including maternal HIV status and CD4 counts if available, maternal age, household composition, infant birth weight, gestational age, and other factors. The study did not measure CD4 counts for HIV-infected mothers, but only used data already available at the Chawama PHC HIV clinic. We did not do confirmatory HIV testing but instead relied on testing previously done at the clinic. Maternal pertussis vaccination status could not be determined given the absence of such records and given that asking mothers to recall their infant vaccination status was unrealistic. While they might have given answers, there would be no way to verify them, and we had little confidence that the subjects could reasonably be expected to know the answer. It was not practical to obtain baseline blood samples from all 2000 mothers to allow retrospective serologic analyses around the subset of infants who might later develop pertussis. At baseline and thereafter at 2- to 3-week intervals, both members of the mother–infant pair underwent NP sampling. Symptom data were solicited in parallel using a standardized checklist to characterize the subject at each sampling point. The goal of scheduling the visits and obtaining samples irrespective of symptoms was to minimize sampling bias. Ad hoc sick visits also resulted in NP swabs of both members of the pair if the visit was triggered by respiratory complaints in either mother or infant. Clinical data and NP sampling was performed by members of the SAMIPS study team (either a registered nurse or a clinical officer). Infants received scheduled vaccines at 6, 10, and 14 weeks with the pentavalent vaccine (diphtheria/tetanus/whole-cell pertussis, Haemophilus influenzae type b conjugate, and hepatitis B) (Pentavac, Serum Institute of India Limited, Pune, India), the 13-valent pneumococcal vaccine, oral rotavirus vaccine, and oral polio vaccine (OPV). BCG and a first dose of OPV are given at birth, but this occurred prior to enrollment. The fidelity of the sample identification rested on bar codes to uniquely identify subjects and subject data linked to bar codes to uniquely identify samples. The subject and sample ID bar codes were scanned at the time of collection using the Xcallibre digital pen system, which was also used to capture data electronically. For every mother–infant pair, we custom printed 4000 subject ID sticker books, consisting of 50 identical study ID barcode labels per subject (half printed in pink as XXXX-0 for mothers and half in green as series XXXX-1 for infants), and affixed to each case report form as needed. Samples were labeled using a set of 35 000 unique sample ID barcodes, printed in duplicate: 1 copy for the case report with symptoms data and the second for the sample tube. Using this chain of barcodes, we linked subjects to samples with digitally recorded dates, and mated these to each individual's symptom data. NP swabs were obtained using flocked-tipped nylon swabs (Copan Diagnostics, Merrieta, California), sized for adults or infants as indicated. In standardized comparative studies vs comparator NP swabs, flocked-tipped nylon swabs yielded a higher rate of culture positivity, had higher colony-forming unit densities, and yielded higher DNA concentrations on quantitative PCR compared with Dacron or Rayon swabs [11, 12]. Swabs were inserted into both nares until they contacted the posterior nasopharynx and were rotated 180 degrees in both directions. The swabs were then placed in commercially prepared tubes with universal transport media (UTM) and stored on ice until transport. Samples were collected from the study clinic each day and taken to the PCR laboratory at the University Teaching Hospital (UTH) and stored at −80°C until PCR testing. Our primary analyses were conducted using the diagnostic testing algorithm developed and validated by the respiratory pathogens group at the US Centers for Disease Control and Prevention (CDC) (Supplementary Tables 1A and 1B). DNA was extracted using the NucliSENS EasyMag system (bioMérieux, Marcy l'Etoile, France) [13]. Pathogen detection was done using a TaqMan genomic assay using the AB7500 Fast Real-Time PCR system (Applied Biosystems, San Francisco, California). Testing starts with a pair of singleplex reactions testing for the targets IS481 and ptxS1. IS481 is the most common insertion sequence in B. pertussis, with multiple copies per genome, making it a very sensitive target for screening [14, 15]. By contrast, ptxS1, coding for pertussis toxin, usually exists as a single or occasionally double copy, making ptxS1 highly specific but less sensitive [16–18]. Because these primers/probes have different annealing temperatures, they were run in parallel on separate 96-well plates. If either IS481 or ptxS1 was positive, DNA was reextracted and a multiplex PCR reaction conducted repeating the tests for IS481 and ptxS1, and now including primers/probes specific to Bordetella parapertussis (PIS1001) and Bordetella holmesii (HIS1001). All primers and probes were purchased from Life Sciences Solutions (a subsidiary of Fisher Biosciences). This paper only provides results for the B. pertussis reactions. Pseudo-outbreaks of pertussis due to accidental contamination of NP swabs at the time of collection, or subsequently in the laboratory, have been reported frequently in recent years [19, 20]. The leading cause of contamination is during sample collection due, ironically, to pertussis vaccines. All wP vaccines used around the world include pertussis DNA, but so too does the leading US multivalent vaccine Pentacel (Sanofi Pasteur). To minimize contamination of our NP swabs during collection, our clinic did not store or administer any vaccines. For infants who required routine vaccinations, these were administered only after all study data and sample collections were complete and then were administered at a location roughly 50 meters away from our clinic, accessed via a separate building entrance. To exclude contamination in the laboratory, all PCR runs included a positive and negative control, the former to confirm that the PCR reaction was successful and generating consistent results across runs, the latter to screen for environmental contaminations within the laboratory. To ensure fidelity of the NP swabs, every patient sample was tested with a primer/probe against the human gene RNAse P. Its product is a constitutive enzyme secreted by all human cells, and therefore tests whether the swab made effective contact with the respiratory mucosa. For our starting point, we referred to the recently completed Pneumonia Etiology Research in Child Health (PERCH) study. PERCH was a 7-country epidemiologic surveillance study of severe pediatric pneumonia, which included the Lusaka, Zambia, site. PERCH was a hospital-based case-control study, and defined its “cases” as children, aged ≥6 weeks, presenting with a clinical syndrome compatible with severe or very severe pneumonia per World Health Organization (WHO) criteria. Of these, 356 PERCH children were aged 1–6 months, and therefore germane to the SAMIPS estimates. A total of 20 PERCH infants with severe pneumonia tested positive for pertussis by PCR, and 14 of 20 (70%) were aged 1–3 months. Overall, we observed that pertussis accounted for roughly 4% of severe pneumonia in Lusaka infants 1–3 months of age. We extrapolated rates from these data to estimate a population incidence in the relevant age category for SAMIPS using the following assumptions: Given that this was a hospital-based cross-sectional study, not a true population-based survey, we adopted a more conservative assumption that incidence could be 3 times lower than implied by PERCH. Therefore, taking instead a rate of pertussis of 2 cases per 1000 infants per month as a plausible lower incidence bound, with a margin of error of ±0.2% as the width of the desired confidence interval, and using the sample size formula for a single proportion: then 1914 subjects would detect this incidence rate for pertussis with 95% confidence. Rounding up, our target sample size was 2000 mother–infant pairs. We defined cases as an infant presenting with any of the signs or symptoms on our screening form with a positive PCR result per CDC criteria. Note that this is distinct from the CDC case definition used in routine surveillance in the United States. That definition can be met in 1 of 2 pathways: (1) if an infant has a cough of any duration with microbiological culture confirmation; or (2) if an infant has 2 weeks of cough plus classic symptoms of pertussis (whooping, paroxysms, apnea, posttussive vomiting). In so doing, the CDC's definition increases the specificity of detection while sacrificing sensitivity relative to our screening case definition. Prior to the study start, in discussions with the scientific advisory group, it was decided that PCR was sufficiently persuasive if following the CDC's protocol (which we were). Hence, cultures were not obtained, meaning that there would be no way to satisfy the first pathway. The second pathway is optimized to identify classic pertussis. We note that infants presenting with these symptoms would likely be classified as “severe pertussis” using the Preziosi scale (see below), conflicting with our objective of also measuring nonsevere pertussis. For our incidence calculations, we used person-time as the denominator, with infants with PCR-confirmed symptomatic pertussis as the numerators. Positive cases were further classified as severe/nonsevere pertussis using the pertussis severity scoring system developed by Monica Preziosi at the WHO [21–23]. The Preziosi Scale was developed for older children, not infants. Because infants may present with different symptoms than older children, we created a Modified Preziosi Scale (MPS) for use among the infants <6 weeks of age. Symptomatic pertussis was defined as a positive PCR result (per CDC criteria) in the presence of any of our screening symptoms. Nonsevere pertussis was defined as an MPS score of 0–6 points; severe is >6 points. This process is summarized in Figure ​Figure1.1. The MPS is included as Supplementary Table 2. Relationship between the total infant population, the symptomatic population, and those with polymerase chain reaction (PCR)–confirmed pertussis and how these were triaged by severity using the Modified Preziosi Scale (MPS). The enrolled population consisted of those who had at least a baseline visit and NP swab. No imputation was performed for missing data. In summaries, missing data did not contribute to the denominators in means and percentages. Subjects were analyzed to the extent that they made study visits. The following endpoints were calculated: the incidence rate of pertussis (all) and severe and nonsevere pertussis (separately) defined as the number of cases divided by the total person-time per 1000 months. In stratified analyses, we calculated the contribution of maternal and infant characteristics on the above measurements. This included maternal HIV serostatus, CD4 count, and infant ages in months. Additionally, we calculated incidence as a function of the number of pentavalent vaccine doses administered prior to onset of symptoms. For this last analysis only, we defined cases as occurring “postvaccination” if they occurred at least 2 weeks after the latest vaccination, thereby granting sufficient time for the infant to have mounted an immune response. All data manipulations and statistics were performed using SAS software, version 9.4 (SAS Institute, Cary, North Carolina).