Menu
Menu
Menu
Menu
Introduction Pneumococcal conjugate vaccines (PCVs) prevent disease through both direct protection of vaccinated individuals and indirect protection of unvaccinated individuals by reducing nasopharyngeal (NP) carriage and transmission of vaccine-type (VT) pneumococci. While the indirect effects of PCV vaccination are well described, the PCV coverage required to achieve the indirect effects is unknown. We will investigate the relationship between PCV coverage and VT carriage among undervaccinated children using hospital-based NP pneumococcal carriage surveillance at three sites in Asia and the Pacific. Methods and analysis We are recruiting cases, defined as children aged 2-59 months admitted to participating hospitals with acute respiratory infection in Lao People’s Democratic Republic, Mongolia and Papua New Guinea. Thirteen-valent PCV status is obtained from written records. NP swabs are collected according to standard methods, screened using lytA qPCR and serotyped by microarray. Village-level vaccination coverage, for the resident communities of the recruited cases, is determined using administrative data or community survey. Our analysis will investigate the relationship between VT carriage among undervaccinated cases (indirect effects) and vaccine coverage using generalised estimating equations. Ethics and dissemination Ethical approval has been obtained from the relevant ethics committees at participating sites. The results are intended for publication in open-access peer-reviewed journals and will demonstrate methods suitable for low- and middle-income countries to monitor vaccine impact and inform vaccine policy makers about the PCV coverage required to achieve indirect protection.
We are conducting prospective hospital-based observational studies in Lao PDR, Mongolia and PNG. We are recruiting children 2–59 months of age presenting with ARI and obtaining NP swabs to determine prevalence and density of VT carriage. We are determining the PCV13 status of each case using written record. Recruitment will occur over at least 3 years and up to 5 years post-PCV13 introduction at each site. Concurrently, we are determining vaccination coverage at the resident village or subdistrict of each recruited case, using either administrative data or vaccination coverage surveys. The Lao PDR PneuCAPTIVE study is embedded within a hospital-based study of ARI aetiology, in collaboration with Lao Oxford Mahosot Hospital-Wellcome Trust Research Unit, the WHO and the Lao PDR Ministry of Health.25 PCV13 was introduced in October 2013, using a 3+0 schedule at 6, 10 and 14 weeks of age and a catch-up programme up to 12 months of age (table 1). We are recruiting cases at Mahosot Hospital, one of the largest paediatric referral hospitals in Vientiane, the capital of Lao PDR. Recruitment started in December 2013. Key aspects of the 13-valent pneumococcal conjugate vaccination (PCV13) programme by site, Lao People’s Democratic Republic (PDR), Mongolia and Papua New Guinea (PNG) The Mongolian PneuCAPTIVE study is embedded within a hospital-based paediatric pneumonia surveillance program to determine vaccine impact, which is conducted in partnership between the Murdoch Children’s Research Institute (MCRI), WHO and the Mongolian Ministry of Health. PCV13 was introduced in June 2016, using a modified 2+1 schedule at 2, 4 and 9 months of age, within the two ‘Phase 1’ districts within Ulaanbaatar, the capital of Mongolia as part of a phased introduction (table 1), with a catch-up programme of two doses 1 month apart for those up to 24 months of age. We are recruiting cases, residing in the two phase 1 districts, at the two district hospitals and the tertiary referral paediatric hospital for Mongolia, the Maternal and Child Hospital. Sampling started November 2015. The PNG PneuCAPTIVE study represents a collaboration between the PNG Institute of Medical Research, Telethon Kids Institute, the University of Western Australia and MCRI. It is an extension of a pneumonia aetiology study that commenced in 2013.26 PCV13 was introduced to PNG in October 2014 using a 3+0 schedule at 1, 2 and 3 months of age (table 1); however, it was not widely distributed in the Eastern Highlands Province until late 2015. We are recruiting cases at the Eastern Highlands Provincial Hospital, the major referral hospital for the Eastern Highland Province, as well as nearby clinics in Goroka, the capital of the Eastern Highlands Province. Recruitment started April 2016. In PNG, we are also recruiting caregivers, as well as contacts, defined as children 0–59 months of age, who have slept in the same house as or played with the case during the preceding 3 weeks. This will enable us to determine whether changes in patterns of VT pneumococcal carriage in the hospitalised cases are reflective of changes within the community, as well as to examine indirect effects in the adult age group. Participant recruitment and data collection are consistent across the three sites; however, there are some local adaptations to the protocol at each site, which are summarised in table 2. These adaptations are due to the PneuCAPTIVE study being nested within other existing studies, described above. Patient eligibility by site: Lao People’s Democratic Republic (PDR), Mongolia and Papua New Guinea (PNG) *Tachypnoea is defined as ≥50 breaths per minute. †A 1-year extension (until June 2019) has been sought for the Mongolian site. Cases are eligible for inclusion in the PneuCAPTIVE study if they are 2–59 months of age and presenting with ARI (defined in table 2 below). All cases with fever or respiratory symptoms are screened for inclusion. In Mongolia, we are restricting recruitment to patients living within the two ‘Phase 1’ districts that have commenced PCV13 in 2016. In PNG, we are restricting recruitment to patients living within 1 hour of the town as follow-up is logistically challenging. Regarding recruitment, in Lao PDR, study staff are screening potential recruits from Monday to Friday each week. However, they are obtaining clinical information from medical records for all eligible cases, including those admitted at weekends to ensure we have a representative sample. In Mongolia, caregivers of all eligible children presenting to the hospital will be approached for recruitment as part of the larger PCV impact study. However, for the purposes of this study, we will select a random sample of 33 cases per month for microbiological testing. In PNG recruitment takes place 4–5 days per week. After determining eligibility and obtaining informed parental consent, we complete a questionnaire to obtain: demographic data, clinical data, PCV13 status and risk factors for vaccination and NP carriage, including prior antibiotic use (see analysis section below for complete list). Vaccination status is determined using written records, either parent-held immunisation records or health centre administrative records. We then collect an NP swab according to WHO guidelines and store it in 1 mL skim milk tryptone glucose glycerol (STGG) medium.27 Swabs are vortexed, aliquoted and stored frozen at −80°C within 8 hours of collection and transported from all three sites to the Pneumococcal Research laboratory at MCRI on dry ice or in liquid nitrogen, where they will be stored at −80°C. All samples are screened for the presence of pneumococci using real-time quantitative PCR (qPCR) assay targeting the pneumococcal lytA gene.22 Genomic DNA are extracted from 100 µL of STGG using a MagNA Pure LC Machine (Roche) using the DNA Isolation Kit III (Bacteria, Fungi) (Roche) following an enzymatic lysis treatment. The pneumococcal load is estimated by reference to a standard curve. All swabs that are lytA positive or equivocal are molecular serotyped using BμG@S Senti-SP v1.5 microarray (BUGS Bioscience) as previously described.28 Serotype-specific pneumococcal density is calculated using the relative abundance of each serotype identified, as determined using microarray and interpreted with the assistance of a Bayesian random effects model as previously described,28 29 and the overall pneumococcal load as determined by the lytA qPCR. The primary outcome, VT carriage, is defined as the NP carriage of at least one pneumococcal serotype included in the PCV13 vaccine, that is, serotypes 1, 3, 4, 5, 6A, 6B, 7F, 9V, 14, 18C, 19A, 19F and 23F. In the context of multiple serotype carriage, VT carriage will be defined as the presence of at least one VT serotype regardless of the presence of other serotypes. A secondary outcome is VT carriage density (CFU/mL), which is defined as an aggregate of the serotype-specific density for each of the VTs carried by the case, and will be reported as a continuous variable using a log scale to account for large variations in density and a skewed distribution. Vaccine history will be defined based on documented evidence of receiving an adequate number of PCV doses to provide a protective immune response against vaccine serotypes at least 14 days prior to study enrolment.30 For children <12 months of age, ‘vaccinated’ is defined as two or more PCV13 doses. For children 12 months of age or older, ‘vaccinated’ is defined as receipt of two doses in the first year of life or at least one dose after the age of 12 months. Conversely, a case will be defined as ‘undervaccinated’ if they have received less than the adequate number of PCV doses (including those never vaccinated). Sensitivity analyses will be conducted using varying definitions of vaccinated including receiving at least one dose of vaccine at any age. For each undervaccinated case recruited (including those with and without VT pneumococci), we are determining their resident village or subdistrict vaccination coverage. In Mongolia and Lao PDR, we are using administrative data. The resident village or subdistrict is identified using relevant administrative codes, which are determined by local staff on enrolment. This determines the health centre’s administrative boundary for the provision of immunisation services, whereas in PNG, we are conducting community surveys within 10 days of discharge in the village where the case is living (table 3). We are surveying all children less than 5 years of age, from households within 10–20 min walk of the case, since this is the group of children with whom the case is mostly likely to interact and therefore influence their carriage status. Vaccination coverage data by site, Lao People’s Democratic Republic (PDR), Mongolia and Papua New Guinea *All children are required to be registered at the health centre servicing their resident subdistrict in order to receive health services. To determine the reliability of our methods in Lao PDR, we plan on comparing our coverage estimates with a National Immunisation Survey, conducted according to WHO guidelines in 2015, when provincial level estimates from this survey become available. In Mongolia, we have validated the newly introduced electronic immunisation record against clinic health records, finding a high degree of concordance.31 We are also in the process of validating the population registers at health centres in Mongolia. In Lao PDR and Mongolia, study staff are double-entering data using Research Electronic Data Capture and Microsoft Access (Microsoft Corporation) database, respectively. In PNG, data are checked by a monitor prior to being entered into Filemaker Pro (FileMaker). We are conducting regular double-entry discrepancy checks and logic checks using Stata Statistical Software. All analyses will be completed using Stata Statistical Software. We will summarise continuous variables using mean and SD (or median and IQR for non-symmetrical data). Categorical variables will be summarised using frequency counts and percentages. We want to investigate the relationship between VT carriage and density among undervaccinated cases (indirect effects) and subdistrict/village PCV coverage. We will use an adaptation of a method used to estimate indirect protection for an oral cholera vaccine, which exploits heterogeneities in vaccine coverage at the subdistrict/village level, comparing VT carriage and density among undervaccinated children from subdistricts/villages with differing levels of vaccine coverage.32 33 This will be done using multivariable models with VT carriage or density in ARI cases as the outcome variable and PCV coverage at the child’s place of residence at the time of admission as the exposure variable. We will use generalised estimating equations to account for clustering at the subdistrict/village level. To identify confounders for adjustment, we have constructed a directed acyclic graph (DAG). DAGs include all variables potentially related to exposure and outcome, connected using unidirectional arrows showing causal relationships between variables. The graph identifies potentially confounding pathways and allows investigators to determine variables that should be controlled for to obtain unbiased effect estimates. As there are likely to be unique confounders between sites, we will develop site-specific DAGs. We will use DAGitty.net (V.2.3) software to identify minimally sufficient confounding subsets for adjustment. For each site, we will construct a similar model using overall pneumococcal carriage as the dependent variable. This model will act as a bias indicator since PCV coverage is not expected to affect levels of overall pneumococcal carriage due to replacement carriage with NVTs, although complete replacement to baseline levels can take several years.21 34 35 Therefore, we will restrict this analysis to the latter part of the study period, when descriptive analyses indicate the replacement is complete. To determine whether a higher PCV coverage is required to achieve indirect effects among completely unvaccinated cases compared with undervaccinated cases, we will conduct a sensitivity analysis among children who have never received PCV. Crude and adjusted monthly VT carriage prevalence will be estimated within rolling 7-month intervals to present smooth curves and assess trends over time. This will be done separately for cases and contacts. To account for differences in age between cases and community contacts, the carriage prevalence will be adjusted using direct standardisation (standardised to the case population over the entire study period). To investigate whether relationship between vaccine coverage and indirect effects as observed among cases with ARI are reflective of the relationship between vaccine coverage and indirect effects in the wider community, we will apply the same model described above to undervaccinated community contacts. We will construct multivariable models with VT carriage and density among undervaccinated community contacts as the outcome variable and PCV coverage at the child’s place of residence at the time of admission as the exposure variable. We will be using the same vaccine coverage data as for the cases. We will compare differences in the PCV13 coverage required to demonstrate indirect effects of PCV13 qualitatively by site and in relation to vaccine schedule and use of catch-up campaigns. Inferential statistics are unlikely to be suitable with the inclusion of only three sites, and comparability between sites is limited due to variations between them. Power calculations were performed using nQuery Advisor+nTerim 4.0. Calculations were based on sample size methods for logistic regression models with a continuous covariate (ie, PCV coverage) and additional covariates, with inflation to account for clustering within villages. Power calculations assumed VT carriage prevalence of 30% in Lao PDR and 40% in Mongolia and PNG at the mean PCV coverage level36 and VT carriage prevalence of 20% in Lao PDR and 30% in Mongolia and PNG at 1 SD above the mean PCV coverage level. Assuming a significance level of 0.05, allowing for adjustment using multiple covariates with an R2 of 0.4, and that 50% of the cases are undervaccinated, a sample size of 1200 cases per site would provide between 87% and 92% power to determine the proportion of cases carrying VT pneumococcus at varying levels of village vaccine coverage. The power calculation has been adjusted to account for clustering by village, with higher variability in Lao PDR and PNG (intraclass coefficient (ICC) 0.1) and lower variability in Mongolia (ICC 0.01). We will describe the number of participants with missing data on individual variables and compare the characteristics of those with and without missing data to determine whether there is evidence of systematic differences in characteristics. If we determine that the differences observed are able to be explained by available data (ie, missing at random), we will consider using multiple imputation to predict the distribution of the missing data in order to account for the bias due to incomplete data.37 Patients were not involved in the design of this study. Public health authorities in Laos and Mongolia were involved in the design and conduct of the study. In PNG, village representatives are approached to ensure that community surveys are conducted appropriately. Recruitment is ongoing, and analysis of data, followed by publication of results, is expected from 2018 onwards. As of August 2017, we have recruited 1039, 481 and 3847 cases from Lao PDR, PNG and Mongolia, respectively. Table 4 describes the baseline characteristics of the children recruited. Case characteristics by site, Lao People’s Democratic Republic (PDR), Mongolia and Papua New Guinea (PNG), 2014–2017 Prospective participants will be fully informed about the potential risks and benefits of participation, and written informed consent will be obtained prior to recruitment. We plan on disseminating results to relevant stakeholders within Lao PDR, PNG and Mongolia, as well as submitting our findings for publication in relevant peer-reviewed journals and conferences.
