The study was conducted among children aged 12–16 months residing in the Mirzapur Demographic Surveillance System (DSS), a rural area 60 km north of Dhaka, Bangladesh. The DSS is comprised of ~240,000 individuals living in 58,300 households visited every 4 months to update births, deaths and migrations. A stratified random sample of 1450 children was selected from the eight unions (administrative unit similar to a county) of the DSS with probability of proportional to eligible population size in each union. Ethics committees at the International Centre for Diarrhoeal Disease Research, Bangladesh (icddr,b) and the Johns Hopkins Bloomberg School of Public Health approved the study and written permission from the participants’ guardian was obtained prior to enrollment.
Data collection took place in three stages: a caregiver survey and oral fluid collection in the household, review of EPI clinic records, and blood collection from a subset of children in four of the eight unions. A wealth index was generated using principal component analysis based on household assets
. Record books from EPI clinics were obtained and records were matched to enrolled children based on EPI registration IDs from their vaccination cards or by matching at least three variables: child name, birth date (±3 days), father’s name or grandfather’s name.
OF samples were obtained by rubbing a foam swab (Oracol; Malvern Medical Developments, Worcester, UK) along the child’s gumline for 1–1.5 minutes as described elsewhere
. OF samples were transported in cold boxes within 4 hours to the local laboratory where 1 mL of transport buffer was added, centrifuged at 2000 rpm for five minutes, and the supernatant was stored at -20°C until testing. OF samples were tested for antibodies to measles virus using a measles virus-specific IgG capture enzyme immunoassay (ELISA) (Microimmune Ltd, Middlesesex, UK), which was validated for use with OF samples with a reported sensitivity of 93% and specificity of 98% compared to serum
. OF samples were categorized as positive, equivocal or negative; equivocal samples were retested and, if equivocal again, were considered positive in binary analyses. To identify poor quality OF samples, specimens testing negative for measles IgG were tested for total IgG antibodies by ELISA (Bethyl Laboratories, Montgomery, TX) and excluded from the analysis if less than 125 ng/mL.
Peripheral blood samples were collected by trained phlebotomists from all enrolled children with parental permission within one month of OF collection in four of eight unions. Serum samples were extracted and stored at -20°C until tested for measles IgG antibodies (Enzygnost ELISA, Siemens, Germany) with a reported sensitivity and specificity of 99.6% and 100%, respectively
. Serum samples were categorized as negative, equivocal or positive as recommended by the manufacturer. Serum samples testing equivocal were re-tested and, if equivocal again, were categorized as positive.
Up to six indicators of MCV1 history were ascertained for each child: 1) maternal report of MCV1 based on recall questions modified from the DHS
; 2) card record of MCV1 based on evidence and dates of MCV1 receipt on the child’s household retained vaccination card, if available; 3) ‘card + history’; 4) EPI record of MCV1 based on dates and evidence of MCV1 abstracted from EPI clinic record books; 5) protective levels of measles IgG antibodies in OF; and 6) protective levels of measles IgG antibodies in blood. MCV1 coverage by OF was adjusted for assay sensitivity and specificity using following equation: Padjusted = (Pobserved – (1-specificity))/(1 – [(1-specificity) + (1-sensitivity)]). Seroprotection in OF and blood was assumed to be vaccine induced because maternal antibodies should have been non-detectable is this age group and the last measles outbreak in the district occurred more than a year before study participants were born according to the WHO measles surveillance laboratory in Dhaka, Bangladesh
[27, 28]. Study participants were not age eligible for the 2010 MCV campaign.
Categorical variables were compared using chi-square tests or Fisher’s exact test and continuous variables with non-normal distributions were compared using the Wilcoxon rank sum test. MCV1 coverage estimates were generated with exact binomial confidence intervals and compared using McNemar’s test for paired samples. Bivariate logistic and log binomial models were used to calculate odds ratios and prevalence ratios, respectively. Sensitivity, specificity and kappa were calculated to evaluate within-child agreement of indicators. Analysis was conducted in Stata 11 (StataCorp LP, College Park, TX, USA).