Study population and study design
The population consisted of pregnant women attending antenatal clinic and prevention of mother-to-child transmission (ANC-PMTCT) services in 30 sentinel sites national representative in Rwanda in 2011 (Fig. 1). All sentinel sites included in the study and were public health centers. Apart from Kigali City, which was represented by 3 urban sites, each province had at least two urban and two rural sites. Site selection was not randomized for the survey. All 15-49 years pregnant women presenting for their first time at the current pregnancy for ANC-PMTCT services during the surveillance period, able to provide consent, and agreeing to donate venous blood for HBV and HIV tests were included in the study. Women transferred from other health facilities and those coming for follow-up ANC-PMTCT services were not considered in the study.
The Rwanda Biomedical Centre (RBC) maintains a national monitoring and evaluation programme of HIV care services. This was a cross-sectional study, expanding on the RBC surveillance programme through additional surveys, where all women were consecutively enrolled. Each health center has an expected sample size calculated based on the previous HIV prevalence. Once the expected sample size was reached at each health center, the recruitment was stopped. The period of data collection was from May to November 2011. The estimated sample size was 13,267.
Data collection
A Five-day training was held to cover standard operating procedures following the approved protocol and laboratory techniques related to the surveillance activities. Three local staff members per health center were trained, including two ANC staff and one laboratory technician. A surveillance form was completed for all eligible women who consented to participate. ANC-PMTCT staff administered the form. Blood specimen were collected by venipuncture for routine HIV rapid test and surveillance-related laboratory tests. The specimens were placed in an ethylene-diamine-tetra-acetic acid (EDTA) tube and labeled using the PMTCT code in order to facilitate the return of laboratory test results for clinical care. Blood specimens were centrifuged; plasma was aliquoted into cryotubes at each health center for HIV rapid testing to allow quick clinical care of HIV-positive people. Specimens for surveillance purpose were labeled and stored in a freezer at a temperature of −10° to −20 °C until they were sent to the National Reference Laboratory (NRL) in Kigali. Sentinel sites staff were supervised at least twice monthly by the RBC surveillance staff. Frozen samples were transported by the site supervisors in insulated boxes containing ice packs replenished regularly.
All HBV testing was performed at the National Blood Transfusion Center in Kigali. The testing for HBsAg used the Abbott ARCHITECT system which included reagents, calibrator, control packs, pre-trigger solution and conditioning solution for the assays. Its sensitivity was 99.80% (95% confidence interval 98.90–99.99%) and specificity 100% (95% confidence interval 98.59–100%). HBsAg testing was performed using Architect HBsAg quantitative reagent kit, Architect HBsAg quantitative calibrator and Architect HBsAg quantitative controls. Total HBsAg were tested using the Vitros Chemiluminescence Immuno-assays (Ortho Clinical Diagnostics, Rochester, NY).
Hepatitis B e antigen (HBeAg) testing to detect replication of HBV was applied to HBsAg positive specimens only. This test was performed using ARCHITECT i2000SR, HBeAg Calibrator, control packs, pre-trigger solution and probe conditioning solution for the assays. HBcAb-IgM to detect recent HBV infection (within 6 months) was also applied to HBsAg positive samples only. This test was performed using also ARCHITECT i2000SR, HBcAb-IgM Calibrator, control packs, pre-trigger solution and probe conditioning solution for the assays.
HIV ELISA-based tests were performed at NRL. The screening test used was the HIV Vironostika Uni Form I Ag/Ab, 4th generation. It was in turn confirmed using the Murex HIV Antigen/antibody combination test. All samples with non-reactive results to the screening test were considered negative. Samples reactive to screening test were confirmed by confirmatory test. Samples were considered as definitively positive if they were reactive to both screening and confirmatory tests. If there was any discrepancy between the screening and confirmatory tests, samples were confirmed using Enzygnost and were considered as positive samples if reactive to Enzygnost.
Data management and analysis
Surveillance forms and laboratory results were double entered using Epi Info 7.0.9.7 (CDC, Atlanta, GA, USA) to minimize errors. To ensure data quality, data verification, and daily back up were performed. All survey data were also backed up on the RBC network server on a daily basis. Data were analyzed using STATA 13 software (StataCorp LP, 4905 Lakeway Drive, College Station, TX, USA). A mixed-effects logistic regression model where the health facility treated as random effects was used to account for the survey design and unknown selection probabilities. Odds ratios were used to measure the magnitude of the effect of HBV in the bivariate analyses of potential risk factors and factors associated with HBV infection at the ≤0.25 significance level were included in the multivariable analysis but kept in the final model only those which were associated at the <0.05 significance level. The following variables were dichotomized as follows age (<25 years and older), marital status (in union or not in union), and number of pregnancies (≤2 or >2). The prevalence of HBsAg and HBeAg among pregnant women was estimated with 95% confidence intervals (95% CI). Model selection for the multivariable model was conducted by minimizing the Akike’s information criterion (AIC) whereby the selected model optimizes both fit and model simplicity. Only HBV and HIV co-infection was examined for associations. The effects of missing values on key variables used in the regression analyses were assessed by fitting indicator variables for missingness for each variable with the outcome; these were tested using maximum likelihood test association with the outcome (missing at random). Missing value records were not included in the analyses because the number of fields and variables with missing data was small. In addition, there was no association between missing data and HIV status. Therefore any bias introduced by this approach will be small.
Ethical considerations
Blood specimen and interview-based data collection was performed for eligible participants who provided signed informed consent. The protocol was reviewed and approved by the Centers for Disease Control and Prevention and by the Rwandan National Ethics committee. All results were returned to health facilities for appropriate care of patients in case of positive results.