Can BCG Vaccination at First Health-facility Contact Reduce Early Infant Mortality?

Background:

The World Health Organization (WHO) recommends Bacillus Calmette-Guérin (BCG) vaccination at birth to protect against TB in countries with high TB burden. Though the BCG vaccine is scheduled at birth, the vaccine is often given with delay, also among children born in Health Facilities (HFs), where it should be possible to obtain the vaccine. In 2010 in rural Guinea-Bissau, only 38% of children received BCG within the first month of life. BCG is supplied in 20-dose vials and unused doses must be discarded 6 hours after reconstitution. This has led to a local practice of not opening a BCG vial unless sufficient eligible children are present for vaccination. Moreover, the HFs seek to assemble larger groups of children eligible for BCG at specific days. Our observations from the field indicate that even on these days, vials may not be opened if the number of children is too low. Similar practises have been demonstrated in other low-income countries, but this practice is not in line with WHO recommendations, which emphasise that multi-dose vials should be opened and used to vaccinate one child despite any wastage.

In 2012, an estimated 73% of all neonatal deaths occurred within the first week of life. Several studies have suggested that BCG vaccination is associated with survival benefits early in life: In three randomised trials from Guinea-Bissau, BCG-at-birth was associated with a 38% lower neonatal mortality and marked reductions already 3 days after vaccination. A WHO-commissioned review and meta-analysis conducted in 2014 including five clinical trials indicated that BCG-vaccinated children had a relative risk of mortality at 0.70 (95% confidence interval 0.49-1.01) compared with BCG-unvaccinated children. Two randomised trials from India, both published after the review was conducted, found no effect of BCG at birth on neonatal mortality. In both India and Guinea-Bissau, all trials were conducted among low birth weight (LBW) children and primary outcome was neonatal mortality. In Guinea-Bissau, BCG-Denmark was used in all trials, whereas BCG-Russia was used in the trials from India. Strain of BCG has been suggested as a possible reason for the discrepancy between the trials in India, and the trials in Guinea-Bissau.

Therefore, early BCG vaccination may be important, not only for protection against tuberculosis, but also for the impact on survival.

Hypothesis:

Increasing the availability of BCG and vaccinating children at the first health-facility contact can reduce early infant mortality due to other causes than accidents by 25%.

Methods:

This study will be a cluster randomized trial of increased BCG availability at the 23 HFs in three rural regions (Oio, Biombo and Farim) in Guinea-Bissau. In half of the HFs, BCG will be provided as per current practice (typically once a week if a sufficient number of children are present for vaccination); in the remaining HFs, additional BCG vaccines will be supplied to make BCG available every day and opening a vial of BCG if there is just one eligible child present. In collaboration with the national vaccination programme, the investigators will ensure that the same strain of WHO-prequalified BCG vaccines will be used at the same time in both the control and intervention arm of the study. The allocation of centres will be crossed over after 12 months and inclusion of participants will carry on for another 12 months. Pregnancies, births and deaths will be monitored through two different levels of surveillance:

Level 1 - Bandim Health Project's (BHPs) rural Health and Demographic Surveillance System (HDSS). The BHP teams survey women of fertile age and children below the age of 5 years in randomly selected village clusters in all health regions across the country, which involves 40 village clusters distributed across the three regions included in this study. Through this system, it is possible to monitor pregnancies, births, vaccinations and child health.

Level 2 - Reinforced community health worker monitoring. In all villages in Guinea-Bissau, community health workers (CHWs) monitor births and deaths through a national monitoring system. The CHWs report aggregated data on pregnancies, number of births and neonatal deaths (in two categories: 0-7 days, 8-28 days) in their capture area to the local health centre at monthly meetings. This data collection will be reinforced by supervision and further data collection from BHP supervisors covering each health centre area. The CHWs will receive a visit every 1-2 months from a supervisor.

All children registered during pregnancy in Oio, Biombo or Farim enter the trial cohort 1 day after birth. A pregnancy can be registered in more than one village, but the child will only enter the trial if the mother gave birth in the village or the nearby health facility, and then was discharged to the village where the pregnancy was registered.

For all registered deaths, a specially trained field worker will visit the household of the deceased child to conduct a verbal autopsy.

Prior to study start, all 23 HFs in the three regions will be visited by a team from BHP. During these visits, refresher training on vaccination technique and assessment of lymph glands will be conducted. The training will include general information on the BCG vaccine, information on reconstitution and durability of the vaccines, how to check that the vials have not been damaged by heat and information on adverse events. During the trial, the HFs will be visited regularly by a nurse from BHP, who will observe and supervise the vaccination sessions for vaccination technique, attention to registration of adverse event, used strain of vaccinations, registration of cold chain and state of vaccine stock.

Statistical analyses:

The primary outcome will be non-accidental mortality rate between 1 day after birth and 42 days after birth, where other vaccines are scheduled to be given. 22,800 infants are expected to be included in the study during the two-year study period. With a baseline mortality of 2.5% (varying between 1.5 and 3.5% by HF) this sample size should result in 88% power to demonstrate an effect if the true mortality reduction is assumed to be 25% (power simulations based on a logistic regression with Generalised Estimating Equation (GEE)-based correction for village cluster). Secondary outcomes are non-accidental hospital admissions, non-accidental neonatal mortality and cost-effectiveness of making BCG available at the first health-facility contact. The effect of making BCG available at the first HF contact will be analysed in logistic regression models with GEE-based correction for village cluster.