A Randomized Trial of Providing BCG Vaccination Immediately

Observational studies in high-mortality areas have shown that the BCG vaccine is associated with marked benefits in child survival. These results cannot be explained by prevention of tuberculosis, which is rare during the firsts years of life. These observations have recently been tested at the Bandim Health Project (BHP) in Guinea-Bissau in randomized trials among low birth weight (LBW, <2500 g) children. According to local practice, BCG is typically postponed in LBW infants until 4-6 weeks of age. Results showed that providing BCG at the time of discharge from the hospital compared with BCG later reduced neonatal mortality by more than 40%. The protective effect started already within the first three days. Based on verbal autopsies, BCG protected the children against neonatal septicemia and pneumonia. Recent immunological studies have indicated that BCG induces epigenetic changes which reprogram the innate immune system to increased pro-inflammatory responses against unrelated pathogens; hence, BCG may indeed have a non-specific beneficial effect protecting against other infections.

BCG is recommended at birth by the World Health Organization (WHO), but is rarely given immediately at birth. In Guinea-Bissau, at the National Hospital's maternity ward, children are currently vaccinated at discharge. For children who are born healthy, this usually means 1-2 days after the delivery. However, very small children and ill children are admitted to the neonatal intensive care unit (NICU), and often stay there for days to weeks before they are discharged - and only receive BCG at the time of discharge. The mortality among children in the NICU is alarmingly high - it was 14% (254/1877) among children admitted to the NICU from November 2007 to May 2013 (unpublished data). A substantial number of children that died had normal birth weight and high Apgar scores. In most cases the underlying cause of death was infectious diseases, which could potentially be affected by BCG vaccination. Hence, the present aim is to conduct a randomized trial at the NICU to study the effect of providing BCG immediately (intervention group) versus BCG at discharge (control group). The primary outcome will be mortality during hospital stay, with cause of death, growth and length of stay as secondary outcomes.

HYPOTHESIS

Providing BCG vaccination upon admission to newborn children in the NICU incubators, rather than at discharge, is associated with 40% reduction in mortality at the NICU for children weighing >1250 g.

METHODS

Setting

The randomized trial will be conducted by the Bandim Health Project in Guinea-Bissau, at the NICU at the National Hospital Simão Mendes. The Bandim Health Project (BHP) maintains a health and demographic surveillance system (HDSS) in six suburban districts of the capital of Guinea-Bissau and covers approximately 102,000 inhabitants. All houses in the study area are visited monthly to register new pregnancies and births. All children below 3 years of age are followed through home visits every third month. The National Hospital where many women from the study area give birth is a few kilometers away. The BHP has worked at the hospital's maternity ward for many years, conducting trials of different health interventions and other observational studies to reduce infant mortality.

Study design

Children will be randomly allocated (1:1) to BCG at admission or at discharge (as per current standard of care).

Enrolment procedures

The BHP team is present at the maternity ward every day. The team will visit the NICU and identify children admitted since their last visit. The mother will have the study explained in the local language (Portuguese Creole), and also receive a written explanation in Portuguese. She will be briefly interviewed about background factors (education, previous deliveries, etc.). Anthropometric measures will be obtained and a Ballard score assessed by a trained nurse.

Randomization

Provided consent, the mother will draw a lot which indicates if the child will receive BCG immediately or BCG at discharge as usual. To ensure equal distribution of different categories of children and calendar time, randomization will be done in blocks within two groups of children; a) children at the NICU who are in an incubator (most severely ill); b) children at the NICU who are not in incubator (less ill, including also children born by caesarean section whose mother is indisposed). Randomization will further be stratified by sex and weight group (1250-1499g, 1500-1999g, 2000-2499g, 2500+ g) in blocks of 16.

Intervention

Immediately after randomization, children who were allocated to BCG vaccine will be vaccinated intradermally with 0.05 ml BCG vaccine (Statens Serum Institute, Denmark). The control group will receive BCG at discharge. All children will receive oral polio vaccine (OPV) at the time of BCG vaccination.

The study is therefore testing BCG+OPV at admission to the NICU (intervention arm) vs. BCG+OPV at discharge from NICU (control arm). Same-sex twin pairs will receive the same treatment.

Blinding

The BHP team will be in charge of randomization and vaccination. These procedures will take place without the participation of the NICU staff. BCG vaccination leaves a small mark on the skin of the child, which vanishes quickly, leaving no sign until a papule is formed several weeks later. To further ensure blinding of hospital staff, a small gauze patch will be placed over the injection site in both groups, for the entire hospital stay.

Follow-up

-At the hospital

The BHP team will visit the NICU every day and ascertain the vital status of all enrolled children and that the plasters have not been removed. All clinical and biological parameters registered by the NICU staff, including weight, temperature and respiratory frequency will also be recorded. The main outcome is survival during the stay at the NICU.

-After discharge

At the day of discharge, the BHP team will assure that all children randomised to the control group receive BCG and OPV before leaving the hospital. The children being discharged will be offered transport to their home. All included children included from Bissau will be followed for adverse events, survival, growth and morbidity with home visits after 3 days, 2 months, 6 months and 12 months of age.

-Statistical analysis

The mortality data will be analyzed in Cox proportional hazards models. Due to the expected high number of twins, all analyses will be adjusted for interdependency between twins. We will analyze the overall effect of BCG on survival, and by cause of death. Length of stay and weight gain at discharge among surviving children will be analyzed using linear regression. To prevent bias which could arise if BCG was particularly beneficial for survival among the smallest children, control and intervention children will be matched based on birth weight, and pairs of children in which one child dies will be excluded from the analysis. There will be a natural variation in the time from birth to enrolment into the trial. All analyses will be adjusted for time from birth to enrolment. All analyses will be conducted stratified by weight group, sex and season (dry (Dec-May) vs rainy (June-Nov)), since we have previously observed that both factors may modify the response to BCG and other vaccines. It is a secondary hypothesis that the early effect is strongest for boys and in the dry season.

Public health campaigns which also affect newborn children, for example polio vaccination campaigns, could interfere with the trial. If such campaigns are implemented also for children in the ICU, the children will be censored in the analysis at the time of the intervention. Likewise if there are strikes, shortage of electricity or other events which interfere with the health services, the children will censored in the analysis after discussion with the DSMB.

-Sample size

The NICU admits around 350-400 children per year who are placed in the incubators. Around 94% weigh more than 1250 g. The median length of stay is 6 days. With a mortality of 12%, we would need to enroll 1262 children to detect a 40% difference in mortality between BCG-vaccinated and BCG-unvaccinated children with 80% power and a significance level of 5%. This corresponds to 122 deaths. If mortality declines (as it has done over the last years in the BHP study area), the event number would have to be slightly higher; with a mortality of 11% we would need to continue to 123 deaths to maintain the same power; if it declines to 10% it would be 124 deaths, etc.

Based on this data, the investigators expect to reach the needed number of events during a period of 4 years, taking into account that some children will have died before they can be enrolled and some children will be too ill to be enrolled.

ETHICAL CONSIDERATIONS The study protocol has been approved by The Danish Central Ethical Committee and the National Ethical Committee in Guinea-Bissau. The trial will be registered at clinicaltrials.gov. In previous studies BCG was safe for LBW children and even more beneficial among the smallest children. A local clinical monitor and a Data Safety and Monitoring Board (DSMB) will be appointed.