Neonatal vaccination with Bacille Calmette-Guérin strains: Effects in infancy

This Ph.D. thesis investigated the importance of BCG vaccine strains for overall morbidity and mortality.

Vaccination with Bacille Calmette-Guérin at birth is recommended in countries where TB and/or leprosy are endemic, and >120 million infants are vaccinated per year. While BCG provides good protection against the most severe forms of TB, the protective efficacy against pulmonary TB ranges between 0-80%. There is accumulating evidence from observational studies and RCTs indicating that the Danish strain of BCG has beneficial NSEs, reducing all-cause mortality by 30-50%. The effect on the risk of hospitalizations had not been assessed prior to this thesis. Among BCG-vaccinated infants, those that develop a BCG skin reaction and/or a tuberculin response have a 30-50% lower mortality than nonreactors. The characteristics of the BCG skin reaction response types had not been investigated prior to this thesis. BCG is a live-attenuated vaccine produced at a host of different laboratories around the world. The vaccine is not a standardized and pharmacologically well-defined product and it has been suggested that different BCG strains should be considered as different vaccines. Different BCG formulations are genetically divergent due to the accumulation of mutations and differences in laboratory production techniques affecting the absolute number of bacteria and the ratio of live-to-dead bacteria. It is unknown whether BCG strains are bioequivalent. In 2018, 184.5 million doses of the WHO-prequalified BCG strains were distributed. The global market share in 2018 for the genetically related BCG-Denmark and BCG-Green Signal was 24%, BCG-Bulgaria 20%, BCGJapan 9% and BCG-Russia 41%. None of these major BCG strains have been evaluated against each other in a randomized trial. Given the widespread use of BCG and its substantial NSEs, we investigated the effects on overall mortality and morbidity and on markers of protection against TB and a well-functioning immune system of neonatal vaccination with different BCG strains in Guinea-Bissau.

We had the following hypotheses:

Hypothesis A. BCG vaccines possess beneficial, non-specific effects related to mortality and morbidity. Since BCG strains are phenotypically distinct, the size of these effects might be different. Compared with BCG-Russia, vaccination with BCG-Denmark and BCG-Japan is associated with fewer admissions, lower in-hospital case-fatality and fewer deaths.

Hypothesis B. The proportion of children having a BCG scar and a positive PPD response is 30% lower among children vaccinated with BCG-Russia compared with BCG-Denmark and BCG-Japan.

Hypothesis C. BCG vaccines are renowned for their safety. Since BCG-Denmark and BCGJapan are probably more potent vaccines, they might be associated with a higher rate of adverse events when compared to BCG-Russia.

During the data collection for this Ph.D., some new ideas evolved. From observations and personal experience of diverse BCG reactions, the Ph.D. student decided to study the importance of the BCG reaction type and size of early BCG skin reactions and the effects of Summary VIII these reaction characteristics on subsequent all-cause mortality. And due to input from our vaccinators, we compared two BCG strains with regard to the content of neonatal BCG doses in a vaccine vial.

BCG is shipped as a vial of freeze-dried bacteria to be reconstituted with a vial containing 1 ml diluent before use. According to manufacturers, these vials contain 20 BCG doses of 0.05 ml for infants or ten doses of 0.1 ml each for children > one year of age and adults. It would, however, only be possible to withdraw and administer these doses if there was zero wastage of liquid in the syringes used for reconstitution and vaccination, as well as in the diluent and vaccine bottles.

In Paper I, we tested whether it is possible to withdraw 20 neonatal doses from vials of BCGRussia and BCG-Denmark. Our experienced vaccinators could only withdraw 13 to 15 doses from both types of vials. This has important implications as vaccine wastage is calculated based on the assumption that the content is 20 doses; hence, wastage of BCG is overestimated, and this leads to unnecessary delays in BCG vaccination due to restrictive vial policies imposed to reduce vaccine wastage.

In Paper II, we supplemented the data on the effect of BCG-Denmark on mortality with data on its effect on hospital admission incidence and in-hospital mortality across three RCTs (6,583 infants). BCG-Denmark did not affect the risk of hospitalization but reduced the severity of disease, the neonatal BCG/no BCG CFR RR being 0.58 (0.35-0.94). Early-BCG especially reduced the risk of fatal neonatal sepsis, the CFR RR being 0.46 (0.22-0.98).

In Paper III, we investigated the effects of providing different BCG strains on morbidity, mortality, BCG skin reactions, tuberculin reactions, and adverse events (ipsilateral lymphadenitis). We conducted a large-scale RCT evaluating BCG-Denmark versus BCG-Russia which was scheduled to enroll 12,000 neonates at the HNSM Maternity Ward. Due to a production halt occurring in 2015 at the BCG manufacturing unit of Statens Serum Institut in Copenhagen, our last stock of BCG-Denmark was depleted by July 2016, almost midway in the trial. We had sought the necessary permits to exchange BCG-Denmark with BCG-Japan for the remainder of the trial and reached 12,000 inclusions in October 2017. Our data indicated no significant differences in morbidity nor mortality by six weeks of age associated with different strains of BCG, but the trial had limited power due to the change in intervention, especially for the mortality comparison. There was, however, somewhat fewer deaths associated with BCG-Japan, the Japan/Russia MRR being 0.71 (0.43-1.19). BCGDenmark and BCG-Japan are likely more immunogenic than BCG-Russia since they were associated with more and larger early BCG skin reactions, tuberculin reactions and adverse events. These strain differences might have important implications both for the protection conferred against TB (specific effect) and for overall health (non-specific effects).

The indications of differences are currently being evaluated in a follow-up trial. We found a significant difference in the proportion of infants with a BCG scar, but much less than the hypothesized 30% difference. BCG-Denmark and BCG-Japan tended to be associated with more adverse events compared to BCG-Russia, but we lacked power to provide definitive conclusions.

There was a high BCG scar prevalence for all BCG strains. The proportion with a positive TST response to 2 TU stimulation was 65% lower for BCG-Russia compared to BCG-Denmark and 46% lower for BCG-Russia compared to BCG-Japan. The incidence of adverse events (lymphadenitis) among infants with home follow-up was 1.0% for BCG-Denmark versus 0.2% for BCG-Russia (p=0.12) and 1.2% for BCG-Japan versus 0.0% for BCG-Russia (p=0.03).

In Paper IV, we studied the idea that BCG reaction characteristics affect all-cause mortality and that the BCG strain administered is one of the main determinants of these characteristics. In a large cohort of 6,012 infants that had received BCG within one week of birth, 97% (5,804/6,012) had a BCG skin reaction by two months and the 2-12-month mortality risk among reactors was 1.9% versus 4.8% for non-reactors. Reacting to BCG versus not reacting was associated with an aMRR of 0.50 (0.26-0.96). Interestingly, the mortality risk was inversely correlated with the size of the BCG skin reaction, the risk being 2.9% for infants with small reactions, 1.8% for medium reactions, and 0.8% for large reactions. The corresponding large/small reaction aMRR was 0.35 (0.20-0.63), and there was a linear trend of decreasing mortality with increasing reaction size (p for trend<0.001). We concluded that the main determinant for both developing a BCG skin reaction and the size of the reaction is the strain of BCG utilized; BCG-Denmark and BCG-Japan are associated with more and larger reactions than BCG-Russia.

In Paper V, we evaluated the effect of having a TST reaction (defined as >1mm) at 2- and 6- months of age versus not having a reaction on all-cause mortality up to 12 months of age. The cohort of infants had received BCG during the neonatal period. Having a TST reaction at two months versus no reaction was associated with a 2-12-month aMRR of 0.34 (0.12-0.94), while at six months, the aMRR was 0.66 (0.15-2.92). We conducted a meta-analysis of the available studies and across four cohorts from Guinea-Bissau, having a 2-month tuberculin reaction versus no reaction was associated with an aMRR of 0.56 (0.38-0.83). Similarly, the effect on mortality of having a 6-month tuberculin reaction versus no reaction was an aMRR of 0.63 (0.39-1.00) across four cohorts. The main determinants for having a tuberculin reaction by six months were large post-vaccination wheals (receiving a large dose of BCG) and being vaccinated with BCG-Denmark or BCG-Japan rather than BCG-Russia.

In conclusion, our findings support that BCG has marked beneficial NSEs that enhance survival substantially. These effects appear to be on the severity of disease (risk of death, inhospital case-fatality risk) rather than on the incidence of infection, as we found no effect on hospitalization risk. Developing a BCG reaction, larger reaction sizes, and positive TSTs are all associated with reduced all-cause mortality, and the main determinant of these reactions is the BCG strain administered. Since all BCG strains tested did produce some degree of BCG reactions and TST reactivity, it would be fair to assume that all are associated with some degree of beneficial NSEs.

A large-scale RCT with adequate power to test strain differences for both specific and nonspecific endpoints have yet to be conducted. We have just finished enrolling neonates in a large comparison of BCG-Japan versus BCG-Russia involving 15,600 neonates and now await a further six months of follow-up before the important results can be published. A third RCT of similar size is planned. Triangulation of the currently available data nevertheless suggests that there would likely be substantial public health advantages obtained at low-cost by:

1) Eliminating restrictive vial policies or at last adjusting wastage calculations to reflect the actual number of doses that can be retrieved from a 20-dose vial.

2) Ensuring provision of at-birth BCG vaccination administered by well-trained personnel and with monitoring of post-vaccination wheal sizes.

3) Prioritizing immunogenic strains such as BCG-Japan over BCG-Russia. BCG-Denmark produced by SSI is no longer available, and the genetically identical BCG-Denmark now produced by AJ Vaccines to our knowledge not yet been examined for its ability to induce scars, TST responses or its overall NSEs.

4) Continuously evaluating both early-life BCG vaccination coverage and the prevalence of BCG reactions, BCG reaction sizes, and TST responses to ensure high standards of vaccination programs. Such an approach could detect whether less-efficient BCG strains and/or insufficient vaccination quality are compromising program efficacy.