To the Editor
We read with great interest the systematic review by Navaratna et al, in particular the section on BCG vaccination to prevent childhood asthma. The authors conclude there is an absence of effect based on two meta-analyses, the first of four cohort studies and the second of two randomised control trials (RCTs).1 For the cohort studies, the combined risk ratio is reported as 0.92 (95% confidence interval (CI) 0.59 to 1.46), but the results of two of these studies appear to have been misinterpreted by the authors, leading to an erroneous conclusion to the meta-analysis.2,3
In the first cohort study, Bager et al report a beneficial effect of BCG vaccination with an adjusted odds ratio for asthma of 0.53 (95% CI 0.03 to 0.95) compared with unvaccinated subjects.2 The 1.63 risk ratio reported in Navaratna et al's figure 3 appears to be the impact of age at vaccination (0 vs 7 years) from Bager et al’s table 2.
In the second cohort study, Marks et al report in their table 2 the effect of BCG on childhood asthma in two separate lines, according to parental history of asthma. In children born to non-asthmatic parents, the relative risk is 1.26 (95% CI 0.92 to 1.73), whereas in children who have one or two asthmatic parents the relative risk is 0.61 (95% CI 0.32 to 1.13).3 Navaratna et al’s figure 3 includes only the former group with a 1.26 risk ratio, with no justification for excluding the latter group in the meta-analysis.
When combining the correct results from the same four cohort studies, we find a strong beneficial effect of BCG on childhood asthma with a risk ratio of 0.76 (95% CI 0.68 to 0.85, P<0.0001) (Figure). Importantly, this means the conclusion by Navaratna et al is incorrect and should state that studies in fact show BCG vaccination prevents childhood asthma.
Navaratna et al’s systematic review did not identify any RCT reporting BCG’s effect on childhood asthma. Instead, the authors included two RCTs in their meta-analysis reporting on infant wheeze, assessed between 13 and 18 months of age, and concluded an absence of effect with a combined risk ratio of 0.93 (95% CI 0.65 to 1.34).1 As asthma cannot be reliably diagnosed before 5 years of age (as the majority of wheezing pre-schoolers do not develop asthma),4 we question the relevance of including these two RCTs in the meta-analysis.
In conclusion, we respectfully disagree with Navaratna et al’s conclusion and, based on our updated meta-analysis of the same four cohort studies, believe that observational studies to date suggest BCG has a beneficial effect on asthma prevention, in line with previous meta-analyses.5,6
Given the limitations and bias inherent to observational studies, RCTs assessing the impact of BCG on childhood asthma are needed. The results of the Melbourne Infant Study: BCG for allergy and infection reduction (MIS BAIR) RCT, in which participants are followed up for at least 5 years, should be available soon, and will more reliably determine the impact of neonatal BCG vaccination on the incidence of asthma.7