Authors' reply to commentary

EDITOR—In his commentary on our paper Fine claims that our observations lack the classic attributes of causality: gradient, strength, and coherence.1

Firstly, the estimates only just reach conventional significance. However, the important finding was not the significance of the estimates for the individual vaccines but the fact that BCG and diphtheria, pertussis, and tetanus vaccines had opposite effects on mortality—the difference being strongly significant (P=0.005).

Secondly, the association of diphtheria, pertussis, and tetanus vaccine with increased mortality could be due to higher vaccination coverage and higher mortality among children of young mothers. However, as we stated in our paper, adjustment for background factors had minimal effect on mortality estimates. Given the low prevalence of young mothers and their slightly higher vaccine coverage (table 2), adjustment for maternal age had little effect. Mortality ratios changed from 1.84 (1.10 to 3.10) to 1.82 (1.08 to 3.07) with one dose of diphtheria, pertussis, and tetanus vaccine and from 1.38 (0.73 to 2.61) to 1.37 (0.72 to 2.59) with 2-3 doses.

Thirdly, our results may be anomalous because mortality declines with age for children who had not been vaccinated with BCG but increases with age among those who were (table 3). These findings are to be expected if diphtheria, pertussis, and tetanus vaccine increases mortality. Exposure to diphtheria, pertussis, and tetanus vaccine increased with age for children vaccinated with BCG (table below), whereas there was little change for those who were not so vaccinated. Presumably children who are not vaccinated with BCG have less contact with immunisation services and are therefore less likely to receive additional doses of diphtheria, pertussis, and tetanus vaccine during follow up.

Fourthly, the mortality ratio of 1.84 (1.10 to 3.10) for one dose of …