Statistics from Altmetric.com
We read with interest the study by Nakajima et al1 which concluded that, in Tasmanian children, there are small age-dependent associations between childhood immunisation and asthma, eczema and food allergy, but that these effects should not deter parents from immunising their children. However, it could be that the small (but significant) effects that were found are due to residual confounding since the authors made no adjustment for socioeconomic status, a factor found to be associated with allergy.2–4
On the other hand, the effects may have been underestimated since the authors included diseases preventable by childhood vaccinations (diphtheria, pertussis, measles, mumps and rubella) in the model, but these (what they call) “confounders” are in fact intermediate variables which possibly “take away” the association between vaccinations and allergy.
Bernsen and van der Wouden1 have proposed that some of the associations between childhood immunisation and atopic conditions that we observed2 may have been due to confounding by socioeconomic status (SES), given that allergies are known to be associated with SES. Unfortunately the Tasmanian Asthma Study did not collect complete information on the SES of the parents of this cohort, for which is the correct measure to adjust, given that immunisation is a childhood exposure. On the other hand, SES itself is not a true causal factor for allergy and is only related to health outcomes because of its associations with other predictors of disease such as health attitudes and behaviours, and environment.3 In our analysis we investigated confounding by breast feeding, parental smoking and infections as likely SES associated predictors of allergic disease, and adjusted wherever it appeared important. We have now used the information on father’s occupation available in the school medical records to define childhood SES and repeated the analyses adjusting for this variable. Father’s occupation was coded using the Australian Standard Classification of Occupations and then grouped into five categories based on skill level.4 Associations remained largely unchanged except for those with food allergies by age 7 which were significantly attenuated (table 1).
Previously, a modestly higher reported incidence of food allergies by age 7 years was associated with immunisation against diphtheria, tetanus, pertussis and polio but not with smallpox. After adjustment for father’s occupation, there remained little evidence for an association between any of the immunisations and food allergies by age 7. In our paper2 we discussed how some biases including parents’ health attitudes may have contributed to the apparent association between the risk of childhood atopic conditions and vaccination. This argument is supported by our finding on attenuation of some estimates when adjusted for a proxy measure of SES such as father’s occupation which, in turn, is likely to be related to health attitudes. The additional analyses further reinforce our conclusions that evidence for an association between immunisation and atopic disease is minimal and should not deter parents from immunising their children.
Bernsen and van der Wouden also suggest that our estimates would have been stronger if we did not adjust for infections which could have been prevented by immunisation. We agree that these are not confounders but intermediate variables potentially in the causal pathway. We adjusted for infections because we wanted to investigate any direct effects of immunisation (ie, immune potentiating effects) rather than effects that could be explained by subsequent infections. However, we have now rerun all the analyses without any adjustments for the relevant infections. This sensitivity analysis showed negligible differences in results between adjusted and non-adjusted models. The interpretations and implications of our findings therefore remain the same.