Statistics from Altmetric.com
We read the paper by Wijga et al1 published in the July 2003 issue of Thorax with great interest. The authors found that the consumption of specific food items such as full cream milk, butter, and brown bread can contribute to a decrease in the risk of asthma and wheezing in pre-school children. These findings agree with previous studies in adults,2 but there are a few methodological problems in the analyses used in the study which may have influenced the results obtained.
Our first area of concern is that trans-generational traditions of families with atopic diseases are not taken into consideration. For instance, families with a history of atopy tend to smoke less, which is described as a “healthy passive smoker effect”.3,4 Grandparents and parents who have asthma tend not to smoke, but their children are more likely to develop atopic manifestations than children from smoking families without asthma. It is also likely that atopic parents change their exposure to pets which may lead to a similar “healthy pet keeping effect”.5 For related reasons, families may also alter their diets resulting in a “healthy cow’s milk effect”. These potential changes within families are supported by avoidance strategies propagated by various national medical associations. Children of atopic parents therefore tend to experience different exposures. Hence, before using statistical models we need to investigate the extent to which the diet of children differs according to the atopic status of their parents. An additional table is therefore needed, comparable to table 2, with consumption frequencies in columns for allergic and non-allergic parents as well as “allergic mothers” and “allergic fathers”. In this respect, we were surprised to see the different proportion of mothers (14.2%) and fathers (29.1%) with allergy. Previous studies have reported the proportions of atopy in both mothers and fathers to be consistently around 30–35%.6–8
This leads to our second area of concern which deals with the logistic regression model used in table 3. By modelling, the authors attempted to control for the effects of maternal and paternal atopy by considering them as confounding variables. We see the child’s diet acting as an intervening variable in the association between parental atopic history and the risk of the child developing asthma (fig 1). The standard epidemiological rule is that intervening variables should not be considered as confounders.9 By neglecting this rule the authors may have achieved biased odds ratios. In addition, children with a parental history of allergy are likely to have a different genetic make up. It is therefore very likely that they react differently—for instance, to cow’s milk—which requires the investigation of interactive effects of parental atopy and diet on asthma in the offspring. To overcome these two challenges the authors need to present another table stratifying for parental allergy. The child’s risk of wheezing should be estimated for each stratum. This would then adequately control for the intervening effects of parental allergy and provide an indication of any interaction. Additionally, information on the risk of exposure to cow’s milk in children with and without parental atopy would help either to change or—since we expect no protective effect in atopic families—to sustain the recommendation to avoid early exposure to cow’s milk.
To add to this argument, the authors also failed to differentiate between allergic and non-allergic (transient?) wheezing. We look forward to seeing additional informative tables.
The first point made by Karmaus and Fussman is that the association we observed between consumption of products containing milk fat (full cream milk, milk products, butter) and a reduced risk of asthma in pre-school children1 could be the result of a “healthy cow’s milk effect”. We think this is unlikely as avoidance of cow’s milk is uncommon in the Netherlands. Dutch guidelines do not advise avoidance of cow’s milk for children with familial allergy after the age of 12 months unless the child has cow’s milk allergy.2 The popular belief is that young children need milk for healthy growth and few parents see milk as potentially harmful. 99% of Dutch children aged 1–4 years use milk (products).3 In the PIAMA population, too, nearly all children used milk—either full cream or semi-skimmed. Apart from 64 children with cow’s milk allergy who were excluded from the analyses, only 27 children (<1%) had not used milk (products) in the previous month. Of these children, 16 had an allergic parent and 11 had non-allergic parents. The data requested by Karmaus and Fussman do not show an association between parental allergy and the prevalence of daily consumption of the foods that we found to be associated with reduced risk of asthma or wheeze: full cream milk, milk products (mainly flavoured and unflavoured yoghurt, either full cream or low fat), butter and brown bread (table 1, first column).
Karmaus and Fussman correctly point out the imbalance between the percentages of allergic mothers and allergic fathers in the study. This imbalance is due to the study design. Maternal allergy was used as the criterion to allocate participants to subgroups of the PIAMA study and in the natural history part of the study non-allergic mothers were oversampled.
The second point of concern raised by Karmaus and Fussman deals with our logistic regression model. They state that, by treating parental allergy as a confounder, we neglected the epidemiological rule that intervening variables should not be considered as confounders. However, in their fig 1, parental allergy is not an intervening variable between the exposure of interest (diet) and the effect studied (child’s asthma). Parental allergy would only be an intervening variable when the child’s diet at age 2 is a cause of parental allergy. Instead, in their figure parental allergy is a classical confounder—that is, it is a factor that is a cause of both the exposure and the disease of interest4 and was therefore dealt with correctly in our analyses.
We agree with Karmaus and Fusmann, however, that effect modification might be present in that children of allergic parents might react differently to dietary exposures, but there was no evidence of effect modification in our data. The prevalences of asthma and of wheeze were lower in daily users in most parental allergy subgroups although, because of the small numbers, even large differences were not statistically significant in most cases (table 1, data shown for only three subgroups). Adjusted logistic regression analysis was carried out in only two subgroups because numbers were too small in the other groups. The adjusted odds ratios for daily consumption of full cream milk, milk products, and butter were consistently below 1 both in children of allergic and in children of non-allergic parents; however, because of the small numbers, the confidence intervals were wider than in the analysis of the group as a whole (see table 3 in our paper).
The last point made by Karmaus and Fussman is that they would have liked us to differentiate between allergic and non-allergic (transient?) wheezing. IgE measurements were not available for these children and they were too young for us to be able to differentiate between transient and persistent wheezers. We look forward to future analyses when a more reliable asthma diagnosis can be made.
We conclude that there is no evidence in our population for a “healthy full cream cow’s milk effect”, that we adjusted correctly for parental allergy in our analyses, and that our data do not suggest that the reported associations between daily consumption of products containing milk fat and reduced risk of asthma and wheeze are only present in children of non-allergic parents.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.