A recent paper on the bias of farming studies did not discuss a healthy worker although this is being a reasonable explanation.
So let’s have a more detailed look here at farm parents. It is an important question if parents are also “protected” (or if some affected parents just moved away).
AFAIK there are 7 studies dealing with this question.
Unfortunately the definitions of allergy history in these papers are vague – does history mean both parents only? Grandparents and sibs? Verified? Also the largest study so far (GABRIEL) never published data on the history of allergy in farming families.
So we can not run run a formal meta-analysis here, while the message is already clear: All except one minor study show that the farming effect is found in parents as well.
Again – was there already “protection” during the childhood of the parents or is it just a delayed healthy worker effect?
I am going to answer this question by looking into effect propagation in the Ehrenstein paper as this is the largest study so far. It basically works by running a 2×2 table in the parental generation (a bit questionable as we have 2 parents) and another 2×2 table in the child generation. Effect should propagate as indicated by movement of observed versus expected cases between cells.
If there is a true effect there should be a horizontal movement, basically just less hayfever in the faming group. If there is a pseudo effect only there should be a vertical movement, eg less hayfever by leaving the farm.
And here is my back on the envelope estimate – correct me if I am wrong. The odds ratio decreases between generation – so the effect size decreases – from about 0.64 to 0.35. But there is no horizontal shift at all indicating a true farm effect in Bavaria / Germany.
Don’t know of any formal test here – as there is no homogeneity of odds ratios across strata, this is making any C-M-H test to be obsolete.