Allergy on farms has a long history. “Hayfever,” one of the main allergic outcomes, has an origin in the farming community while also asthma is increased in farmers due to exposure to grain dust, animal dander and various chemicals,^{1, 2} Nevertheless Blackley³ noted already in 1873 that “these statistics of the occupations of hay‐fever patients bring out prominently the very curious circumstance that the persons who are most subjected to the action of pollen belong to a class which furnished the fewest cases of the disorder, namely, the farming class”. This paradox has been rediscovered in 1989 by Gassner in the Canton of St. Gallen in the Swiss Alps,⁴ followed by studies of Braun‐Fahrländer⁵ among others. 30 years of research at Alpine farms within the framework of the “hygiene hypothesis,”^{6, 7} is raising now the question if allergy protection at farms is just a modern myth.

The main objection against the farming hypothesis is the interpretation of a negative statistical association as a “protective” effect. Only after thorough exclusion of alternative explanations, this interpretation may be justified. Unfortunately previous farming studies did not exclude other reasons but always followed the same line of arguments that (i) allergy and asthma prevalence is lower in the farming environment, (ii) that this particular environment has different exposure conditions, and (iii) these different exposure conditions are responsible for the lower allergy prevalence. (i) and (ii) are certainly true, but (iii) may be false if both statements are not correlated at all, for example, if the propositional calculus is being wrong.

A world‐wide study at least showed that farming is a risk but not a protective factor.⁸ In this study of 44 centers from 61 countries with 388,811 6‐ to 7‐year‐old children, an odds ratio of 1.2 was shown for hayfever when exposed to farm animals. Also more detailed studies could not replicate the lower allergy prevalence.⁹ From the list of nearly two dozen published “causal” factors (listeria, ascaris, toxoplasma, rubella, eurotium, penicillium, acinetobacter, corynebacterium, endotoxin, spores, horse, dog, pig, cow, milk, whey, silage, dung hills, and cleanliness), none could be verified in the general population. Even the most recent claim of a protective farm index¹⁰ looks more like “cherry picking” in a null result as no bacterial species could be identified. Farming studies have never been pre‐registered nor are any datasets available for independent review.

Could therefore be a rather trivial reason for the lower allergy prevalence in farms? Most recently not only confounding but also colliding has received an increased interest. While confounding describes the action of a third variable on exposure and outcome, colliding describes the selection bias that stems from conditioning on a variable that is itself influenced by exposure and outcome,^{11, 12} Colliding may be expected in studies at rural areas where selective migration over generations may have introduced a healthy worker bias with less allergy genes in today's farming population.¹³ Already one of the first papers already showed that not only children but also less than half of their parents had a history of allergic rhinitis.⁵ The literature on a healthy worker bias in rural areas is not fully clear while the largest study concluded¹⁴ that “selective migration over generations could therefore have contributed to a healthy worker effect with less atopy and less severity of symptoms in the farming population today.” Comparing farm children with their neighborhood may therefore introduce a spurious association (Figure 1). In addition, also cryptic confounding is likely as it is known that farm children receive less vitamin D avoiding a known risk factor for allergy.¹⁵

Although a true effect is questionable from an epidemiological viewpoint, there are some animal studies that are seemingly supporting the farm effect. It turns out, however, that we are amidst a replication crisis where inbred mouse strains are having their own problems,^{16, 17} in particular when it comes to immunological readouts.¹⁸ This may be particular true if mice are kept indoors under allergen deprivation and are fed on a vitamin D diet. Results in mice therefore may be as questionable¹⁹ as the epidemiological results. Cow stables have a long tradition in generating myths.²⁰ “Allergy protection” on farms in its present form looks like another myth at least until convincing arguments can be provided that this is not just an epidemiological fallacy.

农场过敏的历史悠久。“花粉热”是主要的过敏结果之一，起源于农业社区，同时由于接触谷物粉尘、动物皮屑和各种化学物质，农民的哮喘病增加，^{1, 2}尽管如此，布莱克利³早在 1873 年就指出“这些关于花粉热患者职业的统计数据突出显示了一个非常奇怪的情况，即最容易受到花粉作用的人属于发生这种疾病最少的阶层，即农业阶层”。1989 年，瑞士阿尔卑斯山圣加仑州的加斯纳 (Gassner) 重新发现了这一悖论^4，随后对布劳恩-法兰德 (Braun-Fahrländer) 进行了研究⁵其中。在“卫生假说” ^6、7的框架内对阿尔卑斯农场进行了 30 年的研究，现在提出了一个问题，即农场的过敏保护是否只是一个现代神话。

反对农业假说的主要反对意见是将负面统计关联解释为“保护”效应。只有在彻底排除其他解释之后，这种解释才可能是合理的。不幸的是，以前的农业研究并未排除其他原因，但始终遵循相同的论点，即 (i) 农业环境中过敏和哮喘的患病率较低，(ii) 该特定环境具有不同的暴露条件，以及 (iii) 这些不同的暴露条件是导致过敏流行率较低的原因。(i) 和 (ii) 肯定是正确的，但如果两个陈述根本不相关，则 (iii) 可能是错误的，例如，如果命题演算是错误的。

一项世界范围的研究至少表明，农业是一种风险，但不是一种保护因素。⁸在这项对来自 61 个国家的 44 个中心的研究中，有 388,811 名 6 至 7 岁儿童，当暴露于农场动物时，花粉热的优势比为 1.2。此外，更详细的研究无法复制较低的过敏患病率。⁹从近两打已发表的“因果”因素（李斯特菌、蛔虫、弓形虫、风疹、蛔虫、青霉、不动杆菌、棒状杆菌、内毒素、孢子、马、狗、猪、牛、牛奶、乳清、青贮饲料、粪丘和清洁度），在一般人群中无法验证。即使是最近声称的保护性农场指数¹⁰在无效结果中看起来更像是“采摘樱桃”，因为无法识别出细菌种类。农业研究从未预先注册，也没有任何数据集可供独立审查。

因此，这可能是农场过敏流行率较低的一个相当微不足道的原因吗？最近，不仅混淆而且碰撞也受到了越来越多的关注。混杂描述了第三个变量对暴露和结果的作用，碰撞描述了选择偏差，这种偏倚源于对本身受暴露和结果影响的变量进行条件化，^{11, 12}在选择性迁移的农村地区的研究中可能会出现碰撞几代人可能已经在今天的农业人口中引入了健康的工人偏见，过敏基因较少。¹³最早的一篇论文已经表明，不仅是孩子，还有不到一半的父母有过敏性鼻炎病史。⁵关于农村地区健康工人偏见的文献并不完全清楚，而最大的研究得出的结论是¹⁴ “因此，几代人的选择性迁移可能有助于产生健康的工人效应，在今天的农业人口中，特应性较少，症状较轻。” 因此，将农场儿童与其邻居进行比较可能会引入虚假关联（图 1）。此外，还可能存在隐秘的混杂因素，因为众所周知，农场儿童摄入的维生素 D 较少，从而避免了已知的过敏风险因素。¹⁵

尽管从流行病学的角度来看，真实的影响值得怀疑，但有一些动物研究似乎支持农场效应。然而，事实证明，我们正处于复制危机之中，近交小鼠品系也有自己的问题，^16、17，尤其是在免疫读数方面。¹⁸如果将小鼠关在缺乏过敏原的室内并以维生素 D 饮食喂养，则情况可能尤其如此。因此，小鼠的结果可能与流行病学结果一样值得怀疑¹⁹。牛厩在创造神话方面有着悠久的传统。²⁰ 目前形式的农场“过敏保护”看起来像是另一个神话，至少在提供令人信服的论据之前，这不仅仅是一种流行病学谬误。