Let me begin with a disclaimer: I am personally affected by hay fever, it is (late) spring, and no wonder I digress on a topic, I know very little about. I do know (even if surprisingly badly), however, my own body, so once I exhausted all tools of modern medicine to combat the mother load of unnecessary illnesses I started experimenting with myself. I read a lot about hay fever from non-scientific sources and tried to put some of the recommendations into practice. In one year, I did a relatively drastic diet doing away with ‘white calories’, refined white sugar, white flower etc. My hay fever did not go away, but turned out to be less of a nuisance than usual.
I told this to my physician who won’t have any of it. He had never heard or read about a connection between sugar consumption and hay fever. As a scientifically trained doctor, he did not know the causal mechanism linking sugar to hay fever and thought of my idea as a fluke. Of course, it is well known that sugar creates a lot of problems for immune system and is the culprit of many civilization diseases like diabetes. But hay fever?
I began screening the scientific literature – some 100 studies and counting. Perhaps screening is still somewhat exaggerated, since I have a very basic understanding of modern medicine, and obviously struggle with technical terms. However, I am semi-literate when it comes to empirical methods and statistics and thought I give it a shot.
A few general observations along the way: Causality issues, the traditional bone of contention in many fields of social sciences, are no strangers in the hard sciences either. Most studies I read are based on some sort of correlational analysis, with observational data. There are a few clinical trials (some on animals), but all in all it is hard to uncover the roots of a phenomenon as complex as hay fever.
What seems to be clear is that hay fever (allergic rhinoconjunctivitis ) is a sort of civilizational disease: specific (Western) civilizations seem to be particularly vulnerable. Hay fever barely exists in some very poor countries, and was already very important in Europe and North America the 19th century, where it has increased tremendously in the last decades. Just check this google ngram using the German word for hay fever. We see that hay fever as a problem mentioned in the literature has increased over time, first until the second world war, and then from the 1970s onwards to an all-time high current times (I have no story for the jumps in the 1940s and 1950s, but the graph would look similar if smoother using the English term). Quite interestingly Eastern Germans only started having hay fever after reunification (Mutius et al. 1998).
Hence, although it is clear that genes do play a role, there has to be a huge environmental or behavioural driving force explaining the trend towards more hay fever. Three dominant hunches are contamination, hygiene, and nutrition. Some scholars argue that air pollution makes the effect of pollen on the immunologic system more aggressive. Others argue that modern hygiene has counter-productive effects, making the human body reacting strongly to hitherto innocuous substances like pollen. There is some evidence from studies comparing people being exposed to livestock (especially on farms, but also pets (e.g. Waser et al. 2004, Smit et al. 2007). There are quite a few studies relating the incidence of hay fever to family size: in larger families small kids in early childhood mutually infect each other on a regular basis, thereby training the immune system against real as opposed to ‘fake’ threats (e.g. Strachan 1989, Ponsonby et al. 1999, Forastiere et al. 1997, Raesaenen et al. 1997). Given that all these observations are based on observational data, and family size might be a proxy for many things, it is still unclear whether these findings are truly valid, and how exactly the causal mechanisms for the hygiene hypothesis works.
The most bewildering literature deals with the effects of nutrition. There are a few (often industry-financed) studies on breast-feeding and different types of formula milk for infants (e.g. Berg et al. 2008). There are several studies relating fatty acids to hay fever (Murray et al. 2005, Laitinen et al. 2005). There are studies arguing that vitamins and a (Mediterranean) diet based on fruits and vegetables reduces hay fever (Ruehl et al. 2007, Zaknun et al. 2012). Yet, there are often complex causal relationships linking nutrition to allergies (e.g. Rosenlund et al. 2012), and quite a few studies are unable to demonstrate said relationship. For instance, Miyake et al. (2006) argue that the reason why Japanese are relatively unbothered by hay fever is less related to vitamins and more to other peculiarities of Japanese diet like the intake of seaweed.
To the best of my knowledge, no study has directly probed the relationship between sugar consumption and hay fever. There may be studies which did not find anything and did abstain reporting a null finding, but all in all, sugar does not figure prominently in the current medical debate. I think that is a pity. My own experience biases me toward a belief that there is some truth to the sugar hypothesis. But sugar is also a well known trouble maker for other civilizational diseases such as diabetes. It is also well known to create problems for the immune system (by blocking vitamins and causing digestive difficulties). It is therefore tempting to test the relationship between sugar consumption and the occurrence of hay fever more systematically.
Since I do not dispose of an adequate micro dataset I return to one of my work-horse tools: country comparisons. A consortium of many research institutes (the ISAAC or International Study of Asthma and Allergies in Childhood link) has collected data for the late 1990s from more than 50 countries around the world (The lancet article 1998). I use this data (average information in cases where several observations are available within a country).
For sugar consumption I use data compiled by a Swedish website, which measures annual intake of per capita. The following graph shows a scatter plot for the two types of information.
Nigeria and Paraguay pop into the eye immediately. These countries have more than 30 percent of the children showing symptoms of hay fever, and relatively low levels of sugar consumption. In contrast, other countries with low sugar levels like Ethiopia or Albania also have very low levels of hay fever. The trend line for all countries shows a moderately increasing relationship between sugar intake and hay fever, but is clearly affected by extreme countries like Nigeria, Paraguay, or Singapore.
Looking at the graph the litmus test has resulted grayish. So let’s add a little more information to the dataset and see whether we can isolate the relationship between sugar consumption and hay fever from confounding factors. To this end, I collected several types of information:
o I added information about the share of young population (up to 15 years old) of the total population. This should proxy household size in these countries and therefore control for the hygiene hypothesis.
o I include World Bank estimates on GDP per capita in purchasing power parities as a proxy for many things: standard of living, life style, access to modern medicine.
o Finally, I include latitude to control for differences in vegetation. One would assume that well known forms of aggressive pollen are more likely in specific climate zones. I use absolute numbers for the latitude to account for the fact that there are two hemispheres, and use a squared term, because it is middle latitudes where (e.g. ragweed) pollen activity is highest (at least within that short period of time that is spring).
o In further tests I use other specifications which also include, continental dummies, dummies for countries producing a lot of ethanol (i.e. sugar extraction for non-food purposes) etc.
I excluded extreme outliers (Nicaragua and Paraguay). I have no story why allergy is so strong in these two countries. The results for the remaining countries are only as good as your willingness to accept this exclusion. Other outliers (such as Malta or Singapore) don’t prove to invalidate the results. Given that the data in the scatter plot is more cone-shaped than equally distributed (pointing to what econometricians call heteroscedasticity), I use regression models with robust standard errors.
The results of these regressions are quite remarkable, considering the paucity of the underlying data. Sugar consumption is the most robust factor, and substantively quite strong. Let me talk you through the other factors first: As expected, incidence of hay fever rises with income levels. The effect is not as strong as I thought. Increasing per capita income by another 1000$ (e.g. from Belgium to Austria) only increases the incidence by a fifth of a percentage point. I.e. you would have to jump 5000$ (e.g. from Belgium to US income levels) to reach another percentage point of affected population. The share of young population does not matter unlike in many micro studies. Maybe because it is a poor proxy, or because other factors of life style matter more than having large families. The model does predict that with latitudes farer away from the equator first the incidence rises, and then (very far away) drops again. Up to latitude 21, the incidence rises, thereafter it falls. At the extremes, e.g. going from the equator to another 1 degree north or south, the (marginal) effect is .3. That is going 3 degrees north or south the incidence of hay fever increases by a percentage point.
Now to my main concern: sugar consumption. To repeat, the metric is kilos per person per year. The mean of the sample is 33.4 kilos, but with tremendous diversity (as seen in the plot above). Adding another kilo of sugar to the average diet in a country increases incidence of hay fever by 0.15 percentage points. I.e. increasing sugar consumptions of Japan (a relatively modest 22 kilo) to Australia (with a whopping 50 kilo) would add another 5 percent(age points) of people suffering from hay fever to the population. That is no trivial quantity. The effect is statistically significant, and as robust as the limited data allow. The following graph gives a visualization of this (partial) effect.
I imagine that medical scientists would ridicule this kind of ‘empirical evidence’. Clearly, my analysis is limited in many ways: it uses macro data, which is only cross sectional (both sugar and allergy data hard to get and compare across time), observational, and I had to omit many other confounding factors. Nonetheless, I would contend, this is a promising start. With good micro data, and perhaps some experimental trials, it would be interesting to see whether changes in diet really alleviate the symptoms of hay fever, or even reduce its incidence. I very much hope that a group of medical scientists one day tries this out. I am willing to help out, as much as I can.