Do allergies really protect against cancer?

A new report on the protective benefits of allergy against brain tumors has just appeared in JNCI, the Journal of the National Cancer Institute.  This is the latest of a number of studies that show an inverse association between allergies, as measured by IgE levels, and cancer, including the most common form of brain tumor, glioma.  The idea is that immune surveillance is increased in people with allergies; their immune system destroys incipient tumors as soon as they are detected.

The authors, Schwartzbaum et al., carried out a case-control study using samples from a serum bank in Norway, where sera and medical information have been archived for decades.  Cases were matched with controls on date of blood collected, 2-year age interval at blood collection, and sex.  Cases were donors diagnosed with glioma some time after they gave blood and controls were donors who were never diagnosed with glioma.

Investigators measured IgE levels in the serum samples, stratified their sample on sex and glioma subtype, and time from blood collection to tumor diagnosis.  They found that "among women, testing positive for allergen-specific IgE...was associated with decreased risk of glioblastoma compared with testing negative...  Among both sexes combined, testing positive for total IgE...was associated with decreased risk of glioma... and simultaneously testing positive for allergen-specific IgE and total IgE was associated with a borderline statistically significantly decreased risk of" of brain cancer.

That is, they found an inverse relationship between evidence of allergies and risk of glioma.  But, not in men. And it turns out that this kind of equivocal finding of an association between allergy and cancer is not unusual.  But, if the association is real, why is it so difficult to demonstrate conclusively?

From a paper published last year (Turner, Cancer Immunol Immunother: "Epidemiology: allergy history, IgE, and cancer"):

Numerous epidemiological studies have investigated potential associations between allergy history and cancer risk with strong inverse associations reported in studies of pancreatic cancer, glioma, and childhood leukemia. Although findings may reflect a state of enhanced immune surveillance and anti-tumor defense among those with atopic allergic disorders (allergic asthma, allergic rhinitis, atopic dermatitis), they may also be due to methodological sources of bias in previous work. Conversely, chronic stimulation of the immune system may be associated with increased cancer risk at selected sites.

So, the results may be real or they may be due to "methodological sources of bias in previous work."  And this is a review of 55 different studies.  One, a 2009 study of glioma in the San Francisco Bay Area, found the by now expected inverse association between allergy and glioma but only in patients taking a specific and new medication.  Which means what, exactly, about the association?  Other studies are based on self-reported allergies, or recall, and neither source is terribly reliable.  Prospective studies have yielded weaker results than retrospective studies based on recall.  Some studies have found opposite associations in men and women, and others no association at all.  A 1987 Swedish study of 54,346 hospitalized asthma patients found a significant deficit in cancers, while a 2009 study of 140,425 hospitalized asthma patients in Sweden found a significant excess of cancer. What's going on? 

The first caveat, as always, is correlation doesn't necessarily equal causation. The second caveat, applicable to all epidemiological studies, is that there are sure to be unmeasured confounding variables that affect the results.  E.g., were smoking or family history of cancer factored in?  If not, should they have been?  Third, were the data collection and statistical methods appropriate? These are common issues that submarine studies of disease.  And so on. 

This kind of rollercoastery history of results is not uncommon when an effect is small and there are other contributing factors that differ throughout a study population or between populations.  It's true for many epidemiological studies, as well as genetic studies.  It's a methodological shortcoming, but also an epistemological one -- when causation is multifactorial and heterogeneous and a risk or protective effect relatively small, we just aren't very good at sorting it all out.

There's more. It's not just how easy it is to show the often paper-thin nature of these kinds of claims and findings.  But increasing numbers of chronic non-infectious disease--as they have been categorized, such as vision or intestinal or cancer problems, or diabetes--turn out to involve inflammation if not immunity.  Things are not unicausal, and perhaps 'immunity' is so complex and comprehensive a system that it is science and its history of discovery that credited it only with resistance to microbial infection.  We still have a lot to learn, and not only should we not believe any given study without being circumspect, but we should also recognize that our categorizations of traits and of genetic effects has been misleading. We need less classification, and more nuanced thinking.