Statistics from Altmetric.com
We recently suggested in a study on 909 adolescent and young adults aged 15+ years (mean 16.1) that the prevalence of autoimmune diseases in coeliac adults is related to the length of gluten exposure, independently of the expected age effect.1 Recently, Sategna Guidetti et al (
) presented a paper which, in the title itself, negates this hypothesis.
However, we feel quite happy with the contribution of Sategna Guidetti et al as we found strong confirmation of our findings in their paper. As mentioned by the authors, their paper stimulates some interesting observations.
The population they studied was affected by a very strong “age” selection as the vast majority were aged over 40 years and hence all had maximum exposure to the risk factors (100% had been exposed to gluten for >20 years, including “actual gluten exposure”) and there was no modulation of effect, just the end point, which surprisingly was identical to our own results. We have not studied a paediatric population, but young adults with a mean age of 16.7 years and the risk factor was evaluated over the whole range of ages before the outcome (autoimmune disease) was expected.
“Age at diagnosis” is a robust variable and is unlikely to be biased. Sategna Guidetti et al showed, very consistently, that age at diagnosis was related to outcome. The actual prevalence of autoimmune diseases was even higher than that observed by us (possibly due to age range?).
The variable “actual gluten exposure”, artificially built by the authors, was largely based on age at diagnosis (hard data) together with minor components related to self reported compliance and follow up.
In summary, if they included in a multivariate model the strong variable “age at diagnosis” which explains a significant part of the variance in the outcome variable, it is very unlikely that a second variable (supposed “actual gluten exposure”) containing the first strong variable adds any further contribution to the outcome variable.
One important prerequisite for a multivariate model is to include variables independent of each other, which was definitely not the case here. In the logistic regression model the variables included were the “strong” ones, as expected in this type of analysis. The outcome (prevalence of autoimmune diseases) was significantly related to present age and age at diagnosis of coeliac disease. What else could contribute to the derived variable “actual gluten exposure”?
To add strength to this finding, we have new prospective data from a cohort of 74 coeliac patients (46 females) diagnosed before the age of five years and followed up for an average period of 18.4 years (range 10–30); their actual mean age is now 20.34 years. Of these, 5/74 developed an autoimmune disease during this follow up period (two dermatitis herpetiformis, one thyroiditis, one MMC, one psoriasis): all of these cases had been exposed to a gluten challenge for 11–48 months after a variable length of time on a gluten free diet. These indeed had “gluten exposure”, unfortunately added on a relatively precocious diagnosis. None of the other 69 patients has developed an autoimmune disease to date.
We thank our colleagues for their significant confirmation to our findings and hope that they will share our will to explore the biological reasons which may explain why age at diagnosis is so strongly correlated with the prevalence of autoimmune diseases in adults.
We thank Ventura et al for their comments on our paper, and we are delighted to have made them happy, but we would like to clarify a few points.
As underlined in our article, we believe that the conclusions of the two papers are not to be considered antithetical for the following reasons.
They were conducted in different coeliac disease (CD) age groups and, as underlined by Londei in his commentary on our article,1 “to date there is no consensus that child and adult CD are the same condition, nor that subjects in whom CD has been diagnosed in adulthood have had CD all their life”.
Study designs were different. In Ventura et al’s survey, actual age between 10 and 25 years was the only selection criterion; CD patients, consecutively recruited over a six month period from 10 paediatric gastroenterology centres, were grouped according to age at diagnosis into three sets: 374 diagnosed before two years of age, 276 diagnosed between two and 10 years, and 259 diagnosed after 10 years of age. Patients who underwent a gluten challenge for diagnostic purposes and its duration were recorded, while neither diet compliance (it is well known that many teenagers are non-compliant!2) nor intestinal mucosa outcomes were mentioned. Conversely, in our study, only patients in whom CD had been diagnosed at our centre, at age ≥16 years (range 16–84), who had been in clinical remission for at least one year, and whose compliance with the diet was ascertained not only by direct enquiry but also by histological outcome of intestinal lesions entered the study: only 422/713 met these stringent inclusion criteria.
When considering an adult versus a paediatric population, we should be aware of a possible screening bias due to different clinical suspicions and presentations.
Although “the conclusions at first glance seem to be similar in the two reports” concerning age at diagnosis as a risk factor, this cannot be viewed as a “surprising” confirmation of Ventura et al’s study hypothesis. Rather this should suggest the application of the “actual gluten exposure” concept to their population, as also proposed by Londei.1
The “strong” age at diagnosis variable can be biased by screening procedures and medical awareness directed to both coeliac and autoimmune (AI) diseases when faced with silent or oligosymptomatic patients.
The “artificially built variable” actual gluten exposure may be perceived with difficulty by a cursory and hasty reading. It is not a “self reported compliance and follow up” but provides a better indication of the effective gluten exposure as it takes into account not only age at CD diagnosis but also age at diagnosis of AI disorders. The beginning of a strict gluten free diet in patients in whom recovery of intestinal lesions was ascertained by histological findings (and not only by self reported compliance, as Ventura et al seem to have gathered), was considered as the end time of gluten exposure. In other words, the period of gluten exposure matched the time of AI disease onset in patients in whom AI disease preceded CD diagnosis, and the beginning of gluten withdrawal (with ascertained compliance by means of the above mentioned criteria) in patients in whom a CD diagnosis was made before AI disease onset, respectively.
Thus when adult CD patients with and without AI associated diseases were compared, age at CD diagnosis, considered as an indirect mirror of duration of gluten exposure, was significantly higher in patients with associated AI disorders, while actual gluten exposure was similar in both groups; moreover, in 35% of patients an AI disease appeared after a diagnosis of CD, even in subjects in whom recovery of intestinal mucosa was ascertained.
This fact and the finding of a 30% prevalence of AI disorders in our patients aged 41 (±15) years compared with 23% in Ventura et al’s study in patients aged 16 (±3.8) years (unless there was a recent revision of which we are unaware, paediatric age comprises adolescence and up to 18 years of age), raises critical questions on the relationship between CD and AI diseases.
The variable “actual gluten exposure” not only reflects more accurately the duration of gluten exposure but eliminates weighty confounding factors that contribute strongly to the apparently significant relationship between age at diagnosis and outcome in the logistic regression model.
In CD, there is a generalised increase in permeability to macromolecules, making it likely that Peyer’s patches are not the only site where gliadins are in contact with the immune system. Gluten peptides may encounter the gut immune system in a fashion that bypasses the normal controlled sampling, leading to sensitisation or loss of tolerance to the antigen.
Despite their diverse aetiology, certain pathogenetic mechanisms are common to all AI diseases: as a rule, they require the presence of self reactive CD4 positive T lymphocytes which are believed to be deleted in the thymus and to be present only when they arise following somatic mutation, producing “forbidden clones”.
The question is why only a minority of CD patients manifest an AI disease?
Most AI diseases show a particular bias for certain HLA haplotypes, usually belonging to class II, which encodes important immune response regulating genes: thus some rational connection may exist between the genetic constitution and susceptibility to AI disorders.
CD seems to meet the criteria of a true AI disease triggered by an environmental agent (gluten) in genetically predisposed individuals. It has been estimated that the HLA contribution to the development of CD among siblings is 36%3 and recent data suggest that a gene or genes other than the HLA unlinked locus must also participate and are likely to be stronger determinants of disease susceptibility than the HLA locus.4,5 The non-HLA locus appears to be inherited as an autosomal recessive trait.4
This may suggest that exposure of the immature system to gliadin in susceptible individuals is a prominent cofactor in modifying the immunological response earlier in life and thus predisposing susceptible individuals, not only to overt CD, but also to AI diseases. In other words, “les jeux sont fait” early in life.
Thus the search for genetic characteristics of CD patients with associated AI diseases could be much more stimulating than meaningless controversies.
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.