Commentary: Prevalence and incidence of celiac disease in patients with rheumatoid arthritis: a case-control study based on the RECORD cohort

Garifallia Sakellariou^1,2*ANNALISA SCHIEPATTI^1,3Anna Zanetti⁴Carlomaurizio Montecucco^1,5Federico Biagi^1,3Carlo Alberto Scirè^4,6

• ¹Department of Internal Medicine and Therapeutics, Università di Pavia, Pavia, Italy, Pavia, Italy
• ²General Medicine Unit, Istituti Clinici Scientifici Maugeri SpA IRCCS Pavia, Pavia, Italy
• ³Gastroenterology Unit, Istituti Clinici Scientifici Maugeri SpA IRCCS Pavia, Pavia, Italy
• ⁴SIR Epidemiology, Research Unit, Milano, Italy
• ⁵Division of Rheumatology, Fondazione IRCCS Policlinico San Matteo, Pavia, Italy
• ⁶School of Medicine, University of Milano Bicocca, Milano, Italy

We read with interest the commentary by Bettacchioli and colleagues (1) on our article on the 23 prevalence and incidence of coeliac disease (CD) in patients with rheumatoid arthritis (RA), based on 24 administrative data of the Lombardy region, Italy (2). With our study as a start, the authors raised 25 some interesting points of discussion regarding some open questions about the optimal testing 26 strategy for CD in other autoimmune disorders. In fact, as an increased prevalence of CD has been 27shown in many autoimmune diseases (3), serological testing for CD is currently recommended only 28 in type 1 diabetes and autoimmune thyroid disorders while the optimal strategy in other autoimmune 29 conditions, including RA, has yet to be established (4,5). There is paucity of data about the 30 prevalence of CD in RA (2,3,6), therefore, this still represents a relatively unknown area, in which 31 further evidence is required for a translation in clinical practice. 322Administrative data for research in uncommon diseases 33 While both RA and CD do not fall under the definition of rare diseases, they still have a relatively 34 low prevalence in the general population (7,8), and this limits the assessment of the co-occurrence of 35 these conditions in cohort studies. While we agree that an optimal diagnosis of CD would be biopsy-36 confirmed, and that detailed RA and CD clinical information would be valuable to study the interplay 37 between these conditions, cohort studies could very hardly assess such an interaction. Even 38 multicentric studies would not reach the necessary power, as suggested by the results of the ESPOIR 39 cohort where only one in 700 RA patients had CD while on our administrative database of 70 061 40 RA patients, derived from a population of about 10 million, we found 171 CD. While ESPOIR 41 enrolled a younger sample, with higher likelihood of new CD diagnoses, this might be less 42 representative of a typical RA population, compared to our population-based sample, and the fact that 43ESPOIR is an historical cohort, enrolled between 2003 and 2005, might further limit 44 generalizability (6). Although diagnoses in administrative databases are prone to uncertainty, we can 45 safely exclude to have classified as CD patients with self-made diagnoses because in Italy the 46 exemption code for CD, that we used to define the disease in our database, can be released only in 47 secondary and tertiary gastroenterology centers based on the results of a duodenal biopsy suggestive 48 of CD. The low probability of an over-estimation is also suggested by a lower prevalence of CD than 49 expected in our population. 50 The impact of screening of CD in autoimmune diseases 51 While we found an increased prevalence of CD in female patients with RA, with and Odds Ratio of 52 2.15 (2), it is remarkable that that we were far from suggesting a universal screening of this group or 53 of all patients with RA. To put our findings into context, the incidence rate ratio (IRR) of CD and 54 type 1 diabetes is estimated between 8 and 9.9, but this is lower in autoimmune thyroiditis (2 to 2.9), 55 although in this population universal screening is recommended. In the same study the IRR for CD in 56 RA it was between 1 and 1.9, but there were no subgroup analyses, and we cannot exclude that 57 females with RA would have a prevalence approaching that of autoimmune thyroiditis (3). 58The sustainability of disease testing is a central aspect, and Bettacchioli et al. estimated such costs in 59 our population (1). Although the cost for screening only females was lower, the overall expense was 60 significant, but this should be interpreted on the background of a general population of more than 10 61 million inhabitants. Moreover, a complete economic evaluation should also consider the impact of a 62 diagnosis of CD, in terms of both increased and reduced costs. A recent systematic literature review 63 estimated the burden of CD diagnosis, taking also into consideration the costs of endoscopy, 64 showing a higher cost per diagnosis than that calculated for our cohort (9). It should also be taken 65 into that the scenario of CD diagnosis in adults is changing, and we cannot rule out that 66 in the future some subsets of patients will be diagnosed without the necessity of a biopsy. This will 67 not imply a relevant increase in costs, as the diagnosis will be based on symptoms and a high 68 autoantibody titer (10). 69 The debate on the optimal testing strategy for CD is an open matter of discussion, and the degree of 71 uncertainty is even higher in patients with rheumatic diseases, as older evidence was based on weak 72 study design (11), while only more recently larger and solid studies have been presented (2,3,6). In 73 the specific setting of RA, considering the typical features of these patients, the limited knowledge on 74 CD in older patients constitutes a further difficulty. To our knowledge, so far, the only attempt to 75 characterize RA patients eligible populations for screening was that of our study (2). Although we do 76 not support universal screening in this subgroup, we recommend keeping a higher level of alertness 77 in these patients, looking for symptoms compatible with CD. We strongly believe that further 78 research will be needed in this area, trying to compromise between the detailed clinical information 79 of the clinical databases and the power of administrative data. 80 The authors declare that the research was conducted in the absence of any commercial or financial 82 relationships that could be construed as a potential conflict of interest. 83