Risk of Small Bowel Adenocarcinoma, Adenomas, and Carcinoids in a Nationwide Cohort of Individuals With Celiac Disease

Background & Aims: The incidence of small bowel cancers is increasing. Associations have been made between celiac disease and small bowel ca ncers, but there have been no detailed studies of large cohorts. Methods: Through the nationwide ESPRESSO cohort study, we r etrieved data from Sweden’s 28 pathology departments on all individual s who received a diagnosis of celiac disease diagnosed from 1965 through 2017. Individua ls with celiac disease, defined as duodenal or jejunal villous atrophy (stage 3 Marsh score) were matched with as many as 5 randomly selected reference individuals from the ge neral population. We used stratified Cox regression to calculate hazard ratios (HRs) for sma ll bowel adenocarcinoma, adenomas and carcinoids. Results: During a median follow up of 11 years, we identifi ed 48,119 individuals with celiac disease (patients) and 239,249 reference individual s. Beginning at 1 year after a diagnosis of celiac disease, 29 patients (0.06%) received a diag nosis of small bowel adenocarcinoma vs 45 reference individuals (0.02%), 7 patients received a diagnosis of carcinoids vs 31 reference individuals, and 48 patients received a diagnosis o f adenomas vs 50 reference individuals. Corresponding HRs were small bowel adenocarcinoma 3 .05 (95% CI, 1.86–4.99), carcinoids 0.59 (95% CI, 0.16–2.10), and adenomas 5.73 (95% CI , 3.70–8.88). HRs were independent of sex and age. Overall, there was 1 extra case of sma ll bowel adenocarcinoma in every 2944 patients with celiac disease followed for 10 years. There was an inverse association between mucosal healing risk of future small bowel adenocar cinoma (HR, 0.18; 95% CI, 0.02–1.61), although the HR failed to attain statistical signif icance. Conclusions: In an analysis of a nationwide pathology database in Sweden, we found the absolute risk of small bowel adenocarcinoma is low in individuals with celiac disease. However, risks of small bowel adenocarcinoma and ad enomas (but not carcinoids) are significantly increased in people with celiac disea s compared to people without this disease.


Small bowel adenocarcinoma
Time on gluten-free diet trends toward protection Screening is not indicated, however physicians should have a high index of suspicion if symptoms or signs develop.

Introduction
Primary small bowel cancer is a heterogeneous group of cancers including adenocarcinomas, carcinoids, lymphomas, sarcomas and other cancers. Adenocarcinomas and carcinoids account for the majority of small bowel cancers and about 2-3% 1, 2 of all gastrointestinal (GI) cancers. They have been named "orphan" neoplasias 2 as they have rarely been studied and there is a lack of evidence-based knowledge of risk factors and associated conditions.
Celiac disease is an immune-mediated disease induced by gluten ingestion and has been suggested as one of few predisposing factors for small bowel cancers. This has triggered a number of transcriptomic studies in individuals with both conditions 3 . A meta-analysis of gastrointestinal malignancy in individuals with celiac disease reported 75 cases of small bowel adenocarcinoma in 79,991 individuals with celiac disease from 8 previous studies (Table 1), corresponding to a pooled odds ratio (OR) of 14.4 with heterogeneity >90% for the included studies 4 . The heterogeneity is partly due to mixing of incidence and morbidity ratios, the latter also including prevalent cancers in the risk estimates. The reported relative risks were also much higher in the peridiagnostic period 5 , where there is an imminent risk of detection bias, and symptoms from the small intestinal cancer may trigger diagnostic work for celiac disease (a form of reverse causation). Earlier studies also suffer from other limitations.
First, they have rarely distinguished between different types of small bowel cancers. Second, most of the studies were performed before the introduction of modern diagnostic techniques such as video capsules and double-balloon enteroscopy, which may have limited small bowel cancer detection in the general population. Third, outcome data have been based on ICD-codes rather than histopathological examination. Fourth, population-based studies in celiac disease, other than a small study of 381 individuals 6 have not reported the risk of small bowel adenomas. Small bowel adenocarcinomas are thought to develop through the adenoma carcinoma sequence 7 .
In this study, we examined the risk of future small bowel adenocarcinomas, adenomas, and carcinoids in a contemporary nationwide cohort of more than 48,000 individuals with celiac disease.

Study population
Individual-level data from Swedish national registries were linked through the unique personal identity number assigned to all Swedish residents 8 10 . In individuals where data on celiac disease serology could be accessed, 88% had an elevated value in close temporal proximity to the celiac diagnosis (85% had positive anti-transglutaminase IgA), consistent with data from other clinical cohorts 11 .

Outcome measures: Small bowel adenocarcinomas, adenomas and carcinoids
Our outcome measure was defined as small bowel adenocarcinomas, adenomas and carcinoids registered with relevant SnoMed codes (see Appendix) in local pathology departments extracted for the ESPRESSO study 9 . Specifically, for adenocarcinoma we also defined an ICD-10 entry of "C17 -malignant neoplasm of small intestine" in the national patient registry as individuals having the outcome adenocarcinoma (in total 6 patients added to the 72 identified from the pathology reports). Compared to diagnoses of small bowel adenocarcinoma registered in the Swedish Cancer registry (defined by International Classification of Diseases (ICD), ICD-7 152.X combined with morphology codes to identify adenocarcinomas (SnoMed "81403" or PAD= "096")), only 39 out of our 78 cases were also identified in the cancer registry. On the other hand, the cancer registry contained 8 entries of small bowel cancer that were neither found in pathology reports or the Patient registry, these were not included as cases as they were judged likely to have been erroneous entries as they were never communicated through clinicians or pathologists. Lymphomas were not included in this study.

Reference individuals
For each individual with celiac disease, the government agency Statistics Sweden randomly identified up to 5 reference individuals from the Swedish Total Population Register 12 matched for age, sex, county and calendar year of the date of celiac diagnosis. Reference individuals were free of celiac disease at matching date, and if diagnosed with celiac disease their followup was censored at the date of celiac diagnosis.

Follow-up
Follow-up started 1 year (365 days) after celiac diagnosis and on the corresponding date in matched reference individuals to avoid including patients with celiac disease who were detected in the process of a clinical workup due to symptoms from the small bowel cancer.
The inclusion of such cases will bias the risk estimates (in a sensitivity analysis, these cases were however included to give a full picture of the association between celiac disease and small bowel cancer).
Follow-up ended at date of death, emigration, outcome (small bowel adenocarcinoma, adenoma or carcinoid in separate analyses) or on administrative end of follow-up (Dec 31, 2017), whichever occurred first. For analyses comparing individuals with celiac disease with mucosal healing to those with persistent villous atrophy, date of follow-up began at date of follow-up biopsy (6-60 months after first diagnosis) as reported in our previous publication 13 .

Statistics
We calculated hazard ratios (HRs) using stratified Cox regression. In the stratified regression, each case is only compared to his/her matched reference individuals and a pooled summary HR is calculated from all strata. We further adjusted for categorical educational attainment and 60 months after the initial diagnosis date 15 . In this analysis we calculated the risk of small bowel cancer in individuals with celiac disease with persistent villous atrophy (Marsh III remained at follow-up biopsy) vs mucosal healing (Marsh 0-II at follow-up biopsy). The definition of mucosal healing was not validated but has been used in several previous publications [15][16][17] . Follow-up biopsy analyses were not performed with internal stratification but were instead adjusted for age, sex, time between biopsies and educational attainment.
Incidences of small bowel adenocarcinoma were calculated as the number of events per 1,000 person-years of follow-up. For small bowel adenocarcinoma we also calculated a conditional logistics regression for the risk of future celiac disease diagnosis in individuals diagnosed with previous small bowel adenocarcinoma. The proportional hazards assumption was verified to hold by creating interaction terms with log(time).
All analyses were performed using SAS 9.4.

Ethics
The current study was approved by the Stockholm Ethics Review Board 2014/1287-31/4) on August 27, 2014. The ethics review board did not require informed consent as it is a strictly register-based study 18 .

Results
We identified 48,119 individuals with celiac disease and 239,249 reference individuals still at risk one year after the celiac disease diagnosis date and corresponding date in reference individuals ( Table 2). Individuals were followed for a median of 11 years. The majority of patients were women and more than 40% were children ( Table 2). We performed similar but separate analyses for outcomes of adenomas and carcinoids of the small bowel (the number and characteristics of study participants in these analyses were very similar to those of the adenocarcinoma cohort, exact numbers can be found in the Appendix).

Small bowel adenocarcinomas
In total, 29  In a conditional logistics regression model, the odds ratio (OR) for future or same date celiac diagnosis given a previous/simultaneously diagnosed small bowel adenocarcinoma was 4.14 95%CI=(2.10-6.17). This analysis was based on 17 individuals with celiac disease and 1 reference individual with earlier small bowel adenocarcinoma (Figure 1 depicts a histogram with time between diagnoses also including those diagnosed within first year after celiac diagnosis).

Small bowel adenomas and carcinoids
Individuals with celiac disease were at a 5.73-fold increased risk of small bowel adenoma (95%CI=3.70-8.88). Interestingly the HRs differed to a large extent between different subgroups (even though none of the interaction terms were statically significant) and the risk was highest during the first year of follow-up (actually year 1-2 after celiac diagnosis) and after more than 15 years of follow-up ( Table 4). The risk of carcinoids was not increased as it was only observed in 3 individuals with celiac disease vs. 28 reference individuals (HR=0.59; 95%CI=0.16-2.10; subgroup analyses were underpowered and therefore not performed: Table   4).

Comparison to previous literature
In this study we found an increased risk of adenocarcinomas in individuals with celiac disease compared to age and sex-matched reference individuals. The HRs were lower than reported by most previous studies [19][20][21][22][23][24] and similarly the absolute risk (0.06%) was 10 times lower than in another recent publication 25 26 . No prior study has examined whether mucosal healing alters the risk of small bowel adenocarcinoma. Our study showed a strong but non-significant protective effect indicating that even our study including more than 8000 patients with celiac disease and a second biopsy lacked statistical power to detect a difference.
Nevertheless, we conclude that mucosal healing is probably associated with lower risk of future small bowel adenocarcinoma. In total 52 patients were identified with both celiac disease and small bowel adenocarcinoma. During the entire follow-up of the ESPRESSO study 3,885 individuals were ever diagnosed with small bowel adenocarcinoma, hence celiac disease was diagnosed in 1.3% of Swedish patients with small bowel adenocarcinoma. This study further showed that celiac disease status was associated with better small bowel adenocarcinoma survival, which is in line with previous publications 27 .

Strengths and limitations
The nationwide cohort design has several strengths. First it distinguishes between different types of small bowel cancers and different locations within the small bowel. Second a large proportion of the cohort was diagnosed in the most recent era after introduction of modern techniques such as video capsule and double-balloon enteroscopy. Third, outcome data were based on histopathological examination and fourth we also report estimates for small bowel adenomas. Further the diagnosis of celiac disease has been validated and suggested to be correct in 95% of the cases 10 . Despite the high validity, it is still possible that some of the individuals diagnosed with celiac disease and small bowel adenocarcinomas were indeed misclassified, however we believe this had very limited impact on our results. Another strength of our study is that we also investigated neuroendocrine tumors, carcinoids, and found no association. This suggests lead time and detection bias to be limited also for adenocarcinoma, as these biases would affect both adenocarcinoma and carcinoid estimates equally. Further the result remained virtually unchanged even after excluding all individuals with comorbidities. A limitation of the study is that the diagnosis of small bowel cancer in Swedish pathology registers has not been validated, and that a pathological review could not be performed.
Indeed, the ESPRESSO histopathology cohort identified more cases than the national cancer registry or the patient registry. On the other hand, only 8 of the 47 cases identified in the cancer registry were not found in either patient registry or in ESPRESSO and may result from unverified data in patients who were unaware of their small bowel cancer. Previous work 28 suggests a 10%-missingness in the cancer registry for digestive tract cancers compared to the patient registry, in our data we have a 50% missing compared to local pathology reports.
Underreporting in the Swedish cancer registry to similar extent (44%) has been suggested for several other cancers such as pancreatic and biliary 29 . We believe that using the cancer registry for small bowel adenocarcinoma may therefore not be as adequate in terms of sensitivity, and therefore believe that our study using histopathologic definitions confers more complete and correct data than in earlier reports 5 .

Conclusion
This study found an increased HR of small bowel adenomas and adenocarcinomas in patients with diagnosed celiac disease, but only a very marginal increase in terms of absolute risk. Our results do not imply a need for surveillance but celiac individuals with signs or symptoms of malignancy should merit further investigation for small bowel adenocarcinoma. Mucosal healing was strongly associated with lower risk of small bowel adenocarcinoma even though the association failed to reach statistical significance. one year after celiac diagnosis) in total 6 were diagnosed before celiac disease, 7 same day and 10 during the first year after diagnosis. NR-Not reported.