Abstract

Aims

Inflammatory markers are established risk factors for atrial fibrillation (AF), but the role of autoimmune diseases is unknown. The aim of the study was to examine the association between coeliac disease (CD) and AF in a large cohort of patients with biopsy-verified CD.

Methods and results

We identified 28 637 patients with CD through biopsy reports (defined as Marsh 3: villous atrophy) from all pathology departments (n= 28) in Sweden. Biopsies had been performed between 1969 and 2008. Age- and sex-matched reference individuals (n= 141 731) were identified from the Swedish Total Population Register. Data on AF were obtained from the Swedish Hospital Discharge Register, the Hospital Outpatient Register, and the Cause of Death Register. Hazard ratios (HRs) for AF were estimated using Cox regression. In the CD cohort, 941 individuals developed AF (vs. 2918 reference individuals) during a median follow-up of 9 years. The corresponding adjusted HR for AF was 1.34 (95% CI = 1.24–1.44). The absolute risk of AF in CD was 321 of 100 000 person-years, with an excess risk of 81 of 100 000. A prior AF diagnosis was also associated with an increased risk of subsequent CD (odds ratio = 1.45, 95% CI = 1.31–1.62).

Conclusions

Atrial fibrillation is more common both before and after CD diagnosis in patients with CD though the excess risk is small. Potential explanations for the increased risk of AF in CD include chronic inflammation and shared risk factors, but ascertainment bias may also have contributed.

Clinical implications

Coeliac disease affects 1–2% of the Western population. Our results indicate that patients with coeliac disease, verified by intestinal biopsy, are at increased risk of atrial fibrillation. This observation is consistent with previous findings that elevation of inflammatory markers predicts atrial fibrillation. Additional studies are needed to clarify the mechanistic link between atrial fibrillation and autoimmune diseases such as coeliac disease.

Introduction

Coeliac disease (CD), commonly known as gluten intolerance, is an immune-mediated disorder that occurs in 1% of the Western world. In patients with CD ingestion of gluten causes small intestinal inflammation and villous atrophy (VA).

Atrial fibrillation (AF) is a growing public health problem associated with substantial morbidity and mortality through an increased risk of ischaemic stroke and heart failure.1,2 Elevation of inflammatory markers has consistently been shown to precede onset of AF.3–6,7 Hypothetically, inflammatory processes could result in atrial fibrosis, which is considered a substrate for AF.8 However, additional studies are desirable to identify the underlying causes of such inflammation.

There are limited data on the association of autoimmune diseases with AF.9 Although several studies suggest an increased risk of cardiovascular morbidity and mortality in CD,10,11 we are only aware of one study directly addressing the association between CD and AF.9 However, that study was underpowered regarding AF but indicated a positive relationship between CD and prior AF [odds ratio, OR = 1.26 (95% CI = 0.97–1.64)].9 Thus, it remains unknown whether CD is associated with risk of AF. We therefore examined the risk of AF in a Swedish nationwide cohort study of more than 28 000 patients with biopsy-verified CD.

Methods

We collected nationwide data on CD through biopsy reports from all (n= 28) Swedish pathology departments. Using the unique personal identity number (PIN),12 these data were matched to both inpatient and outpatient data on AF in the Swedish National Patient Register and in the Swedish Cause of Death Register.

Collection of biopsy data

We defined our study exposure, CD, as having a small intestinal biopsy with VA.

Villous atrophy

Between October 2006 and February 2008, we identified 29 148 patients with CD through computerized biopsy reports from all of the 28 Swedish pathology departments.13 The biopsies had been performed from 1969 to 2008 (Table 1). We defined CD as having a SnoMed pathology code equal to VA and did not require a positive CD serology for the diagnosis (see Appendix for a list of Swedish SnoMed codes translated into the international histopathology grading system by Marsh). However, a study from our group has shown that 88% of individuals with VA and available serology data also had positive CD serology at the time of biopsy.13

Table 1

Characteristics of study participants

 Matched controls (n= 141 731) Coeliac disease (n= 28 637) 
Female 88 411 (62.4) 17 821 (62.2) 
Male 53 320 (37.6) 10 816 (37.8) 
Age at study entry, years (median, range) 29; 0–95 30; 0–95 
Attained age, years (median, range) 40; 1–105 40; 1–100 
 0–19 (%) 58 847 (41.5) 11 801 (41.2) 
 20–39 (%) 26 354 (18.6) 5300 (18.5) 
 40–59 (%) 31.996 (22.6) 6435 (22.5) 
 ≥60 (%) 24 534 (17.3) 5101 (17.8) 
Entry year (median, range) 1997; 1969–2008 1997; 1969–2008 
Follow-upa, years (median, range) 9; 0–40 9; 0–40 
Follow-upa, years (mean ± SD) 10.4 ± 6.4 10.2 ± 6.4 
Calendar year 
 –1989 20 212 (14.3) 4085 (14.3) 
 1990–99 58 722 (41.4) 11 863 (41.4) 
 2000– 62 797 (44.3) 12 689 (44.3) 
Type 1 diabetes 534 (0.4) 922 (3.2) 
Rheumatoid arthritis 799 (0.6) 279 (1.0) 
Autoimmune Thyroid disease 293 (0.2) 193 (0.7) 
 Matched controls (n= 141 731) Coeliac disease (n= 28 637) 
Female 88 411 (62.4) 17 821 (62.2) 
Male 53 320 (37.6) 10 816 (37.8) 
Age at study entry, years (median, range) 29; 0–95 30; 0–95 
Attained age, years (median, range) 40; 1–105 40; 1–100 
 0–19 (%) 58 847 (41.5) 11 801 (41.2) 
 20–39 (%) 26 354 (18.6) 5300 (18.5) 
 40–59 (%) 31.996 (22.6) 6435 (22.5) 
 ≥60 (%) 24 534 (17.3) 5101 (17.8) 
Entry year (median, range) 1997; 1969–2008 1997; 1969–2008 
Follow-upa, years (median, range) 9; 0–40 9; 0–40 
Follow-upa, years (mean ± SD) 10.4 ± 6.4 10.2 ± 6.4 
Calendar year 
 –1989 20 212 (14.3) 4085 (14.3) 
 1990–99 58 722 (41.4) 11 863 (41.4) 
 2000– 62 797 (44.3) 12 689 (44.3) 
Type 1 diabetes 534 (0.4) 922 (3.2) 
Rheumatoid arthritis 799 (0.6) 279 (1.0) 
Autoimmune Thyroid disease 293 (0.2) 193 (0.7) 

aFollow-up time until diagnosis of atrial fibrillation, death from other cause, emigration or 31 December 2008. In reference, individuals follow-up can also end through small-intestinal biopsy.

Inflammation and normal mucosa

We also collected data from biopsy reports in patients with small intestinal inflammation but no VA (Marsh 1–2; n= 13 446) and those with normal mucosa (Marsh 0; n= 244 992).

A regional subset of unique individuals with normal mucosa (121 952 of 244 992 patients) was matched against data on positive CD serology data (gliadin antibodies, endomysial antibodies, and tissue transglutaminase antibodies). Serology data originated from eight university hospitals where patients had undergone biopsy. These university hospitals are responsible for both rural and urban areas, representing 49% of the Swedish population. Linkage with CD serology data was performed to create a cohort of individuals with potential early CD; through this linkage, we found 3736 individuals with a normal mucosa but positive CD serology up to 180 days before biopsy and until 30 days after biopsy. Individuals with inflammation or normal mucosa but positive CD serology were then used as secondary reference groups. The collection of biopsy reports with inflammation and normal mucosa has been described in detail elsewhere.13,14

Matched reference individuals

Records of all patients with CD (n= 29 148), inflammation (n= 13 446) and normal mucosa but positive CD serology (n= 3736) were sent to Statistics Sweden. Each individual with a small intestinal biopsy was then matched with up to five reference individuals (controls) on age, sex, county, and calendar period from the Total Population Register.15 To protect the integrity of individuals, the PINs of the index individuals and their reference individuals were replaced by serial numbers.

Exclusions

We excluded individuals undergoing biopsy where Statistics Sweden had assigned no serial number or controls and where the biopsy may have originated from the ileum (Figure 1). We also excluded controls that could not be matched to an index individual or controls with other data irregularity (Figure 1). These exclusion criteria left us with 46 121 biopsied individuals and 228 632 controls (identical to that of our previous study on mortality10). For the purpose of this study, we subsequently excluded individuals with AF before biopsy or study entry (corresponding date in controls), individuals with additional data irregularities and controls where the matched index individual had been excluded for other reasons (since data were analysed by stratum).

Figure 1

Flow chart of study participants. CD, coeliac disease; AF, Atrial fibrillation. *1223 controls were excluded because their matched individual undergoing biopsy had been excluded (and all analyses were matched on strata, see Methods). #Marsh pathology grade 1–2. §Marsh pathology grade 0 in individuals with positive CD serology (antigliadin, antiendomysial, or anti-transglutaminase antibodies).

Figure 1

Flow chart of study participants. CD, coeliac disease; AF, Atrial fibrillation. *1223 controls were excluded because their matched individual undergoing biopsy had been excluded (and all analyses were matched on strata, see Methods). #Marsh pathology grade 1–2. §Marsh pathology grade 0 in individuals with positive CD serology (antigliadin, antiendomysial, or anti-transglutaminase antibodies).

The final study sample consisted of 28 637 patients with CD and 141 731 controls with no record of AF before study entry. Secondary reference groups consisted of 12 911 individuals with inflammation and 3673 individuals with normal mucosa but positive CD serology (IgA EMA/transglutaminase: n= 355; IgA gliadin or IgG EMA/transglutaminase/gliadin n= 3318). During most of the follow-up of this study, gliadin was the only available CD antibody in clinical practise (for a detailed description of antibody distribution, we refer to our earlier study14).

Outcome measure

We defined AF according to relevant ICD (International Classification of Disease) codes in the Swedish National Patient Register (Discharge diagnoses) and the Cause of Death Register (ICD-7 to ICD-10:): 433.12; 427.92; 427D, I48. Our definition of AF included inpatients and outpatients, as well as individuals diagnosed with AF as a cause of death. We included primary and secondary diagnoses of AF from the Swedish Patient Register, but only AFs that were listed as the main underlying cause of death. The percentage AF-positive study participants with AF only listed as secondary diagnosis was 10.1% in both groups (CD: n= 97; controls: n= 294).

Covariates

Type 1 diabetes mellitus

We used relevant ICD codes in the National Hospital Discharge Register to identify patients with type 1 diabetes. Until version 9, the Swedish ICD classification did not distinguish between type 1 and type 2 diabetes. In this study, we therefore defined type 1 diabetes as having a hospital discharge diagnosis of diabetes before age 30 years (Table 1) (Appendix).

Autoimmune thyroid disease and rheumatoid arthritis were also defined according to relevant ICD codes in Appendix.

Education

When we adjusted for education, we used seven predefined levels, ranging from lower education than 9-year compulsory school to a PhD. This educational information was obtained from the Education Register maintained by Statistics Sweden.

Statistical analysis

Hazard ratios (HRs) for AF were estimated using Cox regression. We used an internally stratified model so that each index individual was only compared with his or her reference individuals within the same stratum (and then a summary risk estimate was calculated). The proportional hazards assumption was examined using log minus log survival curves. Follow-up started on date of first biopsy with VA and the corresponding date in the matched reference individuals.

Follow-up ended with AF diagnosis, emigration, death or on 31 December 2008, whichever event occurred first. In analyses specified a priori, we examined the risk of AF according to follow-up period (<1, 1–4.99, and ≥5 years), age (0–19, 20–39, 40–59, and ≥60 years at first biopsy) and calendar year of first biopsy (1969–1989, 1990–1999, and 2000–2008). We calculated incidence rates as the number of first recorded AF diagnoses divided by person-years until first diagnosis or end of follow-up. In separate analyses, we adjusted for education, country of birth (Nordic country vs. not Nordic country), and for ever having a diagnosis of type 1 diabetes, autoimmune thyroid disease, or rheumatoid arthritis. To increase the specificity of AF, we examined the risk of having an AF diagnosis on at least two separate occasions. In a subanalysis, we also restricted our outcome to having an inpatient diagnosis of AF or death from AF since earlier validation has focused on this subset of patients.16

To examine whether the increased risk of AF in CD was specific to CD, we compared the risk of AF in CD patients with that of individuals undergoing small intestinal biopsy but having either inflammation (Marsh 1–2) or normal mucosa (Marsh 0) but positive CD serology.

Post hoc analyses

In a first post hoc analysis, the individual's age, instead of years of follow-up since biopsy, was used as time-scale in the Cox regression. In a second post hoc analysis, thyroid disease, type 1 diabetes, and rheumatoid arthritis were modelled as time-dependent covariates (Appendix). In a third post hoc analysis,  we used logistic regression to examine the relationship between CD and the prevalence of AF (outcome is AF).

In a sample of study participants (individuals with CD: n= 2524; reference individuals: n= 12 137), we had data on smoking and body mass index (BMI) from the Medical Birth Register. The Medical Birth Register started in 1973, but smoking was not recorded until 1983.17 Smoking data are recorded at the first antenatal visit in approximately gestational week 12 according to three categories specified a priori (number of cigarettes: 0, 1–9/day, and ≥10/day). BMI is based on reported pre-pregnancy weight and height. In women with available smoking and BMI data, we adjusted for these variables in a separate analysis (see appendix).

We also estimated the risk of AF adjusted for the use of antihypertensive medication as a proxy for hypertension. Data on antihypertensive medication (using the following ATC (pharmaceutical) codes C02, C03, C07, and C08) were retrieved from the Swedish Prescribed Drug Register. This analysis was restricted to 25 942 individuals with CD and 129 866 matched controls, all of whom had a follow-up until at least 1 July 2005 when the Prescribed Drug Register started (see Appendix).

It might be hypothesized that small intestinal biopsy could trigger peri-interventional AF. To rule out that the increased risk of AF was due to the mechanical biopsy procedure we excluded all cases of AF in the first week after study entry/biopsy in a separate analysis (see Appendix). Atrial fibrillation may also be precipitated by long-term complications from CD, such as lymphoproliferative malignancy18 and associated treatments. To rule out that AF was increased as a result of lymphoproliferative malignancy, in a sub-analysis we excluded all patients who ever had a diagnosis of such a malignancy (see Appendix).

Finally, birth weight may influence the risk of both cardiovascular disease and CD. In a subset of individuals with data on birth weight, we adjusted for this variable (12 705 individuals with CD vs. 61 449 reference individuals).

Case-control study of prior atrial fibrillation and coeliac disease

Finally, we examined the risk of prior AF in patients with CD using conditional logistic regression (in this analysis, CD was our outcome measure).

SPSS 18.0 (SPSS Inc., Chicago, IL, USA) was used in all analyses. Hazard ratios with 95% confidence intervals (CIs) not including 1 were regarded as statistically significant.

Ethics

The study was approved by the Research Ethics Committee of Karolinska Institutet, Stockholm, Sweden.

Results

The majority of study participants were women (62.2%; Table 1). Some 41.2% of patients with CD had received their diagnosis in childhood (0–19 years). During follow-up, 941 individuals with CD and 2918 reference individuals received a subsequent diagnosis of AF. The median age at first AF was 75 years in both patients with CD and controls.

Coeliac disease and subsequent atrial fibrillation

Coeliac disease was a risk factor for later AF (HR = 1.34; 95% CI = 1.24–1.44) (Table 2). Risk estimates did not change after adjusting for type 1 diabetes, autoimmune thyroid disease, or rheumatoid arthritis (HR = 1.33; 95% CI = 1.23–1.43); nor did the estimate change with adjustment for the level of education (HR = 1.33; 95% CI = 1.22–1.44) or country of birth (HR = 1.34; 95% CI = 1.24–1.44).

Table 2

Risk of atrial fibrillation according to follow-up

Follow-up Observed events Expected events HR; 95% CI P-value Absolute risk/100 000 PYAR (95% CI) Excess risk/100 000 PYAR (95% CI) Attributable percentage 
All 941 702 1.34; 1.24–1.44 <0.001 321 (300–341) 81 (71–92) 25 
Year<1 112 56 1.99; 1.61–2.48 <0.001 396 (323–469) 197 (146–249) 50 
 1–4.99 286 217 1.32; 1.16–1.50 <0.001 283 (250–315) 68 (52–84) 24 
 >5 543 431 1.26; 1.14–1.39 <0.001 332 (304–359) 68 (56–81) 21 
Follow-up Observed events Expected events HR; 95% CI P-value Absolute risk/100 000 PYAR (95% CI) Excess risk/100 000 PYAR (95% CI) Attributable percentage 
All 941 702 1.34; 1.24–1.44 <0.001 321 (300–341) 81 (71–92) 25 
Year<1 112 56 1.99; 1.61–2.48 <0.001 396 (323–469) 197 (146–249) 50 
 1–4.99 286 217 1.32; 1.16–1.50 <0.001 283 (250–315) 68 (52–84) 24 
 >5 543 431 1.26; 1.14–1.39 <0.001 332 (304–359) 68 (56–81) 21 

Reference is general population comparator cohort. The attributable percentage was calculated as (1 – 1/HR).

PYAR, person-years at risk.

Age at diagnosis did not influence the risk of AF (Table 3; P for interaction = 0.324), although patients with childhood CD were younger at end of follow-up (fewer positive events), which resulted in a lower risk estimate for AF with wider CIs. Sex did not influence the risk of AF in patients with CD (Table 3; P for interaction = 0.676).

Table 3

Subgroup analyses in relation to risk of atrial fibrillation

Subgroup Observed events Expected events HR; 95% CI P-value Absolute risk/100 000 PYAR (95% CI) Excess risk/100 000 PYAR (95% CI) Attributable percentage 
Sex 
 Male 479 372 1.29; 1.16–1.43 <0.001 439 (400–479) 98 (79–117) 22 
 Female 462 330 1.40; 1.26–1.55 <0.001 251 (228–274) 71 (59–84) 28 
Age (years) 
 <20 1.27; 0.80–3.10 0.628 4 (0.5–7.0) 1 (−0.7 to 2.3) 22 
 20–39 30 19 1.55; 1.03–2.34 0.034 55 (35–75) 20 (8–32) 36 
 40–59 258 181 1.43; 1.24–1.64 <0.001 382 (335–428) 114 (88–139) 30 
 >60 648 499 1.30; 1.19–1.42 <0.001 2433 (2248–2618) 559 (469–648) 23 
Calendar period 
 –1989 182 164 1.11; 0.94–1.32 0.222 224 (192–257) 22 (12–33) 10 
 1990–1999 514 372 1.38; 1.25–1.53 <0.001 349 (319–379) 96 (80–112) 28 
 2000– 245 168 1.46; 1.26–1.68 <0.001 378 (331–425) 119 (92–145) 31 
Subgroup Observed events Expected events HR; 95% CI P-value Absolute risk/100 000 PYAR (95% CI) Excess risk/100 000 PYAR (95% CI) Attributable percentage 
Sex 
 Male 479 372 1.29; 1.16–1.43 <0.001 439 (400–479) 98 (79–117) 22 
 Female 462 330 1.40; 1.26–1.55 <0.001 251 (228–274) 71 (59–84) 28 
Age (years) 
 <20 1.27; 0.80–3.10 0.628 4 (0.5–7.0) 1 (−0.7 to 2.3) 22 
 20–39 30 19 1.55; 1.03–2.34 0.034 55 (35–75) 20 (8–32) 36 
 40–59 258 181 1.43; 1.24–1.64 <0.001 382 (335–428) 114 (88–139) 30 
 >60 648 499 1.30; 1.19–1.42 <0.001 2433 (2248–2618) 559 (469–648) 23 
Calendar period 
 –1989 182 164 1.11; 0.94–1.32 0.222 224 (192–257) 22 (12–33) 10 
 1990–1999 514 372 1.38; 1.25–1.53 <0.001 349 (319–379) 96 (80–112) 28 
 2000– 245 168 1.46; 1.26–1.68 <0.001 378 (331–425) 119 (92–145) 31 

Reference is general population comparator cohort. The attributable percentage was calculated as (1 – 1/HR).

PYAR, person-years at risk.

Patients with CD were at an increased risk of having at least two AF diagnoses (HR = 1.22; 95% CI = 1.11–1.34). When we restricted our outcome to the definition of AF diagnosis used by Smith et al.16 in their validation of AF in Swedish Registers (Hospital Discharge Register or Cause of Death Register), the HR was 1.32 (95% CI = 1.22–1.43). When the outcome measure was based only on AF in the Swedish Hospital Discharge or Outpatient Registers (and not the Cause of Death Register), the HR was 1.36 (95% CI = 1.26–1.46).

Internal comparison with other individuals undergoing biopsy

Adjusting for age, sex, and calendar year, patients with CD were at a statistically significantly higher risk of AF than patients with normal mucosa but positive CD serology (HR = 1.52; 95% CI = 1.16–1.99) but not compared with individuals with inflammation without VA (HR = 0.97; 95% CI = 0.88–1.08).

Post hoc analysis

Hazard ratios for AF did not change when we modelled thyroid disease, type 1 diabetes, and rheumatoid arthritis as time-dependent covariates (Appendix). The HRs slightly increased when we used age as time-scale instead of the year of follow-up since biopsy (Appendix).

In a post hoc analysis we used logistic regression to examine the relationship between CD and the prevalence of AF (outcome is AF). In such an analysis, an early diagnosis of AF will not influence the risk estimate more than a later diagnosis. In this analysis, patients with CD were at a 1.26-fold increased risk of later AF (95% CI, OR= 1.17–1.36). Adjustment for birth weight did not influence the risk of AF in patients with CD. Data from additional post hoc analyses are given in Appendix, and consistently supported a positive association between CD and AF.

Risk of coeliac disease in patients with atrial fibrillation before coeliac disease diagnosis

Patients with AF were at increased risk of having a later diagnosis of CD (OR = 1.45; 95% CI = 1.31–1.62).

Discussion

In this nationwide population-based study, we found a positive association between CD and AF. The magnitude of the association, however, was relatively small in that CD patients were at ∼30% increased risk of having AF diagnosed when compared with the general population. This association was seen both before and after diagnosis and strongest around the time of diagnosis. If this is a causal relationship, then this suggests that immune-mediated disorders (such as CD) might increase the risk of AF.

Comparison with earlier studies

A recent study by our group found that CD is a risk factor for cardiovascular disease19 and that cardiovascular disease is the most common cause of death in CD.10 In 2004, West et al.9 reported an increased risk of AF in CD (OR 1.26; 95% CI: 0.97–1.64) that was similar to our findings, even though patients with CD had lower rates of diagnosed hypertension and hypercholesterolaemia. In contrast to the study by West et al.,9 we found a statistically significantly increased risk of AF both before and after CD diagnosis. This finding may partly be due to greater statistical power in the present study.

Strengths and limitations

In this nationwide study, we were able to link data on biopsy-verified CD from 28 pathology departments to AF data from two other registers: the Swedish Patient Register and the Cause of Death Register. Earlier validation has shown that 95% of patients (n= 108 of 114) with VA also have CD.13 Therefore, the positive predictive value of a biopsy with VA for CD (95%) actually exceeds that of a CD diagnosis in the Swedish National Patient Register.20 In addition, small intestinal biopsy has a very high sensitivity for diagnosed CD. Small intestinal biopsy has been the standard procedure for the diagnosis of CD in Sweden since the 1970s,21 and 96–100% of all Swedish gastroenterologists and paediatricians report that they perform a biopsy in ≥90% of patients with suspected CD.13 Further, the AF diagnosis has a very high predictive value (97%) in Swedish registers.16 Due to the lack of data, we could not examine the subtypes of AF such as paroxysmal, permanent, or persistent AF. A post hoc analysis did however reveal that in CD patients with AF and anticoagulants, gastrointestinal bleedings do not seem to be more frequent than in controls with AF and anticoagulants.

More than 900 patients with CD developed AF during follow-up, giving us unprecedented statistical power to examine the association between the two diseases. Using biopsy data, we were also able to identify individuals undergoing biopsy but without having VA, which allowed us to form secondary reference groups. Internal comparisons between patients with CD and those with normal mucosa but positive CD serology further supported our assumption that macroscopic intestinal inflammation is detrimental to the risk of AF. In this secondary analysis, data on those with normal mucosa but positive CD serology however only originated from half of the Swedish population. It is difficult to judge whether this has had any impact on our results.

This study has some limitations, greatest of which is that part of the risk increase for AF might be due to ascertainment bias, i.e. around the time of CD diagnosis (or conversely, diagnosis of AF, whichever came first) there is, of necessity, an increased level of health care utilization by the patient. Our finding that the risk of AF was highest in the first year after diagnosis suggests that such a bias is present in our study. However, three facts argue against ascertainment bias as the sole explanation for our findings. First, patients with CD were at a statistically significantly higher risk of AF than were patients having a normal mucosa, even though the latter are probably to have undergone additional investigations because of ‘negative biopsy’. Secondly, patients with CD were still at an increased risk of AF more than 5 years after diagnosis. Finally, logistic regression showed an increased risk of AF in patients with biopsy-verified CD.

Another limitation is the lack of data on folic acid and B12 levels. In a subset of individuals with CD, 22% suffered from folic acid deficiency and 14% from B12 deficiency.13

Potential mechanisms

Our findings support a role of autoimmune disease in the pathophysiology of AF, potentially acting through systemic inflammation, which has consistently been linked to AF risk. A major strength of our paper is that we had access to AF data in individuals undergoing biopsy but not having VA. We believe that chronic inflammation may play a role in our findings, given that patients with CD were at a significantly increased risk of AF compared with those with normal mucosa (and positive CD serology), but not when compared with individuals with inflammation but without VA. More profound systemic inflammation (at the time of diagnosis before a gluten-free diet is introduced reducing the inflammation) could be one interpretation of why the HR is so much higher around the time of diagnosis. We also found a positive association before diagnosis of CD when inflammation caused by undiagnosed CD is likely to have been most intense.

Other potential mechanisms explaining the association of CD with AF include confounding by associated diseases, ascertainment bias, nutritional deficiencies, or shared risk factors. We adjusted for the most common associated diseases, type 1 diabetes, rheumatoid arthritis, and thyroid disease, the latter being a well-known risk factor for AF. Whereas CD is associated with low birthweight22 and low BMI,23 AF is associated with high birthweight24 and high BMI.25,26 Hence, confounding by birthweight and BMI are unlikely explanations for our findings. Although we did not have BMI or smoking data in all study participants, the fact that adjustment for these two risk factors in a small subset of women with available data increased rather than decreased the HR for AF suggests that the association between CD and AF is not mediated by BMI or smoking. Using the Prescribed Drug Register, we obtained proxy data for hypertension. Adjustment for hypertension did not alter the HR for AF and hypertension therefore seems unlikely to explain the increased risk of AF in CD. We are not aware of any other risk factors for AF16,27,28 that are also risk factors for CD. Still, we acknowledge that the absence of a rigorous assessment of hypertension and BMI in all study participants constitutes a study limitation.

Although a robust association of elevated inflammatory markers and incident AF has been reported in several studies,4–6,7 the predictive accuracy of such markers is limited4,6 likely representing the aetiological heterogeneity of AF.

Conclusions

In conclusion, AF is more common in patients with CD both before and after CD diagnosis, although the excess risk is small. Potential explanations for the increased risk of AF in CD include shared risk factors and chronic inflammation, but ascertainment bias may also have contributed.

Ethics approval

This project (2006/633–31/4) was approved by the Research Ethics Committee of the Karolinska Institute, Sweden on 14 June 2006.

J.F.L., the Corresponding Author, has the right to grant on behalf of all authors and does grant on behalf of all authors, an exclusive licence on a worldwide basis to the European Heart Journal.

J.F.L. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

L.E. was supported by a grant from Värmland County. J.F.L. was supported by a grant from the Örebro University Hospital while supervising this article. This project was supported by grants from The Swedish Society of Medicine, the Swedish Research Council – Medicine (522-2A09-195), the Sven Jerring Foundation, the Örebro Society of Medicine, the Karolinska Institutet, the Clas Groschinsky Foundation, the Juhlin Foundation, the Majblomman Foundation, Uppsala-Örebro Regional Research Council and the Swedish Celiac Society. J.G.S. was supported by the Swedish Heart-Lung Foundation. J.W. was funded by a UK National Institute for Health Research Clinician Scientist Fellowship. None of the funders had any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.

Conflict of interest: none declared.

Appendix 1

ICD coding for type 1 diabetes mellitus

Before 1997, the ICD coding for diabetes (ICD-7: 260, ICD-8: 250, ICD-9: 250) did not distinguish between type 1 and type 2 diabetes. We defined individuals with type 1 diabetes as those who were ≤30 years of age at their first hospitalization for diabetes (ICD-7-ICD-10).

ICD coding for type 1 thyroid disease

Autoimmune thyroid disease was defined as follows: ICD-7: 252.00, 252.01, 252.02, 253.10, 253.19, 253.20, 253.29, 254.00, ICD-8: 242.00, 242.09, 244, 245.03, ICD-9: 242A, 242X, 244X, 245C, 245W, ICD-10: E03.5, E03.9, E05.0, E05.5, E05.9, E06.3, E06.5.

ICD coding for rheumatoid arthritis

ICD-8: 712.3, 714.93, ICD-9: 714, ICD-10: M05, M06.

Appendix 2

Additional statistical calculations

Time-scale is age

To rule out that the positive association between CD and AF was due to an incorrect date of diagnosis of CD, we estimated the risk of AF using age as the time-scale rather than years of follow-up. Using age in the analysis did not change the HR [1.34; 95% CI = 1.25–1.44].

Using logistic regression and not taking time of follow-up into account

There is a risk that patients with a newly diagnosed CD are under increased supervision by physicians who might then detect AF just by chance. To avoid this risk, we treated our data as a case–control study with prevalence of AF as the outcome (not taking time of follow-up into account). A logistic regression revealed an OR of 1.26 [95% CI = 1.17–1.36].

Time-dependent covariates
Definition of covariates (other autoimmune diseases)

In this paper, we treated the covariates type 1 diabetes, rheumatoid arthritis, and autoimmune thyroid disease as fixed covariates (diabetes: 1 = ever having a diagnosis of diabetes; 0 = never having a record of a diagnosis of diabetes, etc.). In separate analyses, we also estimated the risk of AF when taking date of diagnosis of these covariates into account (time-dependent covariates). These analyses did not change the risk estimates:

  • Adjusting for type 1 diabetes: HR = 1.36; 95% CI = 1.28–1.44

  • Adjusting for autoimmune thyroid disease: HR = 1.36; 95% CI = 1.28–1.45

  • Adjusting for rheumatoid arthritis: HR = 1.37; 95% CI = 1.28–1.46

Anticoagulant treatment after AF and risk of gastrointestinal bleeding

We identified all individuals with a record of anticoagulant medication (ATC code ‘B01’) in the Swedish Prescribed Drug Register. The Swedish Prescribed Drug Register records outpatient medication since 1 July 2005. Hence, most medication prescribed during the study period (1969–2008) is not covered by this register. We were thereby able to identify 461 individuals with AF who started anti-coagulant treatment after AF (98 with CD and 363 controls). We then used the below ICD codes to identify gastrointestinal bleeding [we restricted our search to ICD10-codes since we were only interested in GI bleedings after the introduction of anticoagulants (starting in 2005)]. Three of 98 (3.1%) patients with CD and 5 of 363 (1.4%) controls with AF and later anticoagulant treatment had a later diagnosis of GI bleeding (Fisher's exact test: P = 0.375). This difference was not statistically significant.

  • Oesophageal bleeding: K22.8

  • K25 to K27 (stomach + small intestine) with the 3rd position ‘0’, ‘1’, ‘2’; ‘4’, ‘5’, and ‘6’

  • Acute haemorrhagic gastritis: K29.0

  • Bleeding in the anus or rectum: K62.5

  • Haematemesis: K92.0

  • Melena: K92.1

  • Non-specific GI bleeding: K92.2

Adjusting for body mass index and smoking in a subset of women with earlier pregnancy

Adjusting for pre-pregnancy, BMI and smoking in a subset of women did not change the risk estimate more than marginally (not adjusted for BMI or smoking: HR = 3.69; 95% CI = 1.18–11.56; adjusted HR = 4.14; 95% CI = 1.21–14.20).

Adjusting for antihypertensive medication as a proxy for hypertension

Adjustment for antihypertensive medication, as a proxy for hypertension did not affect HRs more than marginally in individuals with a follow-up until at least 1 July 2005 (unadjusted HR: 1.40; 95% CI = 1.24–1.59 and adjusted HR = 1.36; 95% CI = 1.21–1.58).

Adjusting for periinterventional atrial fibrillation

Hazard ratio for AF in CD excluding AF that potentially occurred within 1 week was 1.34 (95% CI = 1.24–1.44).

Adjusting for lymphoproliferative cancer

We used ICD-codes 200–204 (ICD7 to ICD9) and C82-C85 (ICD10) in our dataset to identify individuals with an inpatient diagnosis of lymphoproliferative cancer. In a separate analysis, we then excluded anyone who at some stage of life had a lymphoproliferative cancer.

Excluding individuals with a life-time history of lymphoproliferative cancer (analysis based on 28 425 individuals with CD and 141 366 controls), the HR for AF was 1.35 (95% CI = 1.25–1.45).

Adjusting for birth weight

When we restricted our analysis to individuals with data on birthweight, the unadjusted HR for AF in patients with CD was 1.22 (95% CI = 0.49–3.01), and after adjustment for birthweight (categories according to table above) it was 1.26 (95% CI = 0.51–3.14).

Appendix 3

Table

A Comparison of small intestinal histopathology classifications

Classification used in this project Villous atrophy
 
Marsh classification Type IIIa Type IIIb Type IIIc 
Marsh description Flat destructive 
Corazza et al.29 Grade B1 Grade B2 
SnoMed codes M58, D6218, M58005 M58, D6218, M58006 M58, D6218, M58007 
KVAST/Alexander classification III, partial VA IV, subtotal VA IV, total VA 
Characteristics 
 Villous atrophy ++ ++ 
 IEL# 
 Crypt hyperplasia ++ ++ 
Classification used in this project Villous atrophy
 
Marsh classification Type IIIa Type IIIb Type IIIc 
Marsh description Flat destructive 
Corazza et al.29 Grade B1 Grade B2 
SnoMed codes M58, D6218, M58005 M58, D6218, M58006 M58, D6218, M58007 
KVAST/Alexander classification III, partial VA IV, subtotal VA IV, total VA 
Characteristics 
 Villous atrophy ++ ++ 
 IEL# 
 Crypt hyperplasia ++ ++ 

References

1
Stewart
S
Hart
CL
Hole
DJ
McMurray
JJ
A population-based study of the long-term risks associated with atrial fibrillation: 20-year follow-up of the Renfrew/Paisley study
Am J Med
 , 
2002
, vol. 
113
 (pg. 
359
-
364
)
2
Benjamin
EJ
Wolf
PA
D'Agostino
RB
Silbershatz
H
Kannel
WB
Levy
D
Impact of atrial fibrillation on the risk of death: the Framingham Heart Study
Circulation
 , 
1998
, vol. 
98
 (pg. 
946
-
952
)
3
Aviles
RJ
Martin
DO
Apperson-Hansen
C
Houghtaling
PL
Rautaharju
P
Kronmal
RA
Tracy
RP
Van Wagoner
DR
Psaty
BM
Lauer
MS
Chung
MK
Inflammation as a risk factor for atrial fibrillation
Circulation
 , 
2003
, vol. 
108
 (pg. 
3006
-
3010
)
4
Schnabel
RB
Larson
MG
Yamamoto
JF
Sullivan
LM
Pencina
MJ
Meigs
JB
Tofler
GH
Selhub
J
Jacques
PF
Wolf
PA
Magnani
JW
Ellinor
PT
Wang
TJ
Levy
D
Vasan
RS
Benjamin
EJ
Relations of biomarkers of distinct pathophysiological pathways and atrial fibrillation incidence in the community
Circulation
 , 
2010
, vol. 
121
 (pg. 
200
-
207
)
5
Conen
D
Ridker
PM
Everett
BM
Tedrow
UB
Rose
L
Cook
NR
Buring
JE
Albert
CM
A multimarker approach to assess the influence of inflammation on the incidence of atrial fibrillation in women
Eur Heart J
 , 
2010
, vol. 
31
 (pg. 
1730
-
736
)
6
Smith
JG
Newton-Cheh
C
Almgren
P
Struck
J
Morgenthaler
NG
Bergmann
A
Platonov
PG
Hedblad
B
Engstrom
G
Wang
TJ
Melander
O
Assessment of conventional cardiovascular risk factors and multiple biomarkers for the prediction of incident heart failure and atrial fibrillation
J Am Coll Cardiol
 , 
2010
, vol. 
56
 (pg. 
1712
-
719
)
7
Adamsson Eryd
S
Smith
JG
Melander
O
Hedblad
B
Engstrom
G
Inflammation-sensitive proteins and risk of atrial fibrillation: A population-based cohort study
Eur J Epidemiol
 , 
2011
 
Published online ahead of print 19 March 2011
8
Burstein
B
Nattel
S
Atrial fibrosis: mechanisms and clinical relevance in atrial fibrillation
J Am Coll Cardiol
 , 
2008
, vol. 
51
 (pg. 
802
-
809
)
9
West
J
Logan
RF
Card
TR
Smith
C
Hubbard
R
Risk of vascular disease in adults with diagnosed coeliac disease: a population-based study
Aliment Pharmacol Ther
 , 
2004
, vol. 
20
 (pg. 
73
-
79
)
10
Ludvigsson
JF
Montgomery
SM
Ekbom
A
Brandt
L
Granath
F
Small-intestinal histopathology and mortality risk in celiac disease
J Am Med Assoc
 , 
2009
, vol. 
302
 (pg. 
1171
-
1178
)
11
Wei
L
Spiers
E
Reynolds
N
Walsh
S
Fahey
T
Macdonald
TM
Association between coeliac disease and cardiovascular disease
Aliment Pharmacol Ther
 , 
2007
, vol. 
27
 (pg. 
514
-
519
)
12
Ludvigsson
JF
Otterblad-Olausson
P
Pettersson
BU
Ekbom
A
The Swedish personal identity number: possibilities and pitfalls in healthcare and medical research
Eur J Epidemiol
 , 
2009
, vol. 
24
 (pg. 
659
-
667
)
13
Ludvigsson
JF
Brandt
L
Montgomery
SM
Granath
F
Ekbom
A
Validation study of villous atrophy and small intestinal inflammation in Swedish biopsy registers
BMC Gastroenterol
 , 
2009
, vol. 
9
 pg. 
19
 
14
Ludvigsson
JF
Brandt
L
Montgomery
SM
Symptoms and signs in individuals with serology positive for celiac disease but normal mucosa
BMC Gastroenterol
 , 
2009
, vol. 
9
 pg. 
57
 
15
Johannesson
I
The Total Population Register of Statistics Sweden. New Possibilities and Better Quality
 , 
2002
Örebro
Statistics Sweden
16
Smith
JG
Platonov
PG
Hedblad
B
Engstrom
G
Melander
O
Atrial fibrillation in the Malmo diet and cancer study: a study of occurrence, risk factors and diagnostic validity
Eur J Epidemiol
 , 
2009
, vol. 
25
 (pg. 
95
-
102
)
17
SNBHW
The Swedish Medical Birth Register: a summary of content and quality
2003
Stockholm, Sweden
Swedish National Board of Health and Welfare
 
18
Elfstrom
P
Granath
F
Ekstrom Smedby
K
Montgomery
SM
Askling
J
Ekbom
A
Ludvigsson
JF
Risk of lymphoproliferative malignancy in relation to small intestinal histopathology among patients with celiac disease
J Natl Cancer Inst
 , 
2011
, vol. 
103
 (pg. 
436
-
444
)
19
Ludvigsson
JF
James
S
Askling
J
Stenestrand
U
Ingelsson
E
Nationwide cohort study of risk of ischemic heart disease in patients with celiac disease
Circulation
 , 
2011
, vol. 
123
 (pg. 
483
-
490
)
20
Smedby
KE
Akerman
M
Hildebrand
H
Glimelius
B
Ekbom
A
Askling
J
Malignant lymphomas in coeliac disease: evidence of increased risks for lymphoma types other than enteropathy-type T cell lymphoma
Gut
 , 
2005
, vol. 
54
 (pg. 
54
-
59
)
21
Stenhammar
L
Hogberg
L
Danielsson
L
Ascher
H
Dannaeus
A
Hernell
O
Ivarsson
A
Lindberg
E
Lindquist
B
Nivenius
K
How do Swedish paediatric clinics diagnose coeliac disease? Results of a nationwide questionnaire study
Acta Paediatr
 , 
2006
, vol. 
95
 (pg. 
1495
-
497
)
22
Sandberg-Bennich
S
Dahlquist
G
Kallen
B
Coeliac disease is associated with intrauterine growth and neonatal infections
Acta Paediatr
 , 
2002
, vol. 
91
 (pg. 
30
-
33
)
23
Olen
O
Montgomery
SM
Marcus
C
Ekbom
A
Ludvigsson
JF
Coeliac disease and body mass index: A study of two Swedish general population-based registers
Scand J Gastroenterol
 , 
2009
, vol. 
44
 (pg. 
1198
-
1206
)
24
Conen
D
Tedrow
UB
Cook
NR
Buring
JE
Albert
CM
Birth weight is a significant risk factor for incident atrial fibrillation
Circulation
 , 
2010
, vol. 
122
 (pg. 
764
-
770
)
25
Wang
TJ
Parise
H
Levy
D
D'Agostino
RB
Sr
Wolf
PA
Vasan
RS
Benjamin
EJ
Obesity and the risk of new-onset atrial fibrillation
J Am Med Assoc
 , 
2004
, vol. 
292
 (pg. 
2471
-
2477
)
26
Rosengren
A
Hauptman
PJ
Lappas
G
Olsson
L
Wilhelmsen
L
Swedberg
K
Big men and atrial fibrillation: effects of body size and weight gain on risk of atrial fibrillation in men
Eur Heart J
 , 
2009
, vol. 
30
 (pg. 
1113
-
1120
)
27
Benjamin
EJ
Levy
D
Vaziri
SM
D'Agostino
RB
Belanger
AJ
Wolf
PA
Independent risk factors for atrial fibrillation in a population-based cohort. The Framingham Heart Study
J Am Med Assoc
 , 
1994
, vol. 
271
 (pg. 
840
-
844
)
28
Psaty
BM
Manolio
TA
Kuller
LH
Kronmal
RA
Cushman
M
Fried
LP
White
R
Furberg
CD
Rautaharju
PM
Incidence of and risk factors for atrial fibrillation in older adults
Circulation
 , 
1997
, vol. 
96
 (pg. 
2455
-
2461
)
29
Corazza
GR
Villanacci
V
Zambelli
C
Corazza
GR
Villanacci
V
Zambelli
C
Milione
M
Luinetti
O
Vindigni
C
Chioda
C
Albarello
L
Bartolini
D
Donato
F.
Comparison of the interobserver reproducibility with different histologic criteria used in celiac disease
Clin Gastroenterol Hepatol
 , 
2007
, vol. 
5
 (pg. 
838
-
843
)

Supplementary data

Comments

0 Comments