Study participants
Between October 2006 and February 2008 we collected data from small intestinal biopsy reports from all 28 Swedish pathology departments. Biopsies had been conducted between 1969 and 2008 (Table
1). We obtained data on date of biopsy, topography (duodenum or jejunum), morphology codes consistent with villous atrophy (VA) which was used as our definition for CD) and personal identity number [
20]. For each patient with CD, the government agency Statistics Sweden identified up to five reference individuals from the Total Population Register. Matching criteria were age, sex, calendar year and county. We then removed individuals with data irregularities, and thereafter had data on 29,096 individuals with CD and 144,520 reference individuals (identical to the cohort used in our earlier paper on CD and mortality) [
21].
Table 1
Characteristics of participants presented as n (%)
Total | 144,480 | 29,086 |
Age at study entry (years) |
0–19 | 58,851 (40.7) | 11,802 (40.6) |
20–39 | 26,378 (18.3) | 5311 (18.3) |
40–59 | 32,237 (22.3) | 6471 (22.2) |
60- | 27,014 (18.7) | 5502 (18.9) |
Sex |
Female | 89,523 (62.0) | 17,999 (61.9) |
Male | 54,957 (28.0) | 11,087 (38.1) |
Calendar Period |
-1989 | 20,371 (14.1) | 4104 (14.1) |
1990–1999 | 59,860 (41.4) | 12,055 (41.4) |
2000- | 64,249 (44.5) | 12,947 (44.4) |
Nordic | 136,242 (94.3) | 28,129 (96.7) |
Type 1 diabetes | 596 (0.4) | 956 (3.3) |
Autoimmune thyroid diseases | 2502 (1.7) | 1435 (4.9) |
Celiac disease
CD was defined as VA (Marsh stage 3) according to biopsy report [
22]. Having a positive CD serologic test result was not a prerequisite for CD diagnosis although 88% of individuals in a random subset of individuals with VA and available data on serology had positive CD serology at the time of biopsy [
23]. Additional details on the data collection has been published elsewhere [
23]. On average, three tissue specimens were examined for each biopsy report.
Myasthenia gravis (MG)
MG was defined according to relevant international classification of disease (ICD) codes in the Swedish National Patient Register: ICD7: 744.00; ICD8: 733.00; ICD9: 358A; ICD10: G70.0. The Patient Register started in 1964 and contains diagnoses from ICD-7 through ICD-10. The Patient Register included inpatient care but since 2001, hospital-based outpatient care has also been included. We did not have access to data on acetylcholine receptor antibody results, but National Patient Register-based MG diagnosis has been validated against the Stockholm MG Cohort. This analysis found that 147/177 diagnoses in the Patient Register could be confirmed against the Stockholm MG Cohort and the positive predictive value of MG is 83% [
13].
Individuals with MG before CD diagnosis and corresponding controls based on dates were excluded from all prospective analyses (n = 8 and n = 22 respectively). We then excluded another 2 CD patients and 2 controls due to data irregularities (uncertainties regarding date of death) and another 18 reference individuals since their index individual with CD had been excluded and analyses were performed retaining strata. Hence, there remained 29,086 individuals with CD and 144,480 matched controls.
Other covariates
Data on the following potential confounding factors were collected from the government agency
Statistics Sweden: country of birth (Nordic vs. not Nordic), educational level (≤9 years of primary school, 2 years of high school, 3–4 years of high school, college/university) and socioeconomic status according to six categories (according to the European Socioeconomic Classification, ESeC: levels 1, 2, 3 + 6, 7, 8, and 9) [
24]. Missing data on education and socioeconomic status (4 and 31% respectively) were fitted into separate categories for the statistics. We also identified individuals with T1D and autoimmune thyroid disease defined according to relevant ICD codes.
Statistical analyses
We estimated hazard ratios (HR) for MG using Cox regression. We used an internal stratification system and each stratum (consisting of one CD patient and up to five controls) was analyzed separately and a summary HR was calculated. Through visual inspection of log-minus-log curves we found that the proportional hazard assumption was valid. The attributable risk percent (the proportion of MG in patients with CD that could be explained by the underlying CD) was estimated by 1–1/HR. Follow-up started at first biopsy with CD diagnosis and at the same time in matched controls. Follow-up ended when one of the following occurred: death, emigration, first MG diagnosis, or end of study (December 31, 2009).
In a priori analyses, we examined the time-dependent risk of MG (time since CD diagnosis: < 1 year, 1–4.99 years; and > 5 years). We also performed analyses stratified by age at diagnosis of CD (≤19, 20–39, 40–59 and ≥60 years), sex, and calendar period (− 1989, 1990–99, 2000-). In a separate analysis we adjusted for country of birth, education, socioeconomic status, T1D, and autoimmune thyroid disease.
In order to examine the temporal association between CD and MG, we also calculated the odds ratios (ORs) for having a diagnosis of MG prior to the first biopsy for CD. For this analysis, we used conditional logistic regression comparing each individual with CD with his or her matched control.
A post-hoc power analysis found that we had an 80% power at a significance level of 0.05 to detect a HR of 3.1 for MG.
We used SPSS 22 to calculate statistics. P-values < 0.05 were considered statistically significant.