Autoimmune cholestatic liver disease in people with coeliac disease: a population-based study of their association


Dr J. West, Division of Epidemiology and Public Health, University of Nottingham, Queens Medical Centre, Nottingham, UK.


Background : Population data supporting an association between the autoimmune cholestatic liver diseases, primary biliary cirrhosis and primary sclerosing cholangitis and coeliac disease, is limited and at times contradictory.

Aim : To explore the relationship between coeliac disease and both primary biliary cirrhosis and primary sclerosing cholangitis within the General Practice Research Database, a UK-based longitudinal primary care database.

Methods : We identified 4732 people with diagnosed coeliac disease and 23 620 age- and sex-matched controls within the General Practice Research Database. We calculated the prevalence of primary biliary cirrhosis and primary sclerosing cholangitis for both the coeliac disease and control group.

Results : There was a higher prevalence of primary biliary cirrhosis in adults with coeliac disease, compared with controls [0.17% vs. 0.05%, odds ratio 3.63 (95% confidence interval: 1.46–9.04)]. Primary sclerosing cholangitis was also more common in the coeliac disease group [0.04% vs. 0%, fishers exact test (P = 0.03)].

Conclusions : There was a threefold or greater increase in risk of both primary biliary cirrhosis and primary sclerosing cholangitis in people with coeliac disease compared with the general population. The association with primary biliary cirrhosis was weaker than in some reports and it is difficult on the basis of this study to justify screening patients with coeliac disease for either primary biliary cirrhosis or primary sclerosing cholangitis.


The association of coeliac disease (CD) with various forms of liver disease has been described in several European and North American populations.1–14 In particular, an increased prevalence of the autoimmune cholestatic liver diseases; primary biliary cirrhosis (PBC) and primary sclerosing cholangitis (PSC), in patients with CD has been reported.

The association between CD and PBC was first described in four patients by Logan et al.3 Numerous case reports followed but formal epidemiological studies to determine the prevalence of PBC in patients with CD, as shown in Table 1, have produced mixed findings. Considerably less is known about the prevalence of PSC in patients with CD. An association between CD and PSC was first described in1988 by Hay et al.,6 in a report of three patients with PSC later diagnosed with CD. There have been no subsequent studies, however, on the prevalence of PSC in patients with CD, although several groups have published on a frequency of CD among patients with PSC, of the order of 0–2.6%.2, 5, 6, 8, 9

Table 1.  Studies of prevalence of primary biliary cirrhosis in patients with coeliac disease
AuthorAreaCoeliac populationnNumbers with PBCRelative risk
  1. n/a, not available; PBC, primary biliary cirrhosis.

  2. * Positive antimitochondrial antibody only, all four had normal liver function tests.

Bardella et al.10Milan, ItalyOne hospital clinic3361n/a
Kingham and Parker11S. WalesOne hospital clinic1433n/a
Sorensen et al.1SwedenNational in-patient register77352225
DenmarkNational in-patient register896228
Floreani et al.14Padua, ItalyOne hospital clinic1084*n/a

Interpretation of all these reports has been complicated by the increasing recognition and of serological screening for undetected CD. Data on the risk of autoimmune cholestatic liver disease in CD is at times contradictory as well as being limited. We have, therefore, carried out a population-based study to explore the relationship between CD and both PBC and PSC within the General Practice Research Database (GPRD), a UK longitudinal primary care database.


The GPRD, established in 1987, is a longitudinal primary care database and contains the computerized medical records from general practice of more than 8 million of these registered people. Participating practices must record all major diagnoses, medical events and prescriptions, and are audited regularly to ensure they are including at least 95% of this information.15 The data is then aggregated and anonymised to maintain patient confidentiality.

We extracted the records of all subjects within the GPRD between June 1987 and April 2002 with a recorded diagnosis of CD. Where possible we selected five control subjects matched to each individual with CD by age, sex, general practice and follow-up time. When selecting control subjects we excluded individuals who had any record of a gluten-free prescription or a non-specific reference to CD (e.g. ‘gluten-free diet’, ‘gluten sensitivity’). Each subject with CD was assigned a date of diagnosis defined as the date of the first record of CD. As general practitioners enter some data for important historical events retrospectively this date preceded the start of the GPRD record on some occasions. Further details about our study population have been described elsewhere.16

We calculated the prevalence of PBC and PSC for both the CD and control group by dividing the number of diagnoses by the respective number of people in each group. We compared cases and controls using either chi-squared test or Fishers exact test where appropriate and estimated odds ratios (OR) for risk of having either PBC or PSC using conditional logistic regression. We also adjusted for smoking status as cigarette smoking is thought to be positively associated with PBC17 and inversely associated with CD.18 We used stata 7.0 se (Stata Corporation, College Station, TX, USA) for all analyses.


Our study included 4732 subjects with CD and 23 620 matched controls. The groups were closely matched on age and gender, although there were more current smokers in the control group (15.4% vs. 13%, P < 0.01). Further information on the demographic details of our CD and control groups is published elsewhere.16

The prevalence of PBC was 0.17% (n = 8) in CD compared with 0.05% (n = 11) in the general population controls, giving an OR of 3.63 [95% confidence interval (CI): 1.46–9.04] as shown in Table 2. Adjusting for smoking status made little difference to the OR for PBC. The prevalence of PSC was 0.04% (n = 2) in the patients with CD. Although we were unable to calculate the OR for PSC as none of the control population had this condition there was a statistically significant difference when the prevalence in the two groups was compared using fishers exact test (P = 0.03).

Table 2.  Analysis of risk of PBC and PSC in coeliac disease compared with controls
 nOverall number with disease (%)Odds ratio95% CIAdjusted odds ratio*95% CI
  1. PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; CI, confidence interval.

  2. * Adjusted for smoking status

 Control23 62011 (0.05)1 1 
 Coeliac disease  47328 (0.17)3.631.46–9.043.691.48–9.23
 Control23 6200 (0)
 Coeliac disease  47322 (0.04)

In only three (37.5%) of the eight people with both CD and PBC did the diagnosis of PBC precede that of CD (Table 3) and in half of them both diseases were diagnosed within a year of each other. Of those eight people, six (75%) were female and five (62.5%) were non- or never smokers. The average (mean) age at which PBC was diagnosed was higher in the people with CD than in the general population controls (58.6 years vs. 54.6 years).

Table 3.  Age at diagnosis, sex and presenting complaint in people with both coeliac disease and PBC or PSC
Liver diseaseSexAge diagnosed with PBC or PSCPresenting complaintAge diagnosed with coeliac diseasePresenting complaint
  1. n/k, not known; PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; LFT, liver function tests.

  2. * Same patient.

PBCFemale32n/k 2Historical
PBC*Male64Oesophageal varices63Anaemia
PBCFemale64Abnormal LFTs68Anaemia
PSCFemale57Abnormal LFTs54n/k
PSC*Male60Abnormal LFTs63Anaemia


The results of our study support an association between CD and autoimmune cholestatic liver disease. For PBC we have demonstrated that there is a more than threefold increase in risk for those with CD compared with the general population. This was despite a higher prevalence of PBC in the control group than reported by James et al.19 in their description of PBC within the population of Newcastle, UK. As one would expect there were fewer cases of PSC, however, the prevalence within our CD group was still more than threefold the number that would be expected when using even the highest estimate of population prevalence.20

This is a large, population-based study of patients with CD from a primary care data set, and represents the largest population in which the prevalence of PSC in patients with CD has been investigated. Although there were fewer patients with CD and PBC than in the Scandinavian cohort described by Sorenson et al.,1 our study has the advantage of having a large age- and sex-matched internal control group for comparison. A limitation of our study is that it is not possible to comment on the clinical and pathological evidence on which the diagnoses of CD, PBC and PSC were made. However, the accuracy of medical diagnoses within the GPRD has been shown to be high in several validation studies, including those of gastrointestinal disease.21

It is possible that our finding of an increased prevalence of autoimmune cholestatic liver disease in patients with CD reflects greater ascertainment of one disease in the presence of the other. One might expect that the CD group were more likely to have regular medical review and/or investigations that could lead to the diagnosis of hitherto undetected liver disease. If this was the explanation for the association then one might expect the CD group to have been diagnosed with autoimmune cholestatic liver disease at a younger age than the control group. This is the opposite, however, to what we found. The mean age of diagnosis of CD was 56.8 years in those with autoimmune cholestatic liver disease, compared to 36.3 years in the patients without accompanying liver disease. Serological screening of PBC patients for CD cannot, therefore, be excluded but might account for at most three cases, as in the remainder CD was the first diagnosis. We believe that these observations together with the magnitude of the association make it unlikely that it is due to ascertainment bias, although it remains a possibility.

The prevalence of PBC in our United Kingdom CD population (0.17%) was similar to that seen in studies of other European populations, such as in Sweden (0.28%) and Denmark (0.22%)1 that also looked at diagnosed cases. Three studies in which patients with CD have been screened for PBC,10, 11, 14 have had the limitation of being relatively small (Table 1). This may account for the discrepancy in prevalence between the two separate Italian studies that screened patients with CD for PBC using antimitochondrial antibodies. Bardella et al.10 found only one case in 336 patients with CD (0.29%) whereas Floreani et al.14 found four cases among 108 patients (3.7%). Previous studies that have looked at the possible association between CD and PSC have focused on the frequency of CD in patients with PSC.2, 5, 8, 9 The results of our study suggest that as well as CD being more common in patients with PSC, PSC is more commonly seen in CD patients than in the general population; not withstanding the small number of cases of PSC in our study.

Our results suggest that people with CD are at increased risk of both PBC and PSC compared with the general population. The recent finding of IgA serum antibodies against transglutaminase in the liver tissue of patients with CD has also added biological plausibility to this association.22 Both conditions may with time progress to cirrhosis and result in the need for liver transplantation or death. A finding of abnormal liver function tests or symptoms/signs of liver disease in a patient with CD should, therefore, alert the clinician to the possibility of autoimmune cholestatic liver disease. However, until there are more effective early interventions, screening patients with CD for PBC as has been suggested23 is difficult to justify.


Thanks are due to Chris Smith and staff at EPIC (UK) for their help with data management.