An association between primary biliary cirrhosis (PBC) and coeliac disease now seems well established. Since the original description by Logan and colleagues,1 there have been several reports indicating an association and these have been strengthened by larger epidemiological studies.2-11 In this issue (see page 736), Sørensen and colleagues publish their analysis of the prevalence of PBC in two populations of patients with coeliac disease, and confirm the association with a standardised incidence ratio of PBC in patients with coeliac disease of about 26. That separate evaluation of the two populations, Danes and Swedes (two countries with very different prevalences of coeliac disease), gave similar incidence ratios is strong evidence for a real association. If for no other reason, this paper deserves widespread recognition because it highlights the clinical insights that come from accurate and complete national morbidity and mortality registers. We in the United Kingdom would do well to learn from the Scandinavian example.

Last year Kingham and Parker showed, in a well defined population in Swansea, South Wales, that the prevalence of PBC in those followed for coeliac disease was 3% and the prevalence of coeliac disease in those with PBC was 6%.11 In Northern Ireland, routine screening of 67 patients with PBC showed that 11% had IgA endomysial antibodies and of these, four agreed to have a duodenal biopsy which showed villous atrophy, giving a minimum prevalence rate of 7%.8These figures are considerably higher than those given by Sorensenet al who found 24 cases of PBC (0.3%) among 8631 coeliac patients followed in the Danish and Swedish populations. This 10-fold discrepancy between the Welsh and Scandinavian data may in part be due to real differences and in part to different methods. In Wales, all patients with PBC were screened for coeliac disease if they had signs or symptoms of malabsorption (steatorrhoea, haematinic deficiency, a positive family history or antigliadin antibodies present in serum); coeliac patients were screened for PBC if they had abnormal biochemistry. It is not clear in Sorensen et al’s paper whether all patients with coeliac disease were systematically evaluated for PBC. That neither series prospectively measured antimitochondrial antibodies in those with coeliac disease suggests that both figures are likely to be underestimates.

Do these observations mean those patients with either PBC or coeliac disease should be screened for the other? The answer is probably yes. Certainly, I believe that patients with PBC should be screened for coeliac disease: as is evident from the clinical studies some cases of coeliac disease were recognised only when specifically looked for. Moreover, the clinical and biochemical consequences of undiagnosed coeliac disease may be wrongly attributed to progression of the underlying liver disease. How best to screen is uncertain. Floreani and colleagues12 found that antigliadin antibodies were present in up to 7% of patients with PBC but low antibody titres were more likely to be secondary to liver damage and only high titre antibodies were associated with coeliac disease. Antiendomysial antibodies are much more specific for coeliac disease.13Whether a duodenal biopsy is necessary to confirm the diagnosis is more controversial and indeed a duodenal biopsy may be contraindicated in those rare instances of PBC where the thrombocytopenia or prolonged clotting (uncorrectable by parenteral vitamin K) makes the chance of bleeding too high. My own prejudice that duodenal biopsy should be undertaken to confirm or refute the diagnosis is based purely on anecdote. We recently investigated one woman with PBC and high titres of antiendomysial antibodies in whom repeated duodenal biopsies showed no evidence of coeliac disease. Although there are no data on the cost-effectiveness of screening patients with PBC for coeliac disease, screening is certainly justified clinically. Recently, two patients with PBC have been referred to our liver unit for transplantation because of deteriorating liver tests, lethargy and diarrhoea; coeliac disease was diagnosed and treated with a consequent improvement so that transplantation was no longer needed.

Screening patients with coeliac disease for PBC is more problematic, especially if they are asymptomatic. There are many causes of abnormal liver tests in these patients. It is easy and relatively cheap to screen for PBC as antimitochondrial antibodies can be detected in over 95% of cases, but the benefits of diagnosing a progressive disease that cannot be cured is uncertain. As yet there is no definitive treatment for PBC. Ursodeoxycholic acid (UDCA) is the only treatment licensed for PBC and although remarkably free of side effects it will not prevent progression even though most evidence suggests it may slow the rate of progression. Clinical studies have concentrated on symptomatic patients with PBC so it is not yet established whether UDCA has a beneficial effect when given in early PBC.14 Thus, for the asymptomatic patient, little may be gained by diagnosing an untreatable condition and giving the patient the burden of another problem. Conversely, there may be advantages for the patient in making the diagnosis. Symptoms and signs of PBC, such as diarrhoea, upper abdominal pain, lethargy, or abnormalities in liver tests, may be inappropriately attributed to coeliac disease or its complications. Obviously, if the patient is experiencing symptoms of PBC, clarifying the diagnosis will allow for a more rational approach to management.

One of the intriguing questions raised by these observations is why there should be an association between the two diseases. There are several features which the two conditions have in common (table 1) including an association with autoimmune diseases, familial occurrence and an HLA association.15 ,16 However, closer examination suggests that the differences are greater than the similarities: both conditions have a weak HLA association, but with different phenotypes. The primary HLA association of coeliac disease is with the HLA genotypes DQA1*0501 and DQB1*0201 which encode the DQ2 heterodimer which is in linkage equilibrium with the extended HLA haplotype HLA A1,B8,DR3,DQ2. This may account for part of the association of coeliac disease with other autoimmune diseases.17 No such clear HLA association with A1,B8,DR3 exists for PBC.18 ,19 There may be a correlation between the duration of exposure to gluten and the risk of developing autoimmune disease.20 It has been suggested that the generation of the autoantibodies to gliadin and endomysial antigens is an autoimmune one.16 In both diseases there is a female preponderance but this is much more noticeable in PBC. The associated autoimmune diseases differ: of the autoimmune diseases associated with coeliac disease, dermatitis herpetiformis is not closely associated with PBC and diabetes mellitus rarely so.15 Conversely, thyroid disease, seen in up to 20% of patients with PBC, is less common in those with coeliac disease.21 Although diabetes mellitus is associated both with PBC and coeliac disease, only in coeliac disease is the association with insulin dependent diabetes. Again, although both conditions do have an increased incidence in first degree members, the figure varies greatly. PBC is associated with hypergammaglobulinaemia and coeliac disease with selective IgA deficiency. More tellingly, however, coeliac disease, which may be diagnosed at any age, is more common in children whereas PBC has not been diagnosed in children.14 ,15

Despite these differences, it is tempting to use the association to consider whether the pathogenesis of coeliac disease can suggest clues for the pathogenesis of PBC. A simplified pathogenesis of coeliac disease is that gluten, absorbed into the lamina propria, is presented by the dendritic cells to local, sensitised lymphocytes which then stimulate antibody, cytokine and cell mediated enterocyte damage.16 Whether PBC is also triggered by an exogenous antigen is far from certain. Bacteria and viruses have been implicated in the pathogenesis of PBC but few are convinced that any definite trigger has been identified, never mind account for the link between PBC and the antimitochondrial antibody.22

It may be that there is a congenital or acquired defect of immunoregulation in patients with PBC which also promotes an immune response to gliadin. An alternative explanation, raised by Kingham and Parker, is that IgA may be involved. In epithelial cells11IgA antibodies play a major role in mucosal defences and are potentially powerful participants in inflammatory processes. IgA is transported through epithelial cells, allowing antigens or even viruses to enter these epithelial cells.23 ,24 Although both diseases can be postulated to result from exposure to a foreign antigen with genetic and environmental factors determining the time of presentation and rate of progression, only in coeliac disease is the trigger identified. The trigger in PBC remains elusive.