The paper by Kingham and Parker (see page 120) provides a strong take home message regarding the importance of establishing associations between diseases which, although relatively rare, have an impact on the health of a relatively young population. The authors have demonstrated an association between the two disorders and suggest that patients with one disease should be tested for the other. Is this advice justified?

Justification for such targeted screening depends on three things: the robustness of the association; the simplicity and reliability of the diagnostic tests; and the evidence that early detection benefits the patient.

The study aims to make a rigorous assessment of the previously described association between coeliac disease and primary biliary cirrhosis,1 and establish relative prevalences of the two diseases. Gastroenterologists in South Wales have kept a register of specifically defined gastrointestinal diseases since 1984, using ICD coding from all in- and outpatient and day case attendances in the department of gastroenterology. The population under study is accurately defined by use of postcode and ONS statistics. These show a relatively stable population of approximately 250 000 individuals (only 0.2% population growth and little migration or immigration). Multiple case finding methods are used to identify all possible cases of each disease and inclusion criteria are clearly specified. Possible reasons for underdetection of cases are examined.

The study identified 143 patients with coeliac disease and 67 with primary biliary cirrhosis. Point prevalences in 1995 were 54 per 100 000 population for coeliac disease and 20 per 100 000 for primary biliary cirrhosis. Four patients were diagnosed with both diseases, giving a point prevalence of 1.6 per 100 000. Those not detected clinically were diagnosed either by investigation of coeliac patients with abnormal liver function tests, or by investigation of patients with primary biliary cirrhosis with features of malabsorption, anaemia, a family history of coeliac disease, or a positive antireticulin antibody test.

This study confirms and defines the association of these two immunologically mediated diseases. The method is a good example of the new trend towards the application of simple epidemiological principles to describe accurately the incidence and prevalence of diseases.2 In this study the study period and population are well defined. Criteria for defining each disease are clear and objective. Multiple case finding methods are used. Appropriate, simple statistical methods are used. Flaws and possible causes for underdiagnosis are considered. The prevalence rates are comparable with those found in other studies, both for the individual diseases3 ,4 and for the rates of disease association.

These figures indicate that around 3% of patients with coeliac disease may develop or have primary biliary cirrhosis, whereas 6% of those with primary biliary cirrhosis may develop or have underlying coeliac disease. Possible mechanisms for this association are not discussed in depth. The authors conclude with recommendations for screening for primary biliary cirrhosis in coeliac disease by testing for antimitochondrial antibody and testing for antigliadin antibody in all cases of primary biliary cirrhosis in order to detect coeliac disease.

The proposed screening methods are important as they come close to fulfilling the requirements of any screening programme. These include targeting screening at a high risk population by the use of cheap, simple, acceptable tests, which are sensitive and specific. However, the other criteria that need to be fulfilled are knowledge of the natural history of the diseases involved when detected by screening and proof that intervention favourably affects prognosis. These are more problematic, both in this case and with any screening programme. This is because studies of the natural history of a disease and the efficacy of any treatment modality are rarely performed on the patient cohort diagnosed via a screening programme. Thus, the disease course and response to treatment may be different in screened cases compared with those presenting with symptoms or signs of the disease.

In the case of screening those with coeliac disease for primary biliary cirrhosis, 3% of patients with coeliac disease may be affected. The antimitochondrial antibody test is both sensitive and specific for identifying cases of primary biliary cirrhosis5 and, of course, non-invasive. The natural history is relatively well defined5 ,6 and double blind, randomised control trials have confirmed the benefit of treatment with ursodeoxycholic acid.7 The success of liver transplantation in these patients is undisputed. The application of these studies to a screened population is acceptable, as they have included patients diagnosed while in the asymptomatic phase of the illness, who are unlikely to differ notably from a screened population.

The case for screening those with primary biliary cirrhosis for coeliac disease is similarly compelling. Initial testing is non-invasive: antigliadin antibody testing is sensitive, and endoscopic small bowel biopsy is specific. The natural history and increased risk of malignancy (with an excess of lymphomas) in untreated cases is well documented, as is the rapid improvement of nutrition, anaemia and osteoporosis with a gluten free diet.8 However, most studies relate to cases presenting with symptoms or complications of coeliac disease. The increased risk of malignancy in an asymptomatic, screened population, and the effect of dietary therapy, is uncertain. However, the correction of malnutrition, anaemia and osteoporosis is of particular importance in primary biliary cirrhosis, as these are all complications of the disease itself and have considerable impact upon both disease outcome and success of transplantation. In addition, those patients undergoing liver transplantation are at increased risk of malignancy (particularly lymphoma) from immunosuppression and thus identification of any potentially correctable risk is of particular merit.

In conclusion, this is an important paper because it has established the true association of coeliac disease and primary biliary cirrhosis and the need for screening for one disease when the other is identified. It serves as a model for the correct application of epidemiological principles to the study of relatively rare diseases. It may also be used as a model to illustrate the importance and potential benefits of a correctly designed screening programme.