Is measles infection associated with Crohn's disease?

The current evidence does not prove a causal link

The cause of Crohn's disease is likely to be multifactorial, and great interest was generated by two Swedish studies suggesting a high risk of Crohn's disease in those exposed to measles in utero.1 2 The report in this week's issue from Nielsen et al (p 196),3 however, is not alone in suggesting that there is no increased risk.

The two Swedish papers studied largely the same group of patients. Both studies were the result of two index cases of Crohn's disease noted to have been exposed to measles in utero (accounting for two out of four cases in the second study). The first report, in 1994, compared the expected and observed month of birth in patients with Crohn's disease born in 1945-54 in relation to the peak months of measles epidemics. The standardised incidence ratio was 1.46 (95% confidence interval 0.83 to 2.21) for future development of Crohn's disease for births during the three months after the peak incidence of measles. The second paper, in 1996, described a study of maternal measles infection in a cohort of 25 000 babies born in 1940-9.2 Of four such cases three subsequently developed Crohn's disease.

In apparent support of the hypothesis, Thompson et al found a relative risk of 3.01 (1.45 to 6.23) for Crohn's disease among a British cohort of people vaccinated with live attenuated measles vaccine compared with a matched, unvaccinated group.4 However, up to 74% of the original cohort were lost to follow up, and methods of follow up varied between the groups. This report led to concerns that vaccination with live, attenuated measles vaccine could confer the same risk as exposure to measles in utero.5

Now, however, four further studies have failed to confirm evidence of an association. Nielsen et al examined the health records of all possible cases of measles in pregnancy admitted to an infectious diseases hospital in the Copenhagen area in 1915-66.3 The offspring of 25 women who had measles during pregnancy were identified, and none had developed Crohn's disease. In 1995 Hermon-Taylor et al compared the incidence of Crohn's disease with notifications of measles infection in England and Wales, including data after the introduction of measles vaccines.6 They found no association. Also in 1997 Jones et al reported a case-control study of a large cohort of individuals exposed to viral infections during gestation, including 47 people exposed to measles in utero.7 Follow up data on 88% found no cases of inflammatory bowel disease in the index cases, but two among the controls (one with Crohn's disease). A case-control study by Feeney et al in 1997 compared measles vaccination rates in 140 patients with inflammatory bowel disease (83 with Crohn's disease) and matched controls and found no association.8

To reconcile these discrepancies we need an understanding of the investigation of causation. Significant associations were established in the original studies, but associations may be artefactual, indirect, or causal. Artefactual associations may result from chance. The inclusion of index cases which have generated the hypothesis leads to reporting bias, especially if the numbers are small. Spurious associations may also result from differences in methods and completeness of data collection and recall bias (cases more often recall exposure to possible causal factors than controls). Some of these factors may have affected Thompson's vaccination study.4 An indirect association is one in which the factor and disease are associated through a common third factor, such as malaria and altitude, linked through mosquitoes. There are no apparent indirect links in these studies.

A refinement of Koch's postulates has led to the development of six criteria to evaluate the likelihood that an association is causal, the first three of which are the most important.9 Firstly, the greater the strength of the association (the higher the relative risk) the more likely it is that a factor is causal. A dose-response gradient and a consistent association-that is, one repeated in other studies-also suggest causality. The specificity of the association-whether the occurrence of the factor predicts the presence of the disease-the correct temporal association, and the biological plausibility of the association are also relevant. Only the last two criteria are met by the Swedish studies.

Thus, several recent studies of the association between measles and Crohn's disease have failed to confirm the original association, suggesting that the original finding was artefactual. The theory of measles as a causative factor in the development of Crohn's disease therefore cannot be upheld and should remind us of the need for rigorous methodological review when causal associations are proposed.