Crohn's and colitis numbers vary sharply in papers drawing on the same UK survey

This page is research from an investigation by Brian Deer for The Sunday Times of London and the UK's Channel 4 Television into a campaign linking the MMR children's vaccine with autism. | Go to part I: The Lancet scandal | Go to part II: The Wakefield factor

The table here compares data from abstracts and papers by Scott Montgomery, Andrew Wakefield, et al, apparently derived from the same survey: "BCS70", which reported on all those living in England, Scotland and Wales born between April 5 and 11 1970, of which 13,099 were mailed in 1995/6. The abstracts [1] were identical 250-word texts from Gut and Gastroenterology supplements. The papers were from [2] the British Medical Journal; [3] Gastroenterology; and [4] the American Journal of Gastroenterology

The Wakefield group's commission to the BCS70 organisers was to extract data on the numbers of respondents who, in 1995/6, said they suffered from the inflammatory bowel diseases [IBDs], Crohn's and ulcerative colitis. Cases were confirmed with physicians (and for the abstracts reportedly also by histopathology), and in [1] - [3] unconfirmed cases excluded. But as this table shows, the numbers vary - and sharply in the Gastroenterology figures, cited later by Wakefield as a key part of the justification for his MMR crusade

[Note 1] Abstracts. Two identical 250-word abstracts were presented by the Wakefield group in the spring of 1997, described as based on the BCS70 survey: first in the printed programme of a meeting of the British Society of Gastroenterology (Gut); second in a programme for a meeting of the American Gastroenterological Association (Gastroenterology). These abstracts report the highest numbers for confirmed IBDs from the BCS70 cohort. The confirmations of subjects' reports are stated in the abstracts to be by reference to the subjects' physicians, and also by histopathology. These figures should presumably be taken, therefore, to be very accurate and authoritative. However, no subsequent paper in the series here reports the same high numbers, nor makes claims of laboratory confirmation. March 18-21 1997, Brighton, and May 11-14 1997, Washington DC.

[Note 2] British Medical Journal. This paper simply sets out the methodology of BCS70, explains confirmation proceedures with doctors (to which later papers refer), and performs calculations on what is said to be the prevalence of IBD in this cohort. The paper says that there is a "lack of significant association of both social class and sex with inflammatory bowel disease," but explains that this "may be a function of the small number of cases". Prevalence of inflammatory bowel disease in British 26 year olds: national longitudinal birth cohort.

[Note 3] Gastroenterology. The Gastroenterology paper was cited in 2005 by Andrew Wakefield in justification of his 1998 call for replacing MMR with single vaccines administered at 12-month intervals, to prevent autism. The BCS70 cohort passed through childhood before MMR was introduced to the UK in 1988, so there is no direct relevance. Moreover, analysis of the survey data in this paper found no link between inflammatory bowel disease and either prior measles infection or vaccination, contradicting his core theory, which was that measles virus caused IBD and at least some forms of autism.

Nevertheless, Wakefield claimed that the Gastroenterology paper reports findings upon which he based his call for the British government to "suspend" MMR. Relevance is mined from just three cases of Crohn's and four cases of ulcerative colitis where the respondent had apparently previously had natural, or "wild", measles and mumps infection within the same year of life (eg age 3). This was judged by parental recall when the respondent was 10, and must be presumed to be potentially very unreliable, given the nature of the means of gathering this information (with local health visitors asking mothers a multitude of questions, up to perhaps eight years later), and a lack of precision surrounding the clinical symptoms and signs of measles. The authors nevertheless claim statistical significance for this tenuous, convoluted, subgroup analysis of 0.044 (Crohn's) and 0.004 (UC), and combine them to claim 0.001 for IBD. No compensation is allowed for multiple subgroup analyses. Of course, IBD has never been shown to have any relationship to autism, which isn't mentioned in any of this series of papers.

A total of 2784 questionnaires were excluded from the total received. The Gastroenterology paper, which is essentially a collection of retrospective subgroup analyses, such as "mumps infection before age 2 years" and "measles and mumps infection in the same year of life", explains: "After exclusion of cohort members with missing data for the key variables required by these analyses, the number of cohort members was 7019."

The authors say: "Despite the relatively small number of cohort members with IBD at age 26 years, the quality of these unique data may provide us with an unparalleled insight into the etiology of ulcerative colitis and Crohn's disease. BCS70 is one of the few sources that can provide detailed, reliable, and unbiased information about childhood infections in patients with adult-onset IBD."

Material relating to this paper, and apparently to the conference abstracts above, was considered by a special panel of the UK Medical Research Council in March 1998. The minutes report the following conclusion, among others: "The study suggested that there might be a small proportion of IBD cases related in some way to exposure to two paramyxovirus infections within the same year. However, the relationship would only be valid if it were assumed that there had been no selection bias, no recall bias, no prior undiagnosed health problems amongst the group identified, and that misdiagnosis of infection or gut condition had not occurred." Oddly, IBD numbers, again, appear to be different to those given in the abstracts and papers in the table above. According to the minutes, the Wakefield group reported only 15 cases of Crohn's disease from the BCS70 cohort, which can't be reconciled with anything in the published papers. As to the fragility of the data, the minutes note with regard to Crohn's disease: "It was suggested that if one case was changed, all statistical significance would be lost." The panel concluded: "In summary, there was currently no evidence relating measles vaccine or MMR to inflammatory bowel disease." Paramyxovirus infections in childhood and subsequent inflammatory bowel disease.

[Note 4] American Journal of Gastroenterology. The apparent loss of 46 survey respondents in this paper, compared with the previous ones, isn't explained. Nor is it clear whether this explains a fall of two reported cases of Crohn's disease and five cases of ulcerative colitis - which together would mean that seven of the "missing" 46 had IBD (15%). Unlike the conference abstracts, and the BMJ and Gastroenterology papers, which base calculations on physician-confirmed cases of IBD, this paper uses figures for unconfirmed, patient-reported, cases. Confirmed case data for the paper's various calculations aren't shown, other than to state the same overall confirmed figures as in the BMJ paper. The reason for this change in the methodology isn't explained, although the paper states that - despite the numbers worked with being very small indeed - the results wouldn't be "significantly" different if the calculations had been performed on confirmed cases of IBD, as in previous papers in this series.

[4a] The total of 9757 apparent respondents to the survey are whittled down for the first important calculation to 7616, who are those subjects reported as having complete vaccination data available.

[4b] The total used for analysis is whittled down again by introducing the concept of "potential confounding factors", which are described as "sex, social class at birth and crowding in childhood". Those for whom this data was said to be incomplete were excluded from analysis, bringing the analysed cohort down to 7319.

[4c] To perform the calculations, the number of subjects reported to have been vaccinated in childhood against measles are recorded as 4430 (in the text) or 4431 (in a table). Despite all this effort, relating the numbers with IBD to the numbers vaccinated, which was the aim of the project, is said to have produced no statistically significant association.

[4d] Statistical significance, however, is claimed - including in this paper's opening abstract - for a "statistically significant trend (p=0.040) with increasing age of measles vaccination for risk of Crohn's disease." This claim, squeezed from the data, is based on only 11 cases of Crohn's, eight of whom are reported to have been vaccinated before they were two, and one each reported to have been vaccinated at two, three and four years old. Measles vaccination and inflammatory bowel disease: a national British cohort study.

Possible conclusion. Findings from this series of papers - one of which [3] was used in 2005 by Wakefield to justify his call for single shots, administered at 12-month intervals to prevent autism - are within the margin of possible error from variations in the overall cohort and in the numbers recorded as having self-reported, or physician- (and histopathologically-) confirmed, inflammatory bowel disease. In a number of instances, data apparently pointing, weakly, to possible links between infections or vaccinations and subsequent disease, appears to have been forced beyond the point at which the authors might reasonably have concluded that it's as likely to confuse as to illuminate.

Analysis in progress. Any amendments or corrections are very welcome. Please contact Brian Deer through this website. Thanks