Article Text
Statistics from Altmetric.com
Editor,—We were interested to read the study by Evans et al (Gut1997;40:619–22) as we have recently reported a prospective study of colitis drawing similar conclusions.1Our study, however, showed a much stronger association. Forty five of 62 (72.4%) new cases of colitis were taking non-steroidal anti-inflammatory drugs (NSAIDs) or salicylates compared with 38 (7%) of 513 of a control population sample (odds ratio 33.1, 95% confidence interval 17.31 to 63). This difference between the two papers is almost certainly owing to the fact that our group of colitics comprised a much broader spectrum of disease than those reported by Evans et al, who were clearly only studying patients with colitis in relapse.
In addition, we enquired about the usage of over-the-counter NSAIDs and salicylates which was not possible in the study design of Evanset al. We would maintain that exposure misclassification could easily have occurred if over-the-counter usage of these compounds was not considered.
The incident cases of Evans et al were de novo cases of colitis but it is not apparent why they do not regard NSAIDs as causal in these cases, as they acknowledge that these drugs have been reported as causing de novo colitis in the introductory paragraph of their paper. In our study, which was exclusively of de novo colitis, many cases have recovered completely following withdrawal of NSAIDs and salicylates suggesting a cause/effect relationship. In these patients histological differentiation from ulcerative colitis “due to inflammatory bowel disease” was not possible in the majority of cases. However, in a minority (30%) subsequent independent histological assessment has revealed some of the criteria of NSAID colitis.2 These criteria of NSAID colitis had not been established at the time of the study (1989–93) by Evans et al. Thus the interpretation of their cases as Crohn’s or ulcerative colitis according to the criteria of Lennard-Jones3 may not be entirely valid.
In our opinion future epidemiological studies of the possible association of colitis and drugs need standardised histopathological review, preferably by a pathologist blinded to the drug history of the patient, which must be stringently determined by the referring clinician. It is probable that many patients classified as having “colitis due to inflammatory bowel disease” will in fact be suffering from colitis induced by NSAIDs. These drugs may be some of the most important environmental factors in the pathogenesis of colitis.
References
Reply
Editor,—Gleeson and Warren agree with the conclusions of our recent paper, that exposure to NSAIDs is associated with colitis. Where they disagree is on the strength of the association between exposure and colitis. Unfortunately, the abstract of the study they cite contains insufficient data to make a meaningful comparison between the two studies.1-1 However, one possible reason for the difference is that their control group (513 attendees at a minor injuries clinic), recalled less prior NSAID exposure than their cases. Cases (who are ill) often recall past exposure more than controls (who are not ill). Recall bias is a common and difficult problem in epidemiology. The population based dispensed prescribing data we used as our measure of exposure did not suffer from this bias.
We agree that the NSAIDs aspirin and ibuprofen which are available in low dose over-the-counter formulas, might have resulted in some exposure misclassification, but we judge this to be a small effect when compared with recall bias.
Gleeson and Warren suggest that the difference between the two studies can be explained by differences in the spectrum of disease studied. In terms of severity, this might be true. Our cases were all severe enough to warrant hospitalisation. It is possible that milder colitis might be more strongly associated with exposure to NSAIDs. Gleeson and Warren go on to suggest that NSAID colitis might be a more specific condition. Nearly all of our cases had firm diagnoses of ulcerative colitis or Crohn’s colitis. It is possible that a specific NSAID induced, less severe form of colitis exists. We did not study this.
In our study incident colitis was quite strongly associated with current NSAID exposure (odds ratio (OR) 2.96) whereas non-incident colitis was not (OR 1.16). Causality is one possible explanation of an association. Our study was carried out to clarify the “signal” of possible NSAID associated colitis suggested by case reports. It seems illogical to suggest that we should then use case report evidence to strengthen our causality inference. Case reports provide a low standard of evidence of causality. They are anecdotes and the plural of anecdote is not data.
Arguments for a causal association depend upon the strength of the association, presence of biological gradient (dose response), consistency of the findings (lots of studies showing the same), and biological plausibility (a mechanism). We are further towards a causal association than before these studies were performed. However, we still have a lot of work to do before causality can be accepted.
The important point of our paper is that exposure to NSAIDs is associated with a severe adverse event, namely hospitalisation for incident (Crohn’s and ulcerative) colitis. A milder, specific, NSAID colitis may exist. This is an attractive hypothesis which should be tested using epidemiological studies. We would be happy to collaborate with such a venture.
References
- 1-1.↵