The authors respond:
The letter from Frisch and Rostgaard1 seems based on a strong persuasion that the inverse association between appendectomy and ulcerative colitis (UC) is not well substantiated. There are 3 reports showing no association, 2 of which have been authored by Frisch and colleagues,2–4 in contrast to some 20 studies showing a clear inverse association. As pointed out,2,5 some of them have shortcomings, but so do the “negative” studies; short follow-up,2,3 poor quality appendectomy data,3 lack of smoking data,2,3 contamination with prevalent UC cases,2 and low power.4 Taken at face value, 2 of the “negative” studies2,4 are compatible with an important protective effect of appendectomy, given their confidence intervals (CI = 0.69–1.07 and CI = 0.6–4.7). We are confident that any sensible meta-analysis would show a clear inverse association.
Accordingly, in our paper,6 we did not attempt to reproduce this finding once again. Instead, our goal was to make a distinction between 2 families of hypotheses for the inverse association—a mutual cause or a genuine protective effect.
By the definitions used in our paper6 (no modifier codes, no mentioning of Crohn's disease), the incidence of UC increases slightly with age for both sexes. Thus, residual confounding by age would bias our measures towards the null. A new Poisson regression using 10-year age-intervals gave an unchanged asymmetry (RR = 0.64).
The apparent discrepancy in effect between persons age 40 years or older at appendectomy and those less than age 15 years at the start of follow-up is mainly explained by one thin stratum in the preappendectomy period of subjects less than 15 at start of follow-up. This subgroup contributed 88% of the expected incidence with only 4 observations. We concede that the events were too scarce for a reliable estimate of our main measure specified by age at follow-up. This limitation does not concern the main results and the interpretation of our paper, and there are no similar dominant strata in other analyses.
Our findings for patients with and without appendicitis are not logically inconsistent. If a genuine protective effect were linked to the appendectomy itself, as we think, the individuals with and without appendicitis should have similar effects. As shown, their confidence intervals overlap considerably.
It is correct that our subjects survived in the analysis only until the appendectomy. As most of them are quite young, this could not have had any major effect in our study.
We agree that one cannot advocate prophylactic appendectomy at present.6 However, if we could identify healthy subjects at high absolute risk of UC, we should not deny them the potential benefit of mounting a randomized clinical trial of prophylactic appendectomy.
Department of Internal Medicine; Odense University; Odense, Denmark; email@example.com
Department of Epidemiology; Institute of Public Health; University of Southern Denmark; Odense, Denmark
Henrik Toft Sørensen
Department of Clinical Epidemiology; Aarhus and Aalborg University Hospital; Aarhus, Denmark
1. Frisch M, Rostgaard C. Appendectomy and ulcerative colitis [letter]. Epidemiology
2. Frisch M, Johansen C, Mellemkjær L, et al. Appendectomy and subsequent risk of inflammatory bowel disease. Surgery
3. Frisch M, Gridley G. Appendectomy in adulthood and the risk of inflammatory bowel diseases. Scand J Gastroenterol
4. Woods BL, Steinberg EN, Hornung CA, et al. Does appendectomy really protect against ulcerative colitis. Gastroenterology
5. Frisch M, Biggar RJ. Appendectomy and protection against ulcerative colitis. N Engl J Med
6. Hallas J, Gaist D, Sørensen HT. Does appendectomy reduce the risk of ulcerative colitis? Epidemiology