Capsule Endoscopy in Chronic Gastrointestinal Bleeding

The study by Ell et al. [1] is certainly of major importance in the evaluation of wireless capsule endoscopy, as it was conducted with a prospective, comparative and blinded design. However, there are two points that require correction and explanation.

The main finding of the study was a higher rate of bleeding sources identified by capsule endoscopy in comparison with push enteroscopy: 21 of 32 versus nine of 32. The authors state that this finding is significant, with an impressive P value of P < 0.0001. This P value was obtained by using Student's paired t-test; however, this test is only suitable for data that can be measured in numbers, such as age or hemoglobin levels. Data on diagnostic tests are usually dichotomous - either you find the source of the disease or not. Such data may be coded as ”1” and ”0”, but the summary result is a proportion (21 of 32 = 66 %), not a mean (0.66). In consequence, the t-test (which works on the basis of means) is inappropriate for dichotomous data [2]. Instead, rates and proportions should be compared with the chi-squared test or Fisher's exact test. We have done this for the data published in the study, and the result is a P value of 0.0054 - fortunately still remaining highly significant.

In addition, the results are easier to understand from a clinical point of view when a relative risk (RR) is calculated. In the present study, 21 of 32 divided by nine of 32 gives an RR value of 2.3. Capsule endoscopy (as opposed to push enteroscopy) thus evidently doubles the chances of finding the source of chronic gastrointestinal bleeding. The 95 % confidence interval around this figure ranges from 1.3 to 4.3, thus indicating that the diagnostic yield with capsule endoscopy is clearly superior, although the size of the effect contains some uncertainty.

Our second critique relates to the study design. The authors do not describe the sequence or timing of the two procedures. Did all patients first undergo push enteroscopy, which was then followed by CE; or did some patients have capsule endoscopy as the first procedure? Thirteen patients showed signs of active bleeding during the diagnostic work-up, but bleeding may end spontaneously during the work-up procedures, thus influencing the ability to detect its source. An ideal study would therefore randomize the sequence of procedures [3]. Perhaps the authors could describe the sequence in which capsule endoscopy and push enteroscopy were carried out in their patients and also explain whether the 13 patients who had active bleeding on entering the study also had signs of active bleeding when the endoscopic capsule passed through the gastrointestinal tract.

References

• 1 Ell C, Remke S, May A. et al . The first prospective controlled trial comparing wireless capsule endoscopy with push enteroscopy in chronic gastrointestinal bleeding.  Endoscopy. 2002;  34 685-689
• 2 Greenhalgh T. Statistics for the non-statistician, 1: different types of data need different statistical tests.  BMJ. 1997;  315 364-366
• 3 Knottnerus J A, van Weel C, Muris J WM. Evaluation of diagnostic procedures.  BMJ. 2002;  324 477-480