Statistics from Altmetric.com
Editor,—We read with interest the study by Smytheet al (Gut1998;43:571–4); however, we feel that on the basis of the data presented, the pronouncement of death for the test may perhaps be a little premature.
Firstly, despite the low sensitivity and specificity reported, the test still had positive and negative predictive values (66.7 and 57% respectively) which would be clinically useful in allowing patients to come to an informed decision regarding cholecystectomy.
Secondly, the authors conclude that there is no statistical difference between the positive and negative test groups in terms of their outcome after cholecystectomy. The relative benefit of the test expressed as the odds ratio is 2.7 with a 95% confidence interval from 0.7 (no benefit) to 10 (great benefit)—hence the authors cannot reach a conclusion with a study of this size regarding the usefulness of the test. We estimate that if the proportions of subjects in the various outcome groups remained the same, 148 subjects would be needed for the study to have 80% power with an odds ratio of 2.7. The ideal number of subjects for this study would depend on the size of difference in clinical outcome, which would be useful to detect. Obviously, if the true odds ratio is higher than 2.7 then fewer subjects would be required, but at a more realistic but still clinically useful odds ratio of less than 2.7 an even larger study would be necessary.
Thirdly, we obtained different figures for sensitivity, specificity and p value for the χ2 test (with Yates’ correction) of 75%, 47%, and p=0.26, respectively, with respect to symptomatic improvement after cholecystectomy—perhaps the authors’ definition of these parameters was different to our own interpretation of their data.
Fourthly, the results of this study may not be applicable to a wider clinical setting. Cholecystectomy was performed on a highly selected group of subjects, after a variable time period and with the cholecystokinin provocation test result already known. It might have been more appropriate to offer all subjects cholecystectomy or to randomise them to management with or without knowledge of the test result.
Editor,—We agree entirely with the concept that larger numbers in this study (as in any other) would yield narrower confidence intervals. In our study the cholecystokinin provocation test had low sensitivity and specificity (the use of Yates’ correction is controversial) and we disagree that these positive and negative predictive values are clinically useful in counselling patients regarding outcome after cholecystectomy.
We also agree that a randomised blind study may be a more objective way of assessing the usefulness of this test; however, most patients in the study underwent cholecystectomy for symptoms and we have assessed symptomatic relief separately from cholecystokinin positivity. Indeed, most patients were given saline first (they were blinded to the infusion) and their symptoms recorded. We suggest that the comments raised by Campbell and colleagues do not detract from the fact that almost 50% of patients with acalculous biliary pain experience relief after cholecystectomy and the cholecystokinin provacation test is unable to predict those with good outcome.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.