Likelihood Ratios to Determine 'Does This Patient Have Appendicitis?': Comment and Clarification-Reply

In Reply.  —Dr Witt and colleagues and Dr Klovning have pointed out that Table 3 in our article may cause confusion when reconciling the sensitivities and specificities with the LRs. These astute observers have noted that these parameters are not related in the manner one would expect (ie, positive LR=sensitivity/[1-specificity]). The reason for this disconnection is a matter of subsets. The sensitivities and specificities were derived from a meta-analysis of all studies included in the article, whereas the LRs were derived from a subset of those studies with a more appropriate patient sampling. That is, the sampling was of patients presenting with abdominal pain instead of patients undergoing surgery for suspected appendicitis.In retrospect, Table 3 should have displayed parameters derived from the same studies. The accompanying Table is a revision of Table 3 in our original article, now with sensitivities and specificities derived from the same subset of better-quality studies from