An evidence-based medicine approach to studies of diagnostic tests: Assessing the validity of virtual colonoscopy☆☆☆★★★

New technologies are being proposed for colorectal cancer screening. In many centers, screening may be performed with new diagnostic tests, including DNA analysis of colonic cells secreted in stool¹ or “virtual colonoscopy” of the colon using computed tomography (CT) colonography.² With new technologies, rigorously designed trials should be performed and systematically reviewed before these diagnostic tests are widely implemented; the goal of this critical appraisal is to prevent the widespread use of inaccurate diagnostic tests. This article, the second in the evidence-based medicine (EBM) series, provides an EBM format for critically appraising a study of a diagnostic test (Table 1). The article follows the format of other EBM articles in the Journal of the American Medical Association and Gastroenterology.⁵ The EBM concepts have been expanded from previous articles, and examples from the gastroenterology literature have been used.

Table 1. Guidelines for the Evaluation of an Article About a Diagnostic Test

When are diagnostic tests necessary?

Is the pretest probability of a disorder very high or very low?

How do you estimate pretest probability?

Assessing study design

Was a blinded comparison between the new diagnostic test and a gold standard test performed?

Was an acceptable alternative used if no gold standard test was available?

Did all study patients get the gold standard test, regardless of the results of the new diagnostic test?

Did investigators enroll patients in whom the diagnosis was in doubt?

Was the frequency of indeterminate test results reported?

Interpreting the results

Appropriately interpret and apply data about sensitivity, specificity, positive and negative predictive values

Use likelihood ratios to maximize the utility of data from a diagnostic test study

Can these results be applied to your patient?

Are your patients similar to the patients examined in this trial?

Is the test likely to be reproducible with minimal variation in your clinical setting?

Abbreviations:  CEA , carcinoembryonic antigen, CT , computed tomography, EBM , evidence-based medicine, NPV , negative predictive value, PPV , positive predictive value