Accuracy of plain films, and the effect of experience, in the assessment of ankle effusions

To investigate the accuracy of plain radiographs, and the effect of observer experience, in the assessment of ankle effusions compared with an MRI gold standard. Anteroposterior (AP) and lateral radiographs of the ankle of 39 patients were evaluated by four observers, ranging from first-year radiology resident to an attending musculoskeletal radiologist. Observers independently graded the lateral films from 0 to5 at one sitting, and the AP films at a second sitting. All patients had an MRI scan performed within 48 h of the ankle radiographs, on which distention of the anterior recess was used as the gold standard for an effusion. Lateral radiographs had variable sensitivity (range 17 - 63%), but specificity (81-94%) was usually high. AP radiographs similarly had variable sensitivity (15-55%), but their specificity (63-75%) was surprisingly good. Overall, sensitivity and specificity were inversely proportional and more related to individual variability than experience (observer 1, 53% and 81%; observer 2, 17% and 94%; observer 3, 63% and 88%; observer 4, 21% and 94%); however, individual sensitivity and specificity were consistent between AP and lateral radiographs (observer 1, 53% and 81%, 50% and 65%; observer 2, 17% and 94%, 15% and 75%), observer 3, 63% and 88%, 55% and 63%; observer 4, 21% and 94%, 25% and 70%. Positive predictive value was reasonably good for lateral radiographs (range 75 - 86%); however, it was fairly low for AP radiographs (38-61%). Negative predictive value was low for both lateral (50-67%) and AP (47-58%) radiographs. Accuracy was low for both AP (45-59%) and lateral (53-74%) radiographs. As expected, individual accuracy was consistently higher for lateral radiographs than for AP radiographs (observer 1, 65% and 58%; observer 2, 53% and 45%; observer 3, 74% and 59%; observer 4, 54% and 48%). For the diagnosis of ankle effusions the overall accuracy of radiographs was surprisingly low. Quite surprisingly, the diagnosis of effusions on AP radiographs was not much poorer than on lateral films. Results, however, varied little with experience and training. (orig.)