When using CT to screen patients for lung cancer, the Vancouver risk calculator (VRC) has a higher sensitivity, specificity and accuracy than the Lung Imaging Reporting and Data System (Lung-RADS), according to a new study published in Radiology. The authors noted, however, that it would be unwise to replace Lung-RADS altogether.
“With Lung-RADS, screening CT examinations are categorized primarily based on the presence and size of pulmonary nodules by using a scale of 0 to 4 that indicates an increasing likelihood of malignancy. Separate criteria exist for solid and nonsolid nodules,” wrote Charles S. White, MD, department of diagnostic radiology at the University of Maryland in Baltimore, and colleagues. “An alternative approach is to focus more specifically on a series of nodule and patient characteristics to evaluate the likelihood of malignancy by applying a risk calculator model to CT lung cancer screening. The VRC, reported by McWilliams et al., is based on data from two Canadian lung screening trials and is the most widely cited effort using the risk calculator approach.”
The authors then compared the ability of Lung-RADS and the VRC to predict the risk of malignancy in nodules identified from the National Lung Screening Trial (NLST). Overall, Lung-RADS had a sensitivity of 87 percent, specificity of 83 percent and accuracy of 83 percent, but the VRC’s numbers were 93 percent, 90 percent and 90 percent, respectively. The VRC’s sensitivity, specificity and accuracy were also higher on a per-patient basis. For subsolid nodules, on the other hand, Lung-RADS was more specific than VRC, but much less sensitive.
“Of note, the VRC showed a trend toward higher sensitivity for all types of nodules, both solid and subsolid,” the authors wrote. “In a screening setting, higher test sensitivity is typically preferred even at the cost of some specificity and accuracy.”
The team also noted that these statistics may make it seem like providers should replace Lung-RADS with VRC, but “there are compelling reasons why this conclusion is premature.” One reason is that radiologists like the numeric categories used by Lung-RADS, in part because they are similar to the BI-RADS system. Also, Lung-RADS is “an integral part of submission” to CMS-approved registries. So then how can providers take advantage of these findings?
“Rather than replacement, the roles of the VRC most likely include providing a second and complementary method to assess the likelihood of nodule malignancy during CT screening and in the longer term to contribute to future revisions of Lung-RADS by suggesting additional parameters from the VRC that might be incorporated into Lung-RADS to improve its predictive power,” the authors wrote.