A retrospective exploratory study of the variability of radiologists’ measurements in a selected subgroup of subjects enrolled in a clinical trial
6523 A retrospective exploratory study of the variability of radiologists’ measurements in a selected subgroup of subjects enrolled in a clinical trial Background: Independent centralized review of imaging studies is an established means of validating data used in support of corporate ‘go-no go‘ decisions, regulatory approval, and post marketing claims. Despite best efforts at standardizing this process using the Response Evaluation Criteria in Solid Tumors (RECIST), there will be inherent differences in radiologic measurements. To define a concise value for the variation between a specific set of radiologists, the following study was performed. Methods: An experienced radiologist and a research intern reviewed digital Computerized Tomographic (CT) scans from 31 subjects to identify 150 primary and metastatic tumors. Tumors were categorized into 4 categories. [Category (Cat.) 1 - Defined Edge/Radially Symmetric, Cat. 2 - Defined Edge/Irregular Shape, Cat. 3 - Blurred Edge/Radially Symmetric, Cat. 4 - Blurred Edge/Irregular Shape] Fifteen radiologists independently measured each tumor. Mixed effects models controlling for different tumors were used to assess the variance. Results Overall, the between reader variations accounted for 4.15% percent of total errors for the length. The respective between reader and the random errors are 0.32 (95% CI: 0.14, 0.77) and 7.51 (95%CI: 7.08, 8.01). The data shows that the variances are smallest for category 1 tumors (total error: 1.45) and largest for category 4 tumors (total error: 20.79). The total error for categories 2 and 3 are listed in Table 1 below. Conclusions The (low reader error) / (total error) verifies that the radiologists were only responsible for a small portion of the total error, and suggests these 15 radiologists measuring a tumor will be comparable to one of these radiologists measuring the tumor many times. No significant financial relationships to disclose. [Table: see text]