Analysis of interobserver and intraobserver variability in CT tumor measurements.

Abstract

The purpose of this study was to evaluate the variability between radiologists interpreting thoracic and abdominal/pelvic CT scans in selecting specific sites of metastatic tumor for measurement (indicator lesions) and to assess interobserver and intraobserver variability in tumor measurement. Three separate experienced radiologists were asked to review 24 combined thoracic and abdominal CT scans in patients with metastatic tumor. Each radiologist was asked to identify the indicator lesions representative of each patient's tumor bulk. In the second phase of the study, 105 specific foci on 26 combined thoracic and abdominal CT studies (including the original 24) were reviewed twice by the same three radiologists. Up to eight foci were randomly identified per patient, and each observer was asked to determine the slice with the maximum diameter for each tumor focus and to measure it in three dimensions (maximum diameter, its perpendicular, and length). A total of 132 tumor sites were present on the CT studies in phase I, all of which were chosen by at least one observer as an indicator lesion. Of the 116 of these that were separate and nonoverlapped, 57 (49%) were measured by only one observer, whereas 32 (28%) and 27 (23%) were measured by two or all three observers, respectively. Observers were more inclined to pick round or defined/well-defined lesions rather than irregular, oval, or poorly defined ones, although this tendency was not statistically significant. The second phase of the study showed considerable interobserver variability (15%) in CT tumor measurement, which was worse for poorly defined and irregular lesions. Intraobserver variability in measuring individual foci was less (6%). Radiologists interpreting thoracic and/or abdominal/pelvic CT scans for metastatic cancer should measure and report a significant number of each patient's tumor sites, especially larger ones in different anatomic areas. When interpreting a follow-up CT scan of a patient with metastatic cancer, the interpreting radiologist should remeasure the indicator lesions on the previous and on the follow-up CT scans, especially when the results will change the patient's treatment response category.