Radiology: Lung lesion follow-up recommendationsroom for improvement
Screening CT provides high sensitivity for the detection of lung nodules; however, most nodules are benign, requiring serial CT follow-up. “Therefore, interpretations of subsequent screening CT scans are as crucial as the interpretation of baseline CT scans because radiologists must not only detect new abnormalities but also follow the status of any pre-existing indeterminate nodules,” wrote Satinder Singh, MD, of the department of radiology at University of Alabama Hospitals in Birmingham, and colleagues.
Singh et al acknowledged the documented variability among radiologists in the interpretation of baseline screening CTs, and designed the study to evaluate variability in determining changes in pre-existing nodules at follow-up CT scans and in recommendations for further evaluation of the abnormality.
The study focused on a subset of patients enrolled in the National Lung Cancer Screening Trial and included cases of noncalcified nodules at least 4 mm at baseline. The dataset, which was comprised of studies performed between September 2003 and August 2005, oversampled nodules interpreted as showing growth or changes in margins or attenuation.
Nine radiologists, with an experience range of 12 to 42 years, read the studies and recorded bidimensional nodule measurements, the nodule’s dominant attenuation pattern and edge characteristics. They also noted whether the nodule was present at the baseline CT and if it had changed from baseline to follow-up. Finally, readers provided follow-up recommendations.
Among the 100 cases, all of the radiologists agreed that five nodules were not present at baseline. Furthermore, all nine radiologists agreed that the nodule was present at baseline in 76 of 95 nodules, reported Singh et al. Eight readers agreed on the presence of the nodule at baseline for 12 nodules. “In the seven remaining nodules, five to seven readers agreed that the nodules were present at baseline,” wrote Singh.
In 21 of the 76 nodules considered by all readers to be present at baseline, at least two readers disagreed with the majority of readers regarding whether growth had occurred, continued the authors.
Readers were also asked to record diagnostic confidence, and the data analysis revealed that readers were more confident of noted changes in size than in changes in attenuation or margins. They recorded growth changes as definite 72 percent of the time, attenuation changes as definite 39 percent of the time and margin changes as definite 41 percent of the time.
Regarding follow-up recommendations, radiologists provided similar patterns of recommending high-level follow-up. However, low-level or no follow-up recommendations were more variable, according to the authors.
“The radiologist’s follow-up recommendation is probably a more important end result of serial CT screening interpretation than is the assessment of change,” according to Singh and colleagues. In this study, radiologists agreed less on the categories of follow-up recommendations than for the assessment of change. But Singh did note that there was higher agreement for nodules requiring high-level follow-up and suggested that in low-level and no follow-up cases, radiologists may have been less concerned about a distinction because all patients received annual screening CT studies.
Singh and colleagues acknowledged several shortcomings to the study. Specifically, radiologists might have responded differently in an actual clinical environment. They pointed out that technical developments such as thinner slices, automated computer-aided diagnostic measurements, image magnification and training in a standard measurement approach might have reduced variability.
The researchers concluded that reader variability for follow-up CT mirrored earlier findings related to variability in nodule classification at baseline CT, with “moderate to substantial” agreement on nodule growth and screening results but lower agreement with respect to follow-up recommendations, particularly for no follow-up and low-level follow-up cases.