A prospective radiological study of thin-section computed tomography to predict pathological noninvasiveness in peripheral clinical IA lung cancer (Japan Clinical Oncology Group 0201)

Abstract

Purpose: Pathological noninvasiveness needs to be precisely predicted in preoperative radiological examinations of patients with early lung cancer for the application of limited surgery.

Patients and methods: Patients with clinical T1N0M0 peripheral lung cancer were recruited. Radiological findings of the main tumor were evaluated as to ground-glass opacity with thin-section computed tomography. The primary end point was specificity, i.e., the proportion of patients with radiologically diagnosed invasive lung cancer to patients with pathologically diagnosed invasive lung cancer. The precision-based planned sample size was 450. We expected that the lower limit of the 95% confidence interval (CI) for specificity should be satisfied in ≥97% of patients.

Results: We enrolled 811 patients from 31 institutions between December 2002 and May 2004. The primary end point was evaluated in 545 patients. The specificity and sensitivity for the diagnosis of pathologically diagnosed invasive cancer were 96.4% (161/167, 95% CI: 92.3-98.7%) and 30.4% (115/378, 95% CI: 25.8-35.3%), respectively, i.e., a negative result. Nevertheless, the specificity for lung adenocarcinoma ≤2.0 cm with ≤0.25 consolidation to the maximum tumor diameter was 98.7% (95% CI: 93.2-100.0%), and this criterion could be used to radiologically define early adenocarcinoma of the lung.

Conclusions: Although our predetermined criterion for specificity was not statistically confirmed, radiological diagnosis of noninvasive lung cancer with a thin-section computed tomography scan corresponded well with pathological invasiveness. Radiological noninvasive peripheral lung adenocarcinoma could be defined as an adenocarcinoma ≤2.0 cm with ≤0.25 consolidation.