Impact of computed tomography window settings on clinical T classifications and prognostic evaluation of patients with subsolid nodules

Abstract

Abstract OBJECTIVES To investigate the impact of lung window (LW) and mediastinal window (MW) settings on the clinical T classifications and prognostic prediction of patients with subsolid nodules. METHODS Seven hundred and nineteen surgically resected subsolid nodules were reviewed, grouping into pure ground-glass nodules (n = 179) or part-solid nodules (n = 540) using LW. Interobserver agreement on nodule classifications was assessed via kappa-value, and predictive performance of the solid portion measurement in LW and MW for pathological invasiveness and malignancy were compared using receiver-operating characteristic analysis. Cox regression was used to identify prognostic factors. Prognostic significance of T classifications based on LW (c[l]T) and MW (c[m]T) was evaluated by Kaplan–Meier method after propensity score matching. The performance of c(m)T for discrimination survival was estimated via the concordance index (C-index), net reclassification improvement and integrated-discrimination improvement. RESULTS By adopting MW, 124 part-solid nodules were reclassified as pure ground-glass nodules, and interobserver agreement improved to 0.917 (95% confidence interval 0.888–0.946). The solid portion size under MW more strongly predicted pathological invasiveness (P = 0.030), but did not better predict pathological malignancy. For remaining 416 part-solid nodules, c(l)T and c(m)T were both independent risk factors. c(m)T led to T classifications shifts in 321 nodules (14 upstaged and 307 downstaged) with no significant prognostic difference existing between the shifted c(m)T and matching c(l)T group after propensity score matching. The corrected C-index was improved to 0.695 (0.620–1.000) when adopting c(m)T with no significant difference in net reclassification improvement (P = 0.098) and integrated-discrimination improvement (P = 0.13) analysis. CONCLUSIONS As there is no significant benefit provided by MW in evaluating clinical T classification and prognosis, the current usage of LW is appropriate for assessing subsolid nodules.