Longitudinal CT-derived growth metrics associated with invasive lung adenocarcinoma in subsolid pulmonary nodules

Abstract Background Serial computed tomography (CT) is used to monitor subsolid pulmonary nodules, but whether longitudinal changes add diagnostic information beyond the most recent CT findings remains unclear. We characterized growth and densification across the lung adenocarcinoma spectrum and evaluated their incremental value for preoperative identification of invasive adenocarcinoma (IAC). Methods This retrospective, single-center study included 354 patients, each with one surgically resected baseline subsolid nodule and CT examinations at least 365 days apart. Pathologic diagnoses were atypical adenomatous hyperplasia/adenocarcinoma in situ (AAH/AIS; n = 147), minimally invasive adenocarcinoma (MIA; n = 116), and IAC ( n = 91). Annualized volume and mass growth, consolidation-to-tumor ratio (CTR) change, volume doubling time (VDT), and mass doubling time were calculated. Baseline, dynamic, primary parsimonious combined, and last-static CT logistic models were internally validated using 1,000 bootstrap resamples. Results Across AAH/AIS, MIA, and IAC, median annualized volume growth was 16.52, 91.34, and 345.10 mm³/year; annualized mass growth was 0.01, 0.05, and 0.28 g/year; and VDT was 2063.89, 934.65, and 738.32 days, respectively (all P < 0.001). The optimism-corrected areas under the receiver operating characteristic curve (AUCs) were 0.937 for the dynamic model, 0.935 for the primary parsimonious combined model, and 0.941 for the last-static CT model, versus 0.871 for the baseline model (all Holm-adjusted P ≤ 0.0047 for the apparent AUC comparisons). Adding annualized mass growth and CTR change to the last-static CT model did not significantly improve fit or discrimination (likelihood-ratio P = 0.148; DeLong P = 0.668). In an exploratory subgroup with low solid-component burden at last follow-up CT ( n = 210; 11 IAC events), annualized mass growth was associated with IAC after adjustment for CTR change (per 0.1 g/year: odds ratio [OR], 2.478; 95% confidence interval [CI], 1.530–4.013; P < 0.001). Conclusions In this selected surgical cohort, CT tumor burden and growth were associated with pathologic invasiveness. Dynamic metrics improved discrimination over baseline, but their addition to the last-static CT model did not yield a statistically significant improvement. The subgroup finding for annualized mass growth is exploratory and requires external validation.