Comparative clinical utility of the 2024 Fleischner Society updated glossary versus traditional classifications for...

Current pulmonary subsolid nodules (SSNs) classifications remain controversial in terms of interobserver agreement and predictability of pathological subtypes and prognosis. The 2024 Fleischner Society updated glossary introduced a novel structural-component-based classification, necessitating validation of its clinical utility.

Patients with SSNs who underwent resection were retrospectively analyzed from February 2005 to December 2018. Three radiologists independently categorized SSNs on preoperative CT using three classification approaches: traditional lung window [LW]-based and mediastinal window [MW]-based classifications (pure vs. mixed ground-glass nodules according to solid component detected by LW or MW settings, respectively), and Updated classification (simple vs. complex nodules based on structural components). Intra-/interobserver agreement was assessed using κ statistics. Predictors of pathological subtypes and prognosis were identified using logistic and Cox regression, validated via ROC and time-dependent ROC analyses. Survival outcomes were evaluated with Kaplan-Meier curves.

Among 1,316 patients, the MW-based SSN classification showed good-to-excellent intra- and interobserver agreement (κ range: 0.789-0.858), surpassing the moderate-to-good agreement of the Updated (κ range: 0.442-0.655) and LW-based SSN classifications (κ range: 0.527-0.672). For discriminating invasive adenocarcinoma from non‑invasive lesions, MW consolidation‑to‑tumor ratio (MW‑CTR; odds ratio [OR] = 205.351, 95% CI: 36.579-1152.831) and MW‑based SSN classification (OR = 4.158, 95% CI: 2.684-6.442) demonstrated strong discriminative ability. Models incorporating MW-Solid size achieved the highest AUC of 0.889, significantly outperforming the standalone Updated classification (P = 0.012). In survival analysis, the MW-based SSN classification was the only independent prognostic factor among the three classification approaches when evaluated individually (adjusted hazard ratio [HR] = 3.576, 95% CI: 1.658-7.712). The SSN classification combining with MW-CTR provided superior prognostic stratification performance over time (tdAUCs: 0.765-0.846 across 2-15 years; iAUC: 0.807) compared to models using MW-Solid size or LW-Solid size (tdAUCs: 0.703-0.816 across 2-15 years; iAUCs: 0.768 and 0.772). MW-CTR (adjusted HR = 49.924, 95% CI: 21.571-115.544) and LW-CTR (adjusted HR = 53.162, 95% CI: 19.633-143.953) were robust independent prognostic factors.

MW-based SSN classification outperformed the updated and LW-based SSN classification approaches in key clinical utility, evidenced by more reliable observer agreement, optimal discrimination of invasiveness, and stronger prognostic predictive value.