Pulmonary Nodule Risk Model Findings

> Research explainer: This briefing examines verified primary research published 62 days before the briefing date. It is not a same-day research update and does not provide medical advice.

Evidence

This research explainer examines a source published on 2026-06-18. The retrospective study included 568 patients with solitary pulmonary nodules who underwent surgical resection at the Jiading Branch of Shanghai General Hospital between January 2021 and January 2024. Pathology classified 422 nodules as malignant and 146 as benign. Of the malignant nodules, 406 were lung adenocarcinomas, making this an adenocarcinoma-predominant cohort rather than a broad cross-section of all pulmonary nodules.

Eligible nodules were solitary lesions measuring 5–30 mm, with malignant or benign status determined from tissue obtained during thoracic surgery. The study excluded people with prior nodule treatment, multiple nodules, pleural effusion, atelectasis, lymph-node enlargement, ambiguous pathology, or metastatic tumors. Investigators collected demographic data, tumor-marker levels, and high-resolution CT features. Two experienced radiologists assessed nodule characteristics, including density and morphologic signs; disagreements were resolved through departmental discussion.

In multivariable analysis, female sex was associated with higher odds of malignancy (OR 1.729, 95% CI 1.114–2.685; P=0.015). Three CT morphology features were also independently associated with malignancy: lobulation sign (OR 2.250, 95% CI 1.301–3.890; P=0.004), burr sign (OR 2.965, 95% CI 1.740–5.053; P<0.001), and vascular bundle sign (OR 6.721, 95% CI 1.971–22.922; P=0.002). Mixed ground-glass density was associated with malignancy relative to pure ground-glass density (OR 2.627, 95% CI 1.050–6.572; P=0.039), whereas solid density had lower odds relative to pure ground-glass density (OR 0.169, 95% CI 0.102–0.279; P<0.001).

The authors combined these factors into a prediction model. Its reported area under the receiver-operating-characteristic curve was 0.81 (95% CI 0.767–0.849). At the reported operating point, sensitivity was 91.0% and specificity was 47.3%. These measures describe model performance within the study’s data and selected cohort; they do not by themselves show how the model would perform in a different hospital, population, or nodule-evaluation pathway.

Analysis — Interpreting a Surgical Cohort Model

The central signal is not that any one imaging feature determines whether a nodule is malignant. Rather, the analysis found that a set of demographic and CT features was statistically associated with pathology results among people already selected for surgery. The vascular bundle sign had the largest reported odds ratio among the listed factors, but its confidence interval was also wide, from 1.971 to 22.922, indicating substantial uncertainty around the estimated magnitude. The direction of association for nodule density also depends on the specified reference category: pure ground-glass nodules, not all nodules generally.

The model’s performance profile matters for interpretation. Sensitivity of 91.0% means the reported threshold identified a high proportion of malignant nodules in this cohort, while specificity of 47.3% means many benign nodules would also be categorized as positive at that threshold. An AUC of 0.81 indicates discrimination in the analyzed sample, but it is not evidence of improved outcomes. The findings are therefore best understood as a cohort-specific risk-modeling result that may inform research on assessment variables in similar adenocarcinoma-predominant surgical populations.

Limitations

Selection is the major constraint. Every participant underwent surgical resection and had a definitive pathology result, so the study population differs from people whose nodules are observed, evaluated nonsurgically, or identified through other pathways. The single-hospital Chinese cohort may not represent other health systems or more diverse populations. Its predominance of adenocarcinoma further narrows generalizability.

The retrospective design can identify associations but cannot establish that sex or CT features cause malignancy. The abstract reports no external-validation results, so the reported AUC, sensitivity, and specificity should not be assumed to transfer elsewhere. Specificity was modest at the reported threshold, and the study does not demonstrate a benefit for screening outcomes, surgical decisions, or patient outcomes. This briefing reports research findings only and does not provide medical advice or patient-specific conclusions.