Nomogram and risk-score for predicting overall survival and risk stratification in patients with sarcomatoid non-small cell lung cancer: a multicenter study of 135 patients.

Background: To explore the clinical data and CT findings associated with outcomes prognosis of patients with sarcomatoid non-small cell lung cancer (s-NSCLC).

Materials And Methods: In this retrospective study, s-NSCLC patients who underwent contrast-enhanced thoracic CT from January 2013 to June 2023 at three centers were enrolled. Clinicoradiological data, including sex, age, smoking status, tumor-node-metastasis (TNM) classification, tumor size, tumor location, calcification, vacuole/cavity, hydrothorax, low-attenuation area (LAA) ratio and peritumoral ground-glass opacity (GGO) or nodules were calculated. Clinical and CT findings associated with overall survival (OS) were evaluated by a multivariate Cox regression model.

Results: A total of 135 patients with s-NSCLC in three centers were included. s-NSCLC patients were more likely to be elderly male smokers. The mean age and tumor size at diagnosis were 62 years and 5.8 cm, respectively. The median survival time of patients with s-NSCLC was 9 (95% CI: 7, 11) months. The 1-, 2- and 3-year OS rates of the s-NSCLC patients were 36.6%, 26.7% and 21.4%, respectively. s-NSCLC is often peripherally located (98/135, 70.4%). Calcification and vacuole/cavity were rare in s-NSCLC lesions. The hydrothorax was present in 36/135 (26.7%) s-NSCLC patients. The s-NSCLC lesions usually presented with LAA (87/135, 80.6%), and the median LAA ratio was 30.8% (interquartile range, IQR: 10.6%, 50.7%). In the multivariate Cox regression analysis, smoking status (hazard ratio, HR = 1.668 [95% CI: 1.040, 2.678]), tumor size (HR = 1.818 [95% CI: 1.167, 2.832]), peritumoral GGO or nodules (HR = 2.064 [95% CI: 1.090, 3.909]) and M stage (HR = 2.479 [95% CI: 1.476, 4.164]) were associated with increased mortality, whereas surgery (HR = 0.467 [95% CI: 0.274, 0.797]) was associated with decreased mortality. We constructed a nomogram for predicting the 1-, 2- and 3-year OS rates of s-NSCLC patients, and the AUCs were 0.867, 0.905 and 0.911, respectively. Risk-score = 0.512×smoking status + 0.598×tumor size + 0.725×peritumoral GGO or nodules+(-0.761)×surgery + 0.908×M stage.

Conclusion: The multivariate Cox regression model, which includes factors such as smoking status, tumor size, peritumoral GGO or nodules, surgery and M stage, is helpful in predicting the OS and risk stratification of s-NSCLC patients.