A novel nomogram for predicting mediastinal lymph node metastasis in non-small cell lung cancer: a retrospective analysis.

BACKGROUND

Accurate assessment of lymph node metastasis (LNM) is crucial for preoperative staging and treatment planning in patients with lung cancer. While previous research has explored LNM risk in non-small cell lung cancer (NSCLC), clinical validation of multifactorial predictive models is lacking. This study aimed to develop and validate a dynamic nomogram for predicting LNM in NSCLC patients.

METHODS

We retrospectively analysed 619 NSCLC patients and divided them into training (70%) and validation (30%) groups. Univariate and multivariate ordinal logistic regression analyses identified predictive factors for LNM. Variables were selected via least absolute shrinkage and selection operator (LASSO) regression. A dynamic nomogram was developed on the basis of logistic regression results, and its performance was evaluated through receiver operating characteristic (ROC) curve analysis, calibration plots, and decision curve analysis (DCA). The model was further validated with 1,000 bootstrap resamples.

RESULTS

Independent predictors of LNM were ferritin, carbohydrate antigen 125 (CA125), carcinoembryonic antigen (CEA), carbohydrate antigen 199 (CA199), EGFR exon 19 deletion, tumor size, and tumor location. The nomogram exhibited excellent discriminative ability, with an area under the ROC curve (AUC) of 0.846 in the training group and 0.828 in the validation group. DCA indicated greater net benefits across various LNM risk thresholds.

CONCLUSIONS

This study presents a dynamic nomogram that integrates EGFR exon 19 deletion and serum ferritin levels, enhancing preoperative staging and aiding treatment decisions for NSCLC patients.