A combination of tumor and molecular markers predicts a poor prognosis in lung adenocarcinoma

Whether patients with stage IA-IIA lung adenocarcinoma require conventional chemotherapy is still a controversy. An ideal metastasis risk prediction model in lung adenocarcinoma is valuable for determining the prognosis and giving timely, individualized treatment.

Analyzing the clinical cases of 153 lung adenocarcinoma patients using an χ² test, Kaplan-Meier survival curves, and a multivariate logistic regression analysis, we selected the most valuable factors for determining metastasis and constructed metastasis prediction models. We confirmed the importance of the tumor markers (CEA, NSE) and a molecular marker (CAMKII) as independent prognostic factors in lung adenocarcinoma. The result of a five-year survival status was significantly associated with CAMKII and CEA (P < 0.05). A nomogram was created using CEA, NSE, CYFRA 21-1, and CAMKII to estimate the metastasis probability for individuals, specifically, 78 stage I lung adenocarcinoma patients were used to verify the effectiveness of the nomogram. Using machine learning, LASSO selected the subset of variables that minimized the predictive error of the outcome, including CEA, NSE, CYFRA 21-1, CAMKII, tumor size, histologic type, lymph node status, smoking, and age. A ten-fold cross-validation showed the average accuracy of this model was 86.208%, with an area under the curve of 0.857, a sensitivity value of 0.840, and a specificity value of 0.873.

Using both complementary methods, the predictive models illustrated that the combination of tumor markers and a key molecule to predict the prognosis of lung adenocarcinoma patients in early stages is valuable. The postoperative transfer rate of stage I patients can be effectively predicted by these complementary methods.