Nomogram for the preoperative prediction of Ki-67 expression and prognosis in stage IA lung adenocarcinoma based on clinical and multi-slice spiral computed tomography features

Abstract

Objective: This study developed and validated a nomogram utilizing clinical and multi-slice spiral computed tomography (MSCT) features for the preoperative prediction of Ki-67 expression in stage IA lung adenocarcinoma. Additionally, we assessed the predictive accuracy of Ki-67 expression levels, as determined by our model, in estimating the prognosis of stage IA lung adenocarcinoma.

Materials and methods: We retrospectively analyzed data from 395 patients with pathologically confirmed stage IA lung adenocarcinoma. A total of 322 patients were divided into training and internal validation groups at a 6:4 ratio, whereas the remaining 73 patients composed the external validation group. According to the pathological results, the patients were classified into high and low Ki-67 labeling index (LI) groups. Clinical and CT features were subjected to statistical analysis. The training group was used to construct a predictive model through logistic regression and to formulate a nomogram. The nomogram's predictive ability and goodness-of-fit were assessed. Internal and external validations were performed, and clinical utility was evaluated. Finally, the recurrence-free survival (RFS) rates were compared.

Results: In the training group, sex, age, tumor density type, tumor-lung interface, lobulation, spiculation, pleural indentation, and maximum nodule diameter differed significantly between patients with high and low Ki-67 LI. Multivariate logistic regression analysis revealed that sex, tumor density, and maximum nodule diameter were significantly associated with high Ki-67 expression in stage IA lung adenocarcinoma. The calibration curves closely resembled the standard curves, indicating the excellent discrimination and accuracy of the model. Decision curve analysis revealed favorable clinical utility. Patients with a nomogram-predicted high Ki-67 LI exhibited worse RFS.

Conclusion: The nomogram utilizing clinical and CT features for the preoperative prediction of Ki-67 expression in stage IA lung adenocarcinoma demonstrated excellent performance, clinical utility, and prognostic significance, suggesting that this nomogram is a noninvasive personalized approach for the preoperative prediction of Ki-67 expression.

Keywords: Ki-67; Lung adenocarcinoma; Nomogram; Tomography; X-ray computed tomography.

Fig. 1

Study flow diagram

Fig. 2

A middle-aged woman with stage IA lung adenocarcinoma. (A, B) Computed tomography images showing mixed ground-glass opacities in the right lower lobe, with clear boundaries; no lobulation, spiculation, or pleural indentation sign (→); and a bronchial air sign within the lesion, with a maximum nodule diameter of 2.54 cm. Pathological image showing invasive adenocarcinoma, with predominant acinar growth (hematoxylin and eosin staining ×200). (D) Pathological image showing low Ki-67 expression (Ki-67 immunohistochemistry ×40). According to the nomogram, the total score was 80, and the preoperative probability of high Ki-67 expression was approximately 9%. The postoperative pathological diagnosis was a Ki-67 index of 5%, indicating low Ki-67 expression

Fig. 3

An elderly man with stage IA lung adenocarcinoma. (A, B) Computed tomography images showing a solid nodule in the right upper lobe, with shallow lobulation (→), spiculation, no pleural indentation sign, and a maximum nodule diameter of 2.12 cm. Pathological images showing invasive adenocarcinoma, predominantly of the solid type and partly of the acinar type (hematoxylin and eosin staining ×200). (D) Pathological image showing high Ki-67 expression (Ki-67 immunohistochemistry ×40). According to the nomogram, the total score was 224, and the probability of high Ki-67 expression was approximately 85%. The postoperative pathological diagnosis was a Ki-67 index of 25%, indicating high Ki-67 expression

Fig. 4

ROC curves of patient sex, tumor density type, lobulation, maximum nodule diameter and high Ki-67 expression in stage IA lung adenocarcinoma in the training cohort according to the predictive model. AUC, area under the receiver operating characteristic curve

Fig. 5

Nomogram model based on clinical and computed tomography features for predicting high Ki-67 expression in stage IA lung adenocarcinoma. According to the four indexes (sex, tumor density type, tumor lobulation, and maximum nodule diameter) of each patient, the vertical lines between each index and the nomogram points were drawn to obtain the score for each index. The scores of the five indicators are then summed to obtain the total score. Finally, a vertical line is drawn between the total score and the nomogram risk to predict the probability of high Ki-67 expression in stage IA lung adenocarcinoma

Fig. 6

Calibration curve of the nomogram model based on clinical and CT features for predicting high Ki-67 expression in stage IA lung adenocarcinoma. Internal (A) and external (B) validation show a well-calibrated curve between the predictive model and actual observations, indicating the high discrimination and accuracy of the model. Receiver operating characteristic curves of the nomogram prediction model for internal (C) and external (D) validation. Decision curve analysis to evaluate the clinical net benefit of the preoperative prediction model for high Ki-67 expression in stage IA lung adenocarcinoma. Decision curve analysis (E) was used to evaluate the clinical net benefit of the preoperative prediction model for high Ki-67 expression in stage IA lung adenocarcinoma

Fig. 7

Kaplan–Meier curves depicting recurrence-free survival according to the pathological Ki-67 labeling index (LI) (A) and the predicted Ki-67 LI (B). After stratifying by the Ki-67 LI (low or high), patients with both pathological and predicted high Ki-67 LI values had poorer RFS than patients with low Ki-67 LI values (P < 0.001)