Malignancy Prediction Capacity and Possible Prediction Model of Circulating Tumor Cells for Suspicious Pulmonary Lesions

Abstract

More and more undetermined lung lesions are being identified in routine lung cancer screening. The aim of this study was to try to establish a malignancy prediction model according to the tumor presentations. From January 2017 to December 2018, 50 consecutive patients who were identified with suspicious lung lesions were enrolled into this study. Medical records were reviewed and tumor macroscopic and microscopic presentations were collected for analysis. Circulating tumor cells (CTC) were found to differ between benign and malignant lesions (p = 0.03) and also constituted the highest area under the receiver operation curve other than tumor presentations (p = 0.001). Since tumor size showed the highest sensitivity and CTC revealed the best specificity, a malignancy prediction model was proposed. Akaike information criterion (A.I.C.) of the combined malignancy prediction model was 26.73, which was lower than for tumor size or CTCs alone. Logistic regression revealed that the combined malignancy prediction model showed marginal statistical trends (p = 0.0518). In addition, the 95% confidence interval of combined malignancy prediction model showed less wide range than tumor size ≥ 0.7 cm alone. The calculated probability of malignancy in patients with tumor size ≥ 0.7 cm and CTC > 3 was 97.9%. By contrast, the probability of malignancy in patients whose tumor size was < 0.7 cm, and CTC ≤ 3 was 22.5%. A combined malignancy prediction model involving tumor size followed by the CTC count may provide additional information to assist decision making. For patients who present with tumor size ≥ 0.7 cm and CTC counts > 3, aggressive management should be considered, since the calculated probability of malignancy was 97.9%.