Lung Cancer Screening: Comparison of Established and Novel Models for Predicting Malignancy in Pulmonary Nodules
Presented During:
Sunday, May 4, 2025: 9:00AM - 4:00PM
Seattle Convention Center | Summit
Posted Room Name:
Poster Area, Exhibit Hall
Abstract No:
P0180
Submission Type:
Abstract Submission
Authors:
Fabian Doerr (1), Konstantinos Grapatsas (1), Natalie Baldes (1), Filiz Oezkan (2), Michael Forsting (3), Dirk Theegarten (4), Kaid Darwiche (2), Hubertus Hautzel (5), Martin Stuschke (6), Christian Taube (2), Marcel Wiesweg (7), Martin Schuler (7), Servet Bölükbas (1)
Institutions:
(1) University Medical Center Essen, Ruhrlandklinik, Department of Thoracic Surgery, Essen, Germany, (2) University Medical Center Essen, Ruhrlandklinik, Department of Pulmonary Medicine, Essen, Germany, (3) University Medical Center Essen, Institute for Diagnostic and Interventional Radiology, Essen, Germany, (4) University Medical Center Essen, Institute of Pathology, Essen, Germany, (5) University Medical Center Essen, Department of Nuclear Medicine, Essen, Germany, (6) University Medical Center Essen, Department of Radiation Therapy, Essen, Germany, (7) University Medical Center Essen, Department of Medical Oncology, Essen, Germany
Submitting Author:
Fabian Doerr
-
Contact Me
University Medical Center Essen, Ruhrlandklinik, Department of Thoracic Surgery
University Medical Center Essen, Ruhrlandklinik, Department of Thoracic Surgery
Co-Author(s):
Konstantinos Grapatsas
-
Contact Me
University Medical Center Essen, Ruhrlandklinik, Department of Thoracic Surgery
University Medical Center Essen, Ruhrlandklinik, Department of Thoracic Surgery
Natalie Baldes
-
Contact Me
University Medical Center Essen, Ruhrlandklinik, Department of Thoracic Surgery
University Medical Center Essen, Ruhrlandklinik, Department of Thoracic Surgery
Filiz Oezkan
-
Contact Me
University Medical Center Essen, Ruhrlandklinik, Department of Pulmonary Medicine
University Medical Center Essen, Ruhrlandklinik, Department of Pulmonary Medicine
Michael Forsting
-
Contact Me
University Medical Center Essen, Institute for Diagnostic and Interventional Radiology
University Medical Center Essen, Institute for Diagnostic and Interventional Radiology
Dirk Theegarten
-
Contact Me
University Medical Center Essen, Institute of Pathology
University Medical Center Essen, Institute of Pathology
Kaid Darwiche
-
Contact Me
University Medical Center Essen, Ruhrlandklinik, Department of Pulmonary Medicine
University Medical Center Essen, Ruhrlandklinik, Department of Pulmonary Medicine
Hubertus Hautzel
-
Contact Me
University Medical Center Essen, Department of Nuclear Medicine
University Medical Center Essen, Department of Nuclear Medicine
Martin Stuschke
-
Contact Me
University Medical Center Essen, Department of Radiation Therapy
University Medical Center Essen, Department of Radiation Therapy
Christian Taube
-
Contact Me
University Medical Center Essen, Ruhrlandklinik, Department of Pulmonary Medicine
University Medical Center Essen, Ruhrlandklinik, Department of Pulmonary Medicine
Marcel Wiesweg
-
Contact Me
University Medical Center Essen, Department of Medical Oncology
University Medical Center Essen, Department of Medical Oncology
Martin Schuler
-
Contact Me
University Medical Center Essen, Department of Medical Oncology
University Medical Center Essen, Department of Medical Oncology
*Servet Bölükbas
-
Contact Me
University Medical Center Essen, Ruhrlandklinik, Department of Thoracic Surgery
University Medical Center Essen, Ruhrlandklinik, Department of Thoracic Surgery
Presenting Author:
Fabian Doerr
-
Contact Me
University Medical Center Essen - Ruhrlandklinik
University Medical Center Essen - Ruhrlandklinik
Abstract:
Objective:
Today, lung cancer screening programs for patients with a risk profile detect a large number of incidental lung lesions. The correct classification of these nodules regarding their malignancy is paramount. For incidental findings, the Brock model is commonly used today to calculate the probability of malignancy. In this study, we evaluated the predictive accuracy of the Brock model in comparison to the LIONS PREY model. Both models consider patient and nodule related parameters. In contrast to the Brock model, the LIONS PREY model takes into account prognosis-relevant risk factors such as the size progression of a nodule and tobacco consumption (see below).
Methods:
We retrospectively included all patients who underwent resection of a pulmonary lesion at our lung cancer center between August 2022 and August 2024. Patients whose clinical parameters were not fully available for calculating one of the two score systems were excluded from the study. Two study groups were formed (group M: malignant lesions; group B: benign lesions). We evaluated the predictive accuracy of the Brock model and LIONS PREY model based on calibration (observed/expected ratio), discrimination (area under ROC curve), and the total number of correctly classified patients (OCC).
Results:
During the treatment period, 776 patients were resected at our center. In this real-world setting, 424 patients (54.6%) were excluded due to unrecorded parameters. Finally, 304 patients in group M and 48 in group B were analyzed. With an OCC of 95.4%, the LIONS PREY model showed a significantly (p=0.0027) higher precision than the Brock model at 64.1%. The calibration of the LIONS PREY model (O/E ratio: 1.1) was significantly (p=0.031) superior to that of the Brock model (1.4). According to a DeLong analysis, the discriminatory power of the LIONS PREY model (AUC: 0.92; 95% CI: 0.89-0.95) was significantly (p=0.008) better than that of the Brock model (AUC: 0.62; 95% CI: 0.56-0.68).
Conclusions:
The success of a lung cancer screening program is directly linked to a reliable model to predict malignancy. Regarding predictive accuracy, the new LIONS PREY model appears superior to the established Brock model.
Today, lung cancer screening programs for patients with a risk profile detect a large number of incidental lung lesions. The correct classification of these nodules regarding their malignancy is paramount. For incidental findings, the Brock model is commonly used today to calculate the probability of malignancy. In this study, we evaluated the predictive accuracy of the Brock model in comparison to the LIONS PREY model. Both models consider patient and nodule related parameters. In contrast to the Brock model, the LIONS PREY model takes into account prognosis-relevant risk factors such as the size progression of a nodule and tobacco consumption (see below).
Methods:
We retrospectively included all patients who underwent resection of a pulmonary lesion at our lung cancer center between August 2022 and August 2024. Patients whose clinical parameters were not fully available for calculating one of the two score systems were excluded from the study. Two study groups were formed (group M: malignant lesions; group B: benign lesions). We evaluated the predictive accuracy of the Brock model and LIONS PREY model based on calibration (observed/expected ratio), discrimination (area under ROC curve), and the total number of correctly classified patients (OCC).
Results:
During the treatment period, 776 patients were resected at our center. In this real-world setting, 424 patients (54.6%) were excluded due to unrecorded parameters. Finally, 304 patients in group M and 48 in group B were analyzed. With an OCC of 95.4%, the LIONS PREY model showed a significantly (p=0.0027) higher precision than the Brock model at 64.1%. The calibration of the LIONS PREY model (O/E ratio: 1.1) was significantly (p=0.031) superior to that of the Brock model (1.4). According to a DeLong analysis, the discriminatory power of the LIONS PREY model (AUC: 0.92; 95% CI: 0.89-0.95) was significantly (p=0.008) better than that of the Brock model (AUC: 0.62; 95% CI: 0.56-0.68).
Conclusions:
The success of a lung cancer screening program is directly linked to a reliable model to predict malignancy. Regarding predictive accuracy, the new LIONS PREY model appears superior to the established Brock model.
THORACIC:
Lung Cancer
Image or Table
