Use of the TREAT 2.0 Model to Create a Web-Based Lung Nodule Calculator for High-Risk Patients

Area of Science:

• Pulmonary Medicine
• Medical Informatics
• Oncology

Background:

• Indeterminate pulmonary nodules are frequently encountered in clinical practice.
• Cancer probability assessment is crucial for managing these nodules, influenced by factors like age, smoking history, nodule size, and imaging characteristics.
• The Thoracic Research Evaluation and Treatment (TREAT) 2.0 model was previously developed to predict cancer risk.

Purpose of the Study:

• To evaluate the accuracy of a web-based calculator utilizing the TREAT 2.0 model.
• To assess calculator performance specifically when common data patterns are missing in a clinical setting.

Main Methods:

• A web-based calculator was developed based on the TREAT 2.0 model.
• A parsimonious modeling approach identified the 8 most significant missing data patterns.
• Model accuracy and discrimination were assessed using area under the curve (AUC) and Brier scores, with bootstrap internal validation for overfit correction.

Main Results:

• The complete case analysis (all 13 variables) yielded an AUC of 0.91.
• The calculator demonstrated good performance with incomplete data, achieving a mean AUC of 0.79 for submodels with the 8 most common missing variables.
• Validation data sets showed AUCs of 0.83 and Brier scores of 0.11 and 0.14 for complete and incomplete data, respectively.

Conclusions:

• The TREAT 2.0 calculator offers reliable risk predictions for indeterminate pulmonary nodules.
• The calculator maintains accuracy even when essential variables are missing, reflecting real-world clinical scenarios.
• This tool supports clinical decision-making in high-cancer prevalence settings for patients with indeterminate pulmonary nodules.