2022 World Conference on Lung Cancer (ePosters)-EP01.05-010. Using a Pattern Submodel Approach to Predict Lung Cancer in High-Risk Lung Nodule Clinics

Pdf Summary

The study presents the TREAT 2.0 model, a novel prediction model for lung cancer in patients seeking care in high-risk clinics. The model incorporates 12 demographic, clinical, and radiographic factors, as well as the clinical setting in which the patient is seen. The researchers used a novel pattern submodel (PS) approach to handle missing data in the prediction model.<br /><br />Data from 1,401 patients with indeterminate pulmonary nodules (IPNs) were collected from six sites, including patients from different clinical settings and lung cancer prevalence. The PS approach created separate models for each missing pattern, using only the non-missing variables. The outputted probability from this approach is referred to as the predicted risk from the TREAT 2.0 model.<br /><br />The researchers compared the performance of the TREAT 2.0 model to the Mayo and Herder models. The TREAT 2.0 model was found to be more accurate and better calibrated for predicting lung cancer in high-risk IPNs, despite missing data. The study also showed that the TREAT 2.0 model outperformed the TREAT 1.0 model in terms of discrimination and calibration.<br /><br />The study found that the TREAT 2.0 model improved the calibration of predicted risk across different patterns of missingness. However, some missing patterns showed similar or worse performance compared to the TREAT 1.0 model.<br /><br />The researchers concluded that the TREAT 2.0 model provides an improved lung cancer risk prediction for patients with IPNs in high-risk settings. The model is well-calibrated and performs well despite missing data. The study suggests that using models like TREAT 2.0, which consider lung cancer prevalence and evaluation methods, can improve generalizability and increase the use of nodule calculators in clinical practice.

Asset Subtitle

Valerie F. Welty

Meta Tag

Speaker Valerie F. Welty

Topic Early Detection and Screening - Pulmonary Nodule

Keywords

TREAT 2.0 model

lung cancer prediction

high-risk clinics

demographic factors

clinical factors

radiographic factors

clinical setting

missing data

indeterminate pulmonary nodules

prediction models