This project aims to achieve the following objectives:
- Develop and refine polygenic risk scores for lung cancer prediction.
- Assess environmental risk scores based on air pollution exposure for lung cancer prediction.
- Estimate malignancy probability of lung nodules using an AI-derived risk score.
- Quantify chest CT biomarkers in NELSON screening rounds, including emphysema, coronary calcifications, bone density, vertebral height, and body composition.
- Develop and validate multisource data-driven prediction models to identify individuals at the highest risk of lung cancer.
- Design and validate multisource data prediction models for lung nodule management, aiming to reduce unnecessary follow-up scans and referrals.
- Assess the cost-effectiveness of the newly developed prediction models.
Novel data will be generated for the NELSON study, which have not yet been utilized for predicting lung cancer risk and survival. By integrating these new data with existing information, multisource prediction models will be developed for both pre-screening and post-baseline screening participant selection, as well as for nodule characterization. These models aim to facilitate personalized lung cancer screening strategies (Figure 2).
We hypothesize that 15-20% participants will be identified as having low lung cancer risk or short life expectancy. This can prevent about 140,000 Dutch individuals from unnecessarily being screened. Furthermore, the new models are expected to improve the specificity of nodule management by 10% without loss of sensitivity as compared to criteria based on nodule size and growth alone; we expect this will reduce unnecessary work-up by 40-50%. External validation will necessitate studies with the same datasources as included in the final multi-source models.