Adenocarcinoma of Lung; Bronchial Neoplasms; Early Detection of Cancer; Machine Learning
Conditions
Brief summary
This observational, cross-sectional study in lung cancer patients and lung cancer-free controls aims to develop a machine learning model for early detection of LC based on routine, widely accessible and minimally invasive clinical investigations. The model with adequate predictive performance could later be used in clinical practice as an aid in defining the optimal population and timing for lung cancer screening program.
Interventions
OTHERObservational
No interventions.
Sponsors
Jozef Stefan Institute
The University Clinic of Pulmonary and Allergic Diseases Golnik
Study design
Observational model
CASE_CONTROL
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
50 Years to 79 Years
Healthy volunteers
Yes
Inclusion criteria
All patients: * Age ≥ 50 years and \< 80 years at the index date of diagnosis (for Cases) or pseudodiagnosis (for Controls). * Presence of at least one extended blood analysis, spirometry and DLCO report within the 6 months before the index date. * Chest CT scan performed in a non-urgent setting (electively) within the 6 months before the index date (= index CT). * Active smokers at the index date or former smokers that ceased smoking within 15 years before the index date. * Smoking history ≥ 20 pack-years. Additional for Cases only: Confirmed histological diagnosis of bronchogenic lung cancer in the time period ≥ 2010 and ≤ 2020. Additional for Controls only: * Absence of lung cancer at all times ≤ 2020, confirmed by chest CT scan at the index date. * Documented to live without diagnosis of lung cancer for at least 3 years after the index date. Extended criteria for the lung cancer prediction subgroup: In addition to the above stated inclusion criteria, patients in this subgroup have at least one extended blood analysis, spirometry and DLCO report available in the time interval between 3-5 years before the index date.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Develop a model with high predictive performance for early detection of non-small cell lung cancer (NSCLC) in the eligible patient population. | 11 years | The primary outcome is tested by calculating a joint rectangular 95% confidence region for {sensitivity, specificity} and compared with the reported accuracy of NLST study screening criteria. |
Secondary
| Measure | Time frame |
|---|---|
| Demonstrate that the newly developed model achieves higher prediction accuracy than the well-validated model PLCOm2012. | 11 years |
Other
| Measure | Time frame |
|---|---|
| Develop a model with high predictive performance for early detection of small cell lung cancer (SCLC) in the eligible patient population. | 11 years |
| Develop a model for prediction of lung cancer in a time period when the disease is still highly unlikely to be clinically detectable, in a subset of patients who meet the extended eligibility criteria. | 11 years |
| Identify features with the highest discriminatory power of lung cancer prediction and early detection. | 11 years |
| Identify features with the highest discriminatory power to distinguish between lung cancer patients in stage I-II and stage III-IV. | 11 years |
Countries
Slovenia
Outcome results
None listed