Rudolf Kaaks

Ph.D

German Cancer Research Center (DKFZ)

Professor

NLST (Learn more about this study)

NLST-1134

Oct 10, 2023

Development of lung cancer risk scores for determination of risk adapted screening intervals

While randomized trials have convincingly documented that CT-based lung cancer screening can reduce lung cancer-related mortality, questions remain regarding the optimization of screening schedules. So far, formal guidelines issued in various countries recommend annual screening. A major question, however, is whether all individuals who meet minimal criteria to be eligible for CT screening would indeed need to be screened annually, or whether screening can be equally effective sub-groups with comparatively lower lung cancer risks are screened less frequently (e.g. biennially). Previous analyses of the US National Lung Screening Trial, the Dutch-Belgian NELSON trial or the German Lung Screening Intervention Trial have shown that even among individuals who are eligible for LC screening (i.e., who all have a long-term smoking history) there is still wide variation in LC risk and the likelihood of having LC detected in one of future screening rounds. The risk of having a lung tumor detected on a next annual screening occasion may be predicted by risk models based on age and smoking history, plus CT-based risk indicators such as emphysema, consolidation or the presence and characteristics of a pulmonary nodules.
Aim of the present project is to use data from the NLST trial, as well as from screening trials in Italy (ITALUNG) and Germany (LUSI, and HANSE), to develop and validate risk models for a more personalized assignment of screening intervals. A successful “base” model for the prediction of lung cancer risk is PLCOm2012, which uses age, smoking history, and further sociodemographic and anamnestic medical data as risk predictors. This model has shown good risk discrimination and calibration in diverse population around the world, and is currently being tested in various countries (including the HANSE trial in Germany) as a tool for identifying individuals eligible for CT-based screening. Using the NLST data we plan to examine the utility of the PLCOm2012 model for further risk stratification, to improve the assignment of individuals to either an annual or a biennial screening regimen. Furthermore, we plan to augment the PLCOm2012 model with relevant CT-based traits, such as emphysema and consolidation scores and presence and characteristics of pulmonary nodules, to improve the prediction of risk for having lung cancer detected in a future screening round. Finally, we will use the data from independent screening trials in Italy (ITALUNG) and Germany (LUSI, HANSE) to externally validate this augmented PLCOm2012 + CT model. Augmented PLCOm2012 + CT model components will be developed and tested separately for screening participants who do not present any potentially “at-risk” pulmonary nodules on earlier screening tests (i.e., those falling into category 1 of the Lung-RADS classification) and for those who do present with pulmonary nodules potentially at risk (Lung-RADS categories 2 and higher).