Stage- and histology-specific sensitivity for the detection of lung cancer of the NELSON screening protocol-A modeling study

The Dutch-Belgian lung cancer (LC) screening trial (Nederlands-Leuvens Longkanker Screenings Onderzoek [NELSON]) demonstrated low-dose computed tomography (CT) reduces LC mortality by 24% among men. The NELSON protocol differed from previous trials in the eligibility criteria, the use of volume-based nodule management, and increasing screening intervals. The early-stage sensitivity of the protocol is pivotal in determining the optimal screening strategy, such as the interval and age range. The MIcrosimulation SCreening ANalysis-Lung natural history model was used to reproduce LC incidence and mortality by detection method (clinical or screen-detected), sex, histology, and stage in the NELSON trial based on individual-level data. We evaluated screening effectiveness by stage and histology, accounting for population characteristics, trial design, and LC epidemiology. We find stage IA non-small cell LC (NSCLC) sensitivity of 24.6% (other NSCLC) to 41.0% (adenocarcinoma) at baseline screening. At repeat screening rounds, we find this increased to 70.9% for stage IA adenocarcinoma. For stage IB, the sensitivity by histology ranges from 26.4% to 77.1%; for stage II, 39.6%-81.9%. Upon detection, the probability of LC mortality prevention is estimated at 83% for stage IA. The sensitivity for detecting early-stage LC is found to depend on the histology of cancer and is increased for adenocarcinoma at repeat screenings. Despite a low rate of referral to follow-up screening in the NELSON trial, early-stage CT sensitivity and the probability of mortality prevention were similar to previous estimates from the National Lung cancer Screening Trial. Previously demonstrated screening effectiveness may be maintained when implementing new programs, while reducing unnecessary follow-up when considering NELSON evidence.