Mortality and Generalizability of the National Lung Screening Trial
Menée aux Etats-Unis à partir de données portant sur 7 105 personnes incluses dans le "National Lung Screening Trial" et sur 732 vétérans éligibles au dépistage du cancer du poumon (âge : de 65 à 70 ans), cette étude examine la possibilité, pour des patients bénéficiant de soins de routine, de participer au dépistage du cancer du poumon
Introduction : Lung cancer screening (LCS) reduced lung cancer death by 20% in the National Lung Screening Trial (NLST).1 NLST excluded those with life-limiting comorbidities. However, patients in routine care may be less healthy.2 Guidelines recommend screening those in good health, as only those who live beyond an averted cancer death benefit from screening.3,4 Direct comparisons are lacking between NLST participants and screening-eligible or screened patients in routine care. The Veterans Health Administration (VA) is a national pioneer in LCS and can address these knowledge gaps.
We compared all-cause mortality of NLST participants with those of similar age and tobacco use in a national VA cohort.6 We hypothesized trial participants would have lower mortality than screening-eligible or screened patients in routine care and that health metrics could identify patients of similar health to trial participants.
Methods : Full are in the eMethods in Supplement 1. Ethical approval for this study was provided by the University of California, San Francisco human research protection program. Reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines for cohort studies. We restricted 53 454 NLST participants to those randomized to control. We restricted 4503 VA cohort members6 to those meeting NLST eligibility (
≥
30 cigarette pack-years, current use or quit <15 years prior). Care Assessment Needs (CAN) score, a VA metric predicting mortality, and self-rated health and ability to climb stairs4 were captured at baseline. We restricted to ages 65 to 74 years, the range of overlap. Cohort enrollment occurred during 2020 to 2023 with follow-up through July 2025. Participants in the veteran cohort provided verbal informed consent. Participants in the NLST cohort provided written informed consent.
We compared 5-year all-cause mortality using Kaplan-Meier plots and Cox proportional hazards models between NLST controls and (1) cohort members, (2) those who received LCS, (3) cohort members stratified by CAN score, and (4) cohort members stratified by self-rated health and/or ability to climb stairs. Cohort analyses were weighted to reflect the national sampling design. Sensitivity analyses were restricted to cohort members who did not use cannabis (as the parent study examines cannabis’ health effects6) and who enrolled after January 1, 2022, after the early pandemic. P values were calculated using 2-sided log-rank tests, with P < .05 considered statistically significant. Analyses were completed in R version 4.5.2 (R Project for Statistical Computing).
Results : Cohort members (732 individuals; 415 [57%] aged 65-70 years) were more likely male (642 patients [88%]) and less likely White (103 [14%] Black non-Hispanic; 33 [4.5%] not Black, not White, and non-Hispanic; and 569 [78%] White non-Hispanic) than NLST controls but had similar tobacco use (Table). Cohort screening prevalence was 25% (95% CI, 22%-28%) and did not vary significantly by favorable or unfavorable CAN score (<60: 25%; 95% CI, 19%-30% vs
≥
60: 25%; 95% CI, 21%-29%) or favorable or unfavorable self-rated health and stair-climbing ability (26%; 95% CI, 22%-29% vs 18%; 95% CI, 10%-25%).
JAMA Network Open , article en libre accès, 2026