Overdiagnosis in low-dose CT lung cancer screening: A systematic review and meta-analysis of overall magnitude and subgroup variations
A partir d'une revue systématique de la littérature publiée jusqu'en octobre 2024 (26 études), cette méta-analyse évalue le surdiagnostic associé au dépistage du cancer du poumon par tomographie numérique à faible dose de rayonnement
Overdiagnosis is a major concern in low-dose computed tomography (LDCT) lung cancer screening as it can detect indolent or slow-growing cancers. This study aims to quantify overdiagnosis in LDCT screening and explore its variation by several factors. We conducted a systematic review and meta-analysis following PRISMA guidelines. Four databases (PubMed, Web of Science, Scopus, Cochrane Library) were searched up to October 2024 for studies reporting overdiagnosis in LDCT lung cancer screening. Overdiagnosis was quantified using three metrics: percentage in screen-detected cancers, percentage in all cancers in the screening group, and rate per 1000 screened individuals. Pooled estimates were calculated using random-effects models, and subgroup analyses were performed by follow-up time, histology, nodule type, geography, population risk, and gender. Twenty-six studies were included (8 RCTs, 11 cohorts, 7 ecologic studies). Overdiagnosis declined substantially with extended follow-up, with rates decreasing from 37% (95%CI: 14%–60%) to 7% (95%CI:
−
7% to 22%; p = .03) for screen-detected cancers and from 25% (95%CI: 8%–41%) to 2% (95%CI: 4%–8%; p = .01) for all screening-group cancers when follow-up exceeded 5 years. This pattern corresponded to overdiagnosis dropping from 12.26 (95%CI: 3.87–20.64) to 1.46 (95%CI:
−
2.11 to 5.03) per 1000 screened individuals (p = .02). Considerable variations emerged across subgroups: bronchioloalveolar carcinoma 82%, adenocarcinoma 28%, non-adenocarcinoma
−
11%, p < .001; non-solid cancers 66%, part-solid cancers 31%, solid cancers 7%, p = .002; Asian countries 38%, Western countries 22%, p = .009; general population 38%, high-risk population 22%, p = .010. These findings demonstrate that accurate overdiagnosis assessment requires sufficient follow-up duration and must account for substantial variability across clinical and demographic factors.
International Journal of Cancer , résumé, 2025