Updated systematic review of lung cancer risk prediction models finds high discrimination but limited external validation

Review of 91 models (2020-2026) for lung cancer screening selection or nodule classification showed AUCs ranging from ~0.70 to >0.90, but fewer than half were externally validated and calibration was inconsistently reported, highlighting that most models are not yet ready for cl…

• Systematic review and meta-regression of 91 studies proposing 56 pre-screening and 35 post-screening risk models.
• Biomarker and imaging-enhanced models often outperformed others, but calibration was rarely reported and external validation was scarce.
• Main limitation: most models lack external validation, and their performance outside derivation settings remains uncertain.

RadPigeon summaries are original and for information only. They are not clinical advice.