Risk prediction for lung cancer screening: a systematic review and meta-regression.

Clinical Snapshot

Key Findings

• P Value: Not reported in abstract

• Effect Size: AUC range: ~0.70 (moderate) to >0.90 (excellent) across 91 models; biomarker- and imaging-enhanced models consistently outperformed models without these features; 56 models for screening selection (30 incorporating biomarkers), 35 for post-screening nodule classification

• Primary Outcome: Discrimination performance (AUC) of risk prediction models for lung cancer screening selection and post-screening nodule classification, and meta-regression of factors associated with AUC

• Nnt Or Sensitivity: Sensitivity and specificity not reported as pooled estimates; AUC used as primary discrimination metric; fewer than half of models underwent external validation, limiting generalisable performance estimates

• Confidence Interval: Not reported in abstract; meta-regression coefficient confidence intervals not available from abstract data

Clinical Application

The review does not endorse any specific model for immediate clinical implementation. Feasibility of individual models varies considerably: regression-based models (e.g., requiring age, smoking pack-years, family history) are more readily implementable in primary care settings, while biomarker- and imaging-enhanced machine learning models require laboratory infrastructure and specialist input. The finding that fewer than half of models have been externally validated is a critical feasibility barrier to adoption. Australia does not currently have a nationally funded lung cancer screening programme, though the Medical Services Advisory Committee (MSAC) has reviewed evidence and the field is actively evolving. The RACGP and Lung Foundation Australia have published position statements supporting risk-based LDCT screening for high-risk individuals. This review's finding that most contemporary risk models lack external validation is directly relevant to Australian health technology assessment processes, where MSAC requires robust evidence of clinical utility and generalisability before public funding. The TGA does not regulate risk prediction models as therapeutic goods unless embedded in software as a medical device (SaMD), but ARTG listing considerations would apply to commercially deployed decision-support tools. PBS listing of biomarkers used in some models (e.g., specific protein panels) would require separate evaluation. Australian clinicians should note that models developed predominantly in North American or European cohorts may not be directly applicable to Australia's distinct demographic mix, including Indigenous Australians and populations with differing smoking prevalence patterns. Adults at elevated risk for lung cancer who are candidates for LDCT screening programmes, and individuals with screen-detected pulmonary nodules requiring risk stratification for further management. Most applicable to current or former heavy smokers aged 50–80 years meeting contemporary screening eligibility criteria.

Abstract

BACKGROUND: Lung cancer (LC) remains the deadliest cancer, often diagnosed at advanced stages. Screening reduces mortality in high-risk individuals. Eligibility criteria in European and US screening guidelines have recently expanded. Therefore, we conducted an updated systematic review of risk-based models for identifying candidates for low-dose computed tomography screening and post-screening nodule classification. METHODS: We systematically searched Embase and Medline (January 2020-January 2026), identifying studies proposing new risk models in the context of LC screening. We separated models by pre- and post-screening risk stratification. Data extraction included study design, population, model type, risk horizon and model performance metrics. We performed an exploratory meta-regression of areas under the curve (AUCs) to assess whether sample size, model type, validation type and inclusion of biomarkers were associated with performance. RESULTS: Of 2462 records, 91 were included. 56 models were for screening selection (30 included biomarkers) and 35 for post-screening nodule classification. Regression-based models predominated, though machine-learning approaches were increasingly common. Discrimination ranged from moderate (AUC∼0.70) to excellent (>0.90), with biomarker and imaging-enhanced models often outperforming models without. Calibration was inconsistently reported and fewer than half underwent external validation. CONCLUSION: We identified 91 risk prediction models for LC, developed after 2020. Although many demonstrated promising discrimination across both screening selection and post-screening management, most remain insufficiently mature for clinical adoption, as their performance and practical value outside the original study setting are uncertain. Future work should prioritise external validation, updating and comparative evaluation of existing models, and prospective implementation studies rather than continued development of additional models.

References

1. 1.Rezaeianzadeh, R., Leung, C., Kim, S. J., Choy, K., Johnson, K. M., Kirby, M., Lam, S., Smith, B. M., & Sadatsafavi, M. (2026). Risk prediction for lung cancer screening: a systematic review and meta-regression. European Respiratory Review, 35. https://doi.org/10.1183/16000617.0295-2025