Highlights
– In the HANSE prospective cohort (n=5191), PLCOm2012 (6‑year risk ≥1.58%) produced a higher positive predictive value for lung cancer detection than NELSON categorical criteria (2.59% vs 2.17%; p=0.0016).
– The number needed to screen to detect one lung cancer was lower using PLCOm2012 (38.6) than with NELSON criteria (46.1), indicating greater efficiency for risk-model selection.
– Study strengths include a large, prospective, real-world screening cohort with predefined model cutoff; limitations include limited ethnic diversity, short follow-up between rounds, and outcomes focused on detection rather than mortality.
Background and clinical context
Low-dose computed tomography (LDCT) screening reduces lung cancer mortality when applied to appropriately selected high-risk individuals, as demonstrated by landmark trials such as the National Lung Screening Trial (NLST) and the NELSON trial. Trial and guideline implementation strategies have varied: some use simple categorical eligibility rules based on age and cumulative smoking exposure, while others have advocated risk prediction models that integrate multiple individual-level variables to estimate absolute lung-cancer risk. PLCOm2012 is among the most widely evaluated risk models and has been proposed as a more efficient way to select screening candidates than categorical approaches.
Policy-makers and screening program planners remain cautious about shifting from simple categorical rules (easy to communicate and operationalize) to risk-model based selection because of concerns about complexity, equity, model calibration across populations, and real-world effectiveness. The HANSE (Germany) prospective cohort study directly compares a predefined PLCOm2012 threshold with categorical NELSON criteria for LDCT participant selection to address this implementation question.
Study design and methods
HANSE is an ongoing prospective cohort study conducted at three certified German lung cancer centres in Großhansdorf, Hannover, and Lübeck. Between July 23, 2021 and Aug 19, 2022, investigators enrolled 5,191 current or former smokers aged 55–79 years who met either the NELSON categorical risk criteria or had a PLCOm2012 6‑year predicted risk ≥1.58% (the cutoff was predefined to yield a similar group size as the NELSON criteria).
Participants underwent baseline LDCT and a 1‑year follow-up LDCT, with all downstream diagnostic procedures carried out per routine clinical pathways. The prespecified primary outcome was the comparison of positive predictive values (PPVs) for lung cancers detected in the PLCOm2012-selected group versus the NELSON-selected group during the interval between the two scans. The analysis reported here represents the final primary-analysis results.
Key findings
Of 5,191 enrolled participants (2,208 female [43.5%], 2,983 male [57.5%], 5,076 [97.8%] European White), 4,167 met the PLCOm2012‑based selection and 3,916 met the NELSON criteria (some participants met both). Over a median interval of 1.05 years (IQR 0.95–1.08) between scans and with a mean volume CT dose index of 1.15 mGy (SD 0.15), 111 lung cancers were detected.
The PPV (i.e., lung-cancer detection rate) in the PLCOm2012-selected group was 108/4,167 (2.59%; 95% CI 2.13–3.12) versus 85/3,916 (2.17%; 95% CI 1.74–2.68) in the NELSON-selected group (p = 0.0016). This translated to a lower number needed to screen (NNS) to detect one lung cancer for PLCOm2012: 38.6 (95% CI 32.1–46.9) versus 46.1 (95% CI 37.3–57.5) for NELSON criteria.
In plain terms, applying the PLCOm2012 threshold identified more cancers per screened person than the NELSON categorical rule, with a statistically significant and clinically relevant improvement in efficiency.
Interpretation and clinical implications
HANSE provides prospective, real-world evidence that risk-model selection using PLCOm2012 (≥1.58% 6‑year risk) is more efficient than NELSON categorical criteria for detecting lung cancer in a German screening population. The improved PPV and lower NNS suggest that a risk-based approach can prioritize screening for individuals with the highest absolute risk, potentially improving yield and reducing resource use per case detected.
These results support incorporating validated risk models into programmatic LDCT screening pathways. Practical benefits may include fewer scans per cancer detected and potentially reduced downstream diagnostic burden and costs associated with low-yield screening. For health systems constrained by capacity or seeking to maximize benefit per screening resource, risk-based selection appears attractive.
Strengths of the study
– Prospective cohort design with prespecified primary endpoint and predefined PLCOm2012 cutoff.
– Large sample size recruited from certified clinical centres and real-world application of screening and diagnostic follow-up pathways.
– Direct head-to-head comparison of a widely used risk model against a categorical trial-derived selection rule in the same enrolled population.
Limitations and considerations
– Non-randomized design for allocation to selection strategy: participants met one or both criteria and were not randomized to a single selection method. This limits causal inference about downstream outcomes such as mortality.
– Short interval of observation reported here (between baseline and first annual scan) focuses on detection metrics rather than long-term outcomes (lung-cancer-specific or all-cause mortality, overdiagnosis rates, or long-term harms).
– Limited ethnic diversity (97.8% European White) constrains generalisability to other racial and ethnic groups where model calibration and performance may differ. Model recalibration may be required for broader populations.
– Although PLCOm2012 increased efficiency, the absolute difference in PPV (0.42 percentage points) is modest; programmatic impact will depend on prevalence of disease in target populations and operational costs.
– Implementation hurdles remain: collecting model inputs, EHR integration, training, legal and privacy concerns, and ensuring equitable access where risk models might deprioritize some disadvantaged groups.
Safety and harms
HANSE reports LDCT with a mean volume CT dose index of 1.15 mGy, consistent with low-dose protocols. The study focuses on detection performance and did not report mortality or comprehensive harm metrics (radiation cumulative exposure, false positives, unnecessary invasive procedures, psychological impact, or overdiagnosis) in this primary analysis. Any shift to risk-based selection should be accompanied by monitoring of these outcomes within screening programs.
Policy, practice, and research implications
HANSE strengthens the evidence base for adopting validated risk-prediction models (like PLCOm2012) as eligibility criteria in organized lung-cancer screening programmes. Policymakers should weigh this efficiency gain against practical implementation and equity issues. Key actions include:
- Formal evaluation of model calibration across diverse populations and recalibration where necessary.
- Comparative cost-effectiveness analyses to quantify potential resource savings and health benefits at the population level.
- Operational research to integrate risk calculators into primary-care workflows and screening invitations, including digital tools and data privacy safeguards.
- Prospective randomized or population-based implementation trials with longer follow-up to assess mortality benefit, harms (overdiagnosis, false positives), and equity outcomes.
Expert commentary
HANSE’s findings are consistent with accumulating evidence that risk-based selection can improve the yield of LDCT screening relative to simple categorical rules. However, translating detection efficiency into reduced mortality and net population benefit requires longer-term data and thoughtful program design. Guideline panels (e.g., USPSTF, European bodies) will likely consider these data alongside cost-effectiveness and health equity analyses when updating recommendations.
Conclusion
In the HANSE prospective cohort, PLCOm2012 with a 6‑year risk threshold of ≥1.58% selected a screening population with a higher lung‑cancer detection rate and lower number needed to screen than the NELSON categorical criteria. These findings support the implementation of validated risk-prediction models to improve screening efficiency, while underscoring the need for attention to calibration, equity, operational feasibility, and long-term outcomes including mortality and harms.
Funding and trial registration
Funding: Federal Ministry of Education and Research (German Center for Lung Research) and AstraZeneca.
ClinicalTrials.gov registration: NCT04913155.
Selected references
1. Vogel-Claussen J, Bollmann BA, May K, et al.; HANSE investigators. Effectiveness of NELSON versus PLCOm2012 lung cancer screening eligibility criteria in Germany (HANSE): a prospective cohort study. Lancet Oncol. 2025 Nov 10:S1470-2045(25)00490-5. doi:10.1016/S1470-2045(25)00490-5.
2. National Lung Screening Trial Research Team. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med. 2011;365:395–409.
3. de Koning HJ, van der Aalst CM, de Jong PA, et al. Reduced lung-cancer mortality with volume CT screening in a randomized trial. N Engl J Med. 2020;382:503–513.
4. U.S. Preventive Services Task Force. Screening for lung cancer: US Preventive Services Task Force recommendation statement. JAMA. 2021;325(10):962–970.
Author note
This article summarizes and critically appraises the primary-analysis results of the HANSE prospective cohort comparing PLCOm2012 and NELSON selection methods for LDCT screening. It is intended for clinicians, program planners, and policymakers engaged in lung cancer screening implementation and guideline development.