Why lung cancer staging alone can miss high-risk patients

A new single-cell atlas of non-small-cell lung cancer (NSCLC) shows that standard tumour staging alone can get a patient's real risk wrong, in both directions. Built from 2.5 million cells across 454 patients treated at three German hospitals, the atlas found that around a quarter of people staged as stage I were actually at high risk of poor outcomes, while a small group of stage IV patients had survival as good as early stage disease. This article looks at what the atlas found, how the combined risk model performed against staging alone, and what it could mean for how lung cancer risk is assessed in future.

Fluorescence microscopy image of lung cancer cells, illustrating the single-cell analysis behind the new atlas

Key points

  • A single-cell atlas of 2.5 million cells from 454 people with non-small-cell lung cancer combined tumour genetics with standard staging to build a more detailed risk model.

  • A quarter of stage I patients were secretly high risk: 20 of 79 people staged as UICC stage I were reclassified into high-risk groups, with a 5-year survival of 35%, against 68.6% for stage I overall.

  • Some stage IV patients did far better than their stage suggested: a subgroup of 5 of 20 stage IV patients reclassified to low risk had 80% 5-year survival, and none had the standard actionable mutations doctors currently test for.

  • The combined model outperformed staging alone in an independent validation group of 47 patients, with a C-index of 0.678 versus 0.576 for staging alone.

  • The full atlas is freely available to researchers at lucas-plus.com, which could support further studies into how cellular profiling complements staging.

What the atlas found

Researchers from LMU Munich, Charité Berlin and TU Berlin built LuCaS+, a single-nucleus atlas of non-small-cell lung cancer, published in Nature Cancer on 12 November 2026. It combines data from 454 patients treated at three German hospitals: a discovery group of 364 patients from Berlin and Cologne, and a separate validation group of 90 patients from Munich. Across more than 2.5 million cells, the team profiled 136 malignant tumour clusters and 60 non-malignant cell types, then used this detail to build a full model that combines standard UICC staging with the cellular information from each tumour.

Why does current staging miss some high-risk stage I patients?

Standard staging looks mainly at tumour size and spread, not at what is happening inside the tumour at a cellular level, so it can group together patients whose underlying biology is very different. Among patients staged as UICC stage I, whose overall 5-year survival was 68.6%, the full model reclassified 25.3% (20 of 79 patients) into the two highest-risk groups, where 5-year survival was only 35.0%. In other words, roughly 1 in 4 people with a stage I diagnosis, normally considered highly treatable, had cellular hallmarks closer to advanced disease.

Can some stage IV patients do far better than their stage suggests?

Yes, the data points the other way too. Among patients staged as UICC stage IV, the model reclassified 5 of 20 into the lowest-risk groups. These patients had an 80% 5-year survival rate, compared with just 8% for the 15 patients who remained in the highest-risk groups. None of these more favourable stage IV patients carried any of the standard actionable mutations lung cancer teams currently test for, including EGFR, ALK, MET, RET, ROS1 and NTRK1-3, meaning that routine genomic testing alone would not have identified them as having a better prognosis.

How was the combined model validated?

The research team tested the full model against staging alone in an independent group of 47 patients from a separate hospital. The full model performed significantly better at predicting survival, with a C-index of 0.678, than staging alone, with a C-index of 0.576 (P = 0.043), suggesting the findings may generalise beyond the original discovery group. A C-index measures how well a model ranks patient risk: 0.5 is no better than chance and 1.0 is perfect, so this is a modest but meaningful improvement.

What could this mean for lung cancer care?

At this stage, the atlas is a research tool rather than a clinical test, and the findings need to be confirmed in larger and more diverse patient groups before they could change practice. But the direction of the findings reinforces a point Lung Cancer Europe has long made through its Get Checked! campaign and its calls for broader biomarker testing: a stage label is not the whole picture, and more detailed molecular information can identify both higher-risk early-stage patients who might benefit from closer monitoring, and lower-risk advanced-stage patients who might otherwise be over-treated. The full LuCaS+ dataset is freely available to researchers at lucas-plus.com, which could support further studies into how cellular profiling complements staging across Europe.

Frequently asked questions

What is a single-cell atlas?

A single-cell atlas is a detailed map of the different cell types inside a tumour, built by analysing the genetic activity of individual cells rather than a whole tissue sample. This atlas profiled more than 2.5 million cells from 454 people with non-small-cell lung cancer.

Does this replace UICC or TNM staging?

No. The researchers combined cellular data with standard UICC staging to create a full model, rather than replacing staging. The atlas is also a research resource at this stage, not an approved clinical test.

Does this apply to small cell lung cancer?

No. The atlas covers non-small-cell lung cancer, which includes the LUAD and LUSC subtypes studied here. Small cell lung cancer has different biology and was not part of this research.

Where can researchers access the data?

The full LuCaS+ atlas is freely available to researchers at lucas-plus.com.

These findings add detail rather than certainty. The atlas is a research resource, and its risk model still needs validation in larger, more diverse patient groups before it could inform clinical decisions. But it adds to a growing body of evidence that a stage label alone does not capture a person's true risk, and that more detailed molecular and cellular information could help identify who needs closer monitoring and who is already being treated appropriately. Member organisations and researchers can explore the full LuCaS+ dataset at lucas-plus.com, and read the original paper in Nature Cancer for the complete methodology and results.

About Lung Cancer Europe

Lung Cancer Europe is a pan-European advocacy organisation working for equal access to high-quality lung cancer care across Europe. Our members are national lung cancer advocacy organisations, and we bring their evidence and experience to European policymakers. Through our annual Lung Cancer Europe Reports, our Charter and the Get Checked! campaign, we work to improve diagnosis, treatment and support for everyone affected by lung cancer.

References

1. Jurmeister P, Keyl P, et al. A single-cell lung cancer atlas reveals cellular and molecular drivers of metastasis and clinical outcome. Nature Cancer (2026). Article.

Next
Next

Broader biomarker testing linked to longer lung cancer survival, study finds