Lung cancer
The deadliest cancer in Europe is fought with evidence built around the patients it is least likely to kill.
The first Exclusion Debt Report looks at lung cancer across seven European markets: France, Germany, Ireland, the Netherlands, Portugal, Spain and the UK. It asks one question across the whole evidence chain, from clinical trial design to screening to the datasets that train AI: who is missing, and what does it cost?
This report in five numbers
From the report and the sources it cites: GLOBOCAN 2024 estimates for Europe; an Australian population study of older adults (2021); a cohort of 17.5 million people in England (2025); four Italian cancer centres (2022); a review of 61 Phase III clinical trials (2023). Where a figure comes from one country or analysis, the report says so.
What exclusion debt means
Exclusion debt is what accumulates when evidence about a disease is built without the people who have it.
The gap is created at specific moments: the design of a clinical trial, the setting of a screening threshold, the assembly of a dataset. It grows, because every decision built on the partial evidence carries it forward. And it can't be undone after the fact: once the evidence is built, the people who were missing can't be added back by analysis. That's why the report ends with a standard for how evidence is collected, not a plea for better statistics. More on exclusion debt
Three people, one disease
None of the three is a real person. Each is a composite built from the evidence in the report, and each stands for tens of thousands of people across the seven markets.
Brigid, 81
Lives on a farm in the west of Ireland, sixty miles from the nearest cancer centre, with COPD and heart failure. Her treatment was approved on a clinical trial whose participants were, on average, in their early sixties, and which left out people with heart failure.
Eva, 49
A primary school teacher in Rotterdam who has never smoked. Her cancer carried an EGFR mutation, the kind that responds well to targeted therapy, but it was found late: screening across Europe is designed for long-term smokers.
Rahim, 58
A bus driver in east London, of Bangladeshi heritage. Men of Bangladeshi heritage have the highest lung cancer rates of any group in England, and the clinical trials across his city record his age and tumour type, but not his ethnicity.
Seven ways the evidence leaves lung cancer patients out
| Who carries the burden | What the evidence shows | Why it matters |
|---|---|---|
| Older and multimorbid patients | About 53% of older metastatic patients excluded from typical clinical trials; median age 62 among those enrolled, against 70 overall | Treatments used on patients the evidence never covered |
| Women and never-smokers | Up to 25% of cases are never-smokers; EGFR drives about 75% of never-smoker adenocarcinoma; screening evidence built on male smokers, with the NELSON clinical trial 84% men | A rising, targetable group found too late |
| Deprived communities | Twice the incidence; 35% higher risk of aggressive adenocarcinoma | Highest burden, furthest from clinical trials and screening |
| Minority ethnic communities | Distinct incidence and subtype patterns; persistent under-enrolment; ethnicity often unrecorded | The gap can't be measured, so it can't be managed |
| Geography and access | 54% of those on standard treatment had no clinical trial nearby; about 15% enrol; no national screening yet in Ireland, the Netherlands or Portugal | Postcode decides inclusion in the evidence |
| Lived experience | Patient-reported outcomes secondary or exploratory in 89% of Phase III clinical trials; how missing data was handled stated in 45% | Decisions made without knowing what matters to patients |
| Datasets and AI | Cancer datasets heavily weighted to European ancestry | Bias trained into the next generation of tools |
The seven aren't separate problems. The older patient is also more likely to have other conditions, to live far from an academic centre and to live in a deprived area, so the patient left out on one count is usually left out on several. And the most consequential gaps are the least measured: nobody can audit a fairness that was never recorded.
The standard: a gap that can only be closed at the start
Who must be in it
The people who actually get lung cancer, deliberately and in proportion: older patients, including those 75 and over, and people with other conditions; women and never-smokers; people in the most deprived areas; minority ethnic communities, recorded at the detail the disease demands; people in rural areas and smaller markets; and people with language, literacy or digital-access barriers.
What must be captured
Age with performance status and frailty; sex-specific history; smoking status with never-smoker as its own category, plus exposure to air pollution and at work; ethnicity, self-identified; deprivation; rurality and travel time to the nearest site; other conditions and medicines; language, literacy and digital access; lived experience, with patient-reported outcomes planned in advance; and consent status.
How it must be collected
Designed in before the protocol is final, not patched in after. Built on a standing, consented community of people affected by the disease, maintained between clinical trials. Made from real people, not synthetic proxies. And auditable: every dataset should be able to answer, and publish, who is missing.
What each part of the system should do
- Clinical trial sponsors: benchmark every protocol's planned population against the real-world age, comorbidity, sex, deprivation and ethnic profile before finalising it, and justify each exclusion in writing.
- Screening programmes: revisit criteria that rest on smoking history alone, build pathways for never-smokers with high exposure and for women, and publish uptake by deprivation and ethnicity.
- HTA bodies and regulators: require sponsors to report how their clinical trial population compares with the real one, and treat unexplained gaps as a limitation of the evidence.
- Data custodians and AI developers: report the ancestry and make-up of training data, and don't deploy models in populations missing from it without validating them there first.
- Patient organisations and advocates: press for the standard as a condition of funding and approval, and replace the smoker's stigma with the never-smoker reality.
The limits of this analysis
The report is a synthesis of published evidence, not primary research. The inequality figures come from England, the enrolment figures from four Italian centres over six months, and the figures on who would get into a clinical trial from Australia; the research-funding comparisons are older. The severity ratings are the authors' judgement. The full report sets out every limit, and lists all 25 sources.
Start with the people the evidence leaves out.
300 people living with lung cancer have joined us, 39% of them in the most deprived fifth of areas. Ask them before your protocol is final, and refer the ones who say yes, with every referral followed through. Our lung cancer members