Smoking causes most lung cancer and quitting helps at any age — neither fact depends on a genotype. The interesting genetics is one locus in the nicotine receptor cluster, and an argument from 2008 about whether it acts through smoking or through the lung that has never been fully settled.
Solid lines are connections this site curates. Dashed lines mean the two ends share a research paper — worth knowing, and not a claim that one explains the other.
Start with the part that is not in doubt. Smoking causes the large majority of lung cancer; the risk falls after quitting and keeps falling for years; and nothing on this page changes either statement or should be read as softening it.
rs8034191 sits at chromosome 15q25, inside a cluster of genes for nicotinic acetylcholine receptor subunits — CHRNA5, CHRNA3 and CHRNB4. If that sounds familiar it is because the same cluster carries rs1051730, the strongest common variant for how much a person smokes, which this site covers on its smoking behaviour page.
The lung cancer association is not small by GWAS standards. It was found in about 2,000 cases and 2,600 controls, replicated across five further studies, and reached a combined p-value of 5 x 10-20. The authors estimated the locus accounts for 14% of lung cancer cases at the population level — an attributable risk, not a personal one.
Two papers reported this region in Nature in the same month of 2008, and they framed it differently.
The first observed that the risk looked statistically similar whether or not people smoked, and pointed out that these receptor subunits are expressed in lung tissue itself — alveolar epithelial cells, pulmonary neuroendocrine cells, lung cancer cell lines — and that they bind compounds found in tobacco smoke. On that reading the variant does something in the lung, not only in the brain.
The second found the same cluster affecting cigarettes smoked per day and nicotine dependence, and conferring risk of lung cancer and peripheral arterial disease. Its authors described the finding as a case study of gene-environment interaction: the variant changes how much a person smokes, and smoking does the damage.
Both papers are real, both were large, and the disagreement is about mechanism rather than about whether the association exists. Nearly two decades on, the honest summary is that some of the effect runs through smoking quantity and it is still argued how much is left over. A page that picked a side here would be telling you more than anybody knows.
A 2017 analysis of 29,266 cases and 56,450 controls found 18 susceptibility loci, ten of them new — and the most useful thing in it is that the loci are not the same for different kinds of lung cancer. Four were associated with lung cancer overall and six specifically with adenocarcinoma, the subtype that occurs most often in people who never smoked. Other loci sit near genes that maintain the ends of chromosomes.
So "lung cancer genetics" is really several different genetics, depending on which lung cancer.
It does not estimate anyone's risk, and it is not used in any screening decision. Where lung cancer screening exists it is offered on age and smoking history — how much, for how long, how long ago — because those are the things that actually predict it and can be answered without a laboratory.
The useful actions are the ordinary ones: not starting, stopping if you have started, and taking a persistent cough, coughing blood, breathlessness or unexplained weight loss to a doctor rather than to a genome. Our smoking page covers the one place where genetics has something practical to say — which quitting medicine suits whom — and even there the finding is about a blood measurement rather than about this variant.
The 15q25 locus. Hung et al. genotyped 317,139 SNPs in 1,989 lung cancer cases and 2,625 controls across six central European countries, identifying 15q25 at P = 9 x 10-10, with replication in five further studies (2,513 cases, 4,752 controls) giving P = 5 x 10-20 overall and an estimated attributable risk of 14%. The association region contains CHRNA5, CHRNA3 and CHRNB4; the CHRNA5 non-synonymous variant D398N (rs16969968), which substitutes an amino acid at a conserved site in the second intracellular loop, is among the strongest markers, and rs8034191 is in the same haplotype block. Reported risks were statistically similar irrespective of smoking status or propensity to smoke.
Mechanism, unresolved. Thorgeirsson et al., publishing on the same cluster in the same month, reported effects on cigarettes per day and nicotine dependence together with risk of lung cancer and peripheral arterial disease, and interpreted the locus as acting through nicotine addiction. The competing interpretation rests on expression of these subunits in alveolar epithelial cells, pulmonary neuroendocrine cells and lung cancer cell lines, and on their binding of N'-nitrosonornicotine and related tobacco carcinogens — a plausible route to a direct pulmonary effect. Mediation analyses since have generally supported a substantial smoking-mediated component without excluding a residual direct effect; self-reported cigarettes per day is a coarse exposure measure, which limits how cleanly the two can be separated.
Current architecture. The 2017 OncoArray-based analysis (29,266 cases, 56,450 controls) identified 18 genome-wide significant loci, ten novel, with pronounced heterogeneity by histology: four associated with lung cancer overall and six with adenocarcinoma specifically. eQTL analysis in 1,425 normal lung tissue samples highlighted RNASET2, SECISBP2L and NRG1; other implicated loci include the cholinergic receptor CHRNA2 and telomere-maintenance genes. Effect sizes for individual common variants remain modest and no polygenic score is in clinical use for lung cancer.
Clinical position. Lung cancer screening with low-dose CT is offered on the basis of age and smoking history in the guidelines that recommend it; genotype has no role in eligibility, diagnosis or treatment selection, which is driven by histology and by somatic tumour testing (EGFR, ALK, ROS1, KRAS and others) — a different kind of genetics entirely, performed on the tumour rather than inherited.
What a 23andMe/AncestryDNA export or raw VCF can and can't tell you about Lung Cancer comes down to these specific, well-studied positions — not a diagnosis.
No. It is one common position associated with risk across large populations. The authors estimated the locus accounts for about 14% of cases at population level — that is an attributable fraction across everyone, not a probability for one person, and it cannot be turned into one.
That is the open question. The 2008 study reporting this variant found risks statistically similar irrespective of smoking status, and these receptor subunits are expressed in lung tissue and bind tobacco carcinogens. A second paper the same month found the cluster acting through how much people smoke. Both are large and real, and how much of the effect is smoking-mediated is still argued.
Two possible reasons, which is exactly the argument. It changes smoking behaviour, and smoking causes lung cancer. And the receptors themselves are present in lung cells and bind compounds in tobacco smoke, which would be a direct route. The evidence supports the first strongly and has never ruled out the second.
No. Screening, where it is offered, is decided on age and smoking history — the things that actually predict lung cancer and that need no laboratory. No guideline uses this or any other common variant for eligibility.
No, and the difference matters. Tumour testing looks for changes acquired by the cancer itself — EGFR, ALK, ROS1 and others — and those results choose treatments. This page is about inherited variation, which does not.
Free to reuse. This page's text is original writing from freely-available research, licensed CC BY 4.0 — reuse it, including commercially, with attribution to MyGeneLog. It's general research-derived information, not medical advice or a diagnosis — see Terms of Use.