Korea finds 65.3% of its stomach cancers while the disease is still localized. In the United States the figure is 32%. That single comparison explains most of what looks remarkable about Korean cancer statistics, and understanding it is more useful than any of the survival numbers on their own.
It also comes with a caveat that most articles quoting these figures leave out, so this one leads with it: a screening programme improves its own survival statistics whether or not it saves a single additional life. Both things are true at once. Here is how they fit together.
Stage at diagnosis is the whole story
Cancer survival is reported by how far the disease had spread when it was found. The categories are localized, regional and distant, and the difference between them dwarfs almost everything else.
| Stage at diagnosis | Korea, 2019–2023 | United States, 2016–2022 |
|---|---|---|
| Localized | 97.6% | 78.1% |
| Regional | 62.2% | 39.0% |
| Distant | 7.5% | 8.1% |
| All stages combined | 78.6% | 39.8% |
Look at the distant row. Korea 7.5%, the United States 8.1%. Once stomach cancer has spread, the two countries are indistinguishable, and Korea is marginally worse. Now look at the all-stages row, where Korea is nearly double. Almost the entire difference in the headline number comes from the mix of stages each country is diagnosing, not from what either does about the disease once found.
There is a fourth category the three-row version of this table hides, and it should be named. Cases of unknown stage are 5.1% of Korean diagnoses, with 47.1% survival. Quoting three stages implies they account for everyone. They do not.
Where the stage mix comes from
Korea has run a national gastric screening programme since 1999. Every adult aged 40 and over is offered either an upper-GI series or a gastroscopy every two years. In 2024, gastric screening had the highest participation of the six cancers the programme covers, at 77.4%.
The United States has no equivalent population screening programme for stomach cancer, which is why gastroscopy tends to appear there only after symptoms. Symptoms arrive late in this disease. That is the mechanism behind 32% against 65.3%, and it is not a story about the skill of either country’s surgeons.
Colorectal cancer shows the same pattern with a narrower gap, because the United States does screen for it.
| Stage at diagnosis | Korea, 2019–2023 | United States, 2016–2022 |
|---|---|---|
| Localized | 39.0% | 34% |
| Regional | 38.8% | 37% |
| Distant | 16.0% | 23% |
| Unknown | 6.2% | 6% |
Both countries screen, both find a similar share early, and the all-stages survival gap narrows to 75.6% against 65.4%. The clearest difference is at the far end, where Korea diagnoses 16.0% of colorectal cancers as already distant and the United States 23%.
The caveat that has to travel with these numbers
Survival is counted from the date of diagnosis. Move the diagnosis earlier and the measured survival gets longer, even if the date of death does not change at all. This is called lead-time bias, and it is not an argument against screening. It is a property of the statistic.
A second effect works the same direction. Screening preferentially catches slow-growing disease, simply because a slow cancer spends more years in a detectable state where a two-yearly test can find it. Fast, aggressive cancers are more likely to appear between screening rounds with symptoms. So the screened population’s average case is gentler than the unscreened population’s, which flatters survival again. That one is called length bias.
Put together, they mean this: a country that screens will report better survival than a country that does not, even in a hypothetical world where treatment is identical and no life is extended. Anyone using Korea’s survival figures to argue that Korean oncology outperforms American oncology is making a claim these numbers cannot support.
What the numbers can support is narrower and still worth a lot. Stage at diagnosis is not incidental. In Korea, stomach cancer found localized has 97.6% five-year relative survival and found distant has 7.5%. Whatever share of that gradient is measurement artefact, it is not all of it, and the thing that moves you from one end of it to the other is being looked at before you have symptoms.
What this means if you are deciding
Three things follow, and none of them is a promise.
A test’s value depends on whether it is finding disease you would otherwise have found late. For an adult with no gastric screening programme at home, a gastroscopy is examining an organ nobody has looked at. For colorectal cancer, where screening exists in most of our readers’ countries, the honest question is different and narrower: whether you are actually up to date, and whether a colonoscopy is more conclusive than the detection test you have been offered.
Population statistics are not personal ones. These figures describe tens of thousands of people over five years. They tell you that stage matters. They cannot tell you your own outcome, and this site will not use them as if they could.
Someone else’s number deserves the same scrutiny as ours. Every figure here names its cohort and its diagnosis window because those things change what a number means. Korea’s data is diagnosis years 2019–2023 and the American data is 2016–2022, so the side-by-side comparisons above are not like-for-like, and we would rather say that than quietly line up two columns and let them look equivalent. If you find these figures quoted somewhere without a window attached, that is worth noticing.
Checking the figures
The Korean numbers come from 국가암등록통계, the national cancer registry statistics, published through 국가암정보센터 and updated to the 2019–2023 cohort in January 2026. The American numbers come from the SEER programme’s own stat-facts pages, on the SEER 21 registry set excluding Illinois. Both are free to read.
One caution if you go looking. Registry statistics are revised on a yearly cycle and the previous vintage stays online, so it is easy to end up comparing a figure from one edition against a figure from the next. Korea’s stomach cancer mortality-to-incidence ratio, for instance, is 0.21 in the current edition and was 0.24 in the previous one. Both are correct for their own vintage, and pairing one with the other is a mistake that leaves no visible trace.