Korea's gastric cancer screening works — until about age 70, on new national data

Korea established its National Cancer Screening Program in 1999, initially offering stomach, breast and cervical cancer screening to Medical Aid Program beneficiaries, and by 2019 had expanded it to six cancers — stomach, liver, cervical, breast, colorectal and lung — for both Medical Aid recipients and health insurance beneficiaries [s2]. Over two decades the programme has been credited with reducing the incidence of stomach, liver, colorectal and cervical cancers and improving five-year relative survival [s2].

What has been much less clear is where the benefit stops. A nationwide cohort study published online on 20 July gives an answer with an unusual amount of statistical power behind it [s1].

What was done

The study linked the Korean National Health Insurance Service database to national mortality records [s1]. It included individuals aged 40 and over who underwent screening esophagogastroduodenoscopy between 2005 and 2010, propensity score-matched to non-screened controls [s1]. The primary outcome was gastric cancer-specific mortality; comorbidity was measured with the Charlson Comorbidity Index [s1].

The matched cohort ran to 555,904 individuals — 277,952 in each arm [s1].

The age gradient

Hazard ratios for gastric cancer-specific mortality, screened versus unscreened, moved steadily with age [s1]:

• Age 50–59: HR 0.56 (95% CI 0.34–0.92)
• Age 60–64: HR 0.65 (95% CI 0.40–1.07)
• Age 65–69: HR 0.82 (95% CI 0.58–1.17)
• Age 70–74: HR 0.99 (95% CI 0.75–1.30)
• Age 75–79: HR 0.92 (95% CI 0.68–1.24)
• Age 80 and over: HR 2.21 (95% CI 1.49–3.27)

Read the intervals, not just the point estimates. Only two bands have confidence intervals that exclude 1.0: the 50–59 group, where screening was associated with a 44% lower gastric cancer mortality, and the 80-and-over group, where it was associated with more than double the mortality [s1]. The four bands between them are consistent with benefit, with no effect, and — at their upper bounds — with modest harm.

The gradient itself is the finding. Point estimates rise monotonically from 0.56 through 0.99 across the 50–74 bands before the reversal in the oldest group [s1]. A clean dose-response with age is harder to produce by chance than any single estimate in the series.

Comorbidity does the same thing

Independently of age, comorbidity burden modified the association. Among people with a Charlson index below 3, the hazard ratio was 0.86 (95% CI 0.74–1.01); among those with an index of 3 or above, 1.88 (95% CI 1.42–2.49) [s1].

In combined age-and-comorbidity analyses, point estimates favoured screening across every age group up to 75–79 among people with a Charlson index of 0 — while hazard ratios exceeded 1.0 in most age groups among people with an index of 3 or above [s1].

So the two modifiers are not redundant. A healthy 78-year-old and a multimorbid 60-year-old land in different places.

Why an association above 1.0 is not a contradiction

A hazard ratio of 2.21 for a screening test does not mean the endoscope causes gastric cancer death. Two mechanisms are more plausible, and the study speaks to one of them directly.

The first is procedural harm. The authors report that post-endoscopy respiratory or cerebrocardiovascular events increased with age and comorbidity [s1]. Upper endoscopy in a frail 82-year-old carries risks — sedation, aspiration, cardiovascular stress — that it does not carry in a healthy 55-year-old. Those risks compound with the downstream investigations and treatments a positive finding triggers.

The second is confounding by indication, which propensity matching reduces but cannot eliminate. Older people who present for screening endoscopy may differ from those who do not in ways not captured by administrative data — including having symptoms that prompted the referral. In an oldest-old group, that kind of residual confounding can be large relative to the true effect.

The honest reading is that the two explanations are not separable here.

What it does and does not establish

It establishes that in a very large matched Korean cohort screened between 2005 and 2010, the association between screening endoscopy and lower gastric cancer mortality weakened with age and with comorbidity, and reversed direction beyond 80 or at high comorbidity burden [s1].

It does not establish a threshold. The authors write that benefit becomes questionable beyond age 70, and that at 80 and over, or with substantial comorbidity, potential harms may outweigh benefits [s1] — "questionable" and "may" are load-bearing. The 70–74 hazard ratio of 0.99 is a point estimate whose interval runs from 0.75 to 1.30 [s1]; that is not a demonstration of no benefit, it is a demonstration that this study cannot tell.

It is also an observational cohort, not a trial. Screening was chosen, not randomised.

Two further constraints on generalising. The screening episodes date from 2005 to 2010 [s1]; endoscopic technique, sedation practice and gastric cancer incidence have all moved since. And this is a benefit-to-harm balance calculated inside a country that runs national stomach cancer screening as standard [s2]. It does not transfer to settings where the disease is rarer, because the procedural harms travel and the detected cancers do not.

What to watch

Whether comparable analyses in other populations that screen endoscopically reproduce the age gradient — and whether guideline bodies move from a lower age bound to an upper one. The practical shape of the question this study raises is not whether to screen, but whether a national programme should stop inviting people at a defined age, or stop on the basis of comorbidity and life expectancy instead. The study's own age-and-comorbidity analysis points at the second.

This article is informational and does not constitute medical advice.

Sources

• [s1] Myeong S, Jung J, Bae KH, et al., "Age-dependent effects of gastric cancer screening endoscopy on mortality: a nationwide cohort study," Endoscopy, published online 20 July 2026. https://doi.org/10.1055/a-2891-4257
• [s2] Lee K, Suh M, Choi KS., "Current Status of the National Cancer Screening Program in Korea: History, Achievements, and Future Directions," Journal of Preventive Medicine and Public Health, published online 21 June 2025. https://doi.org/10.3961/jpmph.25.268