Thyroid cancer in South Korea: what the evidence says about the epidemic, the correction, and what it cost

Between 1999 and 2012 South Korea multiplied its thyroid cancer diagnoses tenfold. It is the largest documented cancer detection epidemic anywhere, it has been taught for a decade as the definitive case of overdiagnosis, and the country is now further into the aftermath than any other.

This article collects what the published evidence establishes, links to our reporting on each part, and is explicit about where the questions a patient actually has run past what anyone has measured.

The curve, and the curve that did not follow it#

Deaths did not move with any of it. Across 2005 to 2018 the annual count of thyroid cancer deaths stayed between 330 and 394 while annual diagnoses ranged from 12,833 to 44,798, and the age-standardised mortality rate fell from 1.28 to 0.80 per 100,000.

That divergence is the whole argument. A nationwide review of how the tumours were found showed 94.4 percent of the 1999–2008 increase was tumours under 20 mm detected mainly by screening, and 97.1 percent of the total increase was localised or regional disease — the signature of looking rather than of illness.

Who was screened, and who died#

Korea never had a national thyroid screening programme. The scans were opportunistic, and the strongest single predictor identified in a national survey of women was having attended for a breast ultrasound — 2.71 times the odds of thyroid screening against no breast screening, where mammography alone raised them only 1.47-fold.

The result was distributed by income. Higher income went with more screening, more diagnoses and more postoperative complications; lower income went with higher mortality. For lung cancer and stroke, analysed as controls by the same team, higher income predicted lower incidence and lower mortality — the ordinary direction. And when Korea began to correct course, incidence and complications fell fastest among the higher-income group that had been over-examined first.

What the operation costs#

Overdiagnosis matters only because treatment does something to people. Pooled across 199 studies and 304,352 patients, transient recurrent laryngeal nerve injury ran from 2 percent after near-total thyroidectomy to 10 percent after completion thyroidectomy, with permanent injury worst after reoperation. Highest risk of hypoparathyroidism belongs to the combination that follows a cancer diagnosis: total thyroidectomy with lateral neck dissection.

Those pooled figures are also a lesson in measurement. A Korean series of 1,000 consecutive transaxillary robotic operations that examined every patient's vocal cords found transient nerve injury in 4.1 percent — roughly four times what pooled international data assigns to that approach, because a transient palsy is counted only where somebody looks for it.

The alternative to operating#

Of 8,969 Korean patients with papillary thyroid microcarcinoma treated between 1994 and 2017, 107 died and two died of the thyroid cancer. In June 2025 the Korean Thyroid Association published the country's first guideline permitting active surveillance for such tumours: ultrasound and thyroid function every six months for two years then annually, with surgery if the tumour progresses.

In a Korean cohort of 927 patients, what predicted choosing surveillance over immediate lobectomy included prior awareness that it existed and higher income. Pooled data records progression in 14.53 percent and no thyroid cancer deaths in either the surveillance or the immediate surgery arm — alongside higher complication and recurrence rates among those who did come to surgery later.

How the numbers get misreported#

In 2024 a study of 434,228 Korean patients reported thyroid cancer-specific mortality rising from 0.76 to 2.70 per 1,000 person-years among those diagnosed between 2013 and 2018 — deaths climbing while diagnoses fell. The Korea Central Cancer Registry replied in the same journal that the figure counts deaths among diagnosed patients and is therefore a case fatality rate, not mortality, and that population mortality fell over the same span.

Where the evidence runs out#

1. Nobody knows what is driving the renewed rise in incidence since 2015. It appears in distant-stage disease as well as localised, which is not what overdetection produces, and the authors who documented it decline to explain it.
2. Registry data records diagnoses, not disease, and cannot separate more cancer from more looking. Stage at diagnosis is itself partly a function of imaging intensity.
3. Nothing in this evidence explains why lower-income Koreans die of thyroid cancer more often. Later presentation, comorbidity and treatment access are all candidates and none has been tested.
4. No randomised comparison exists between active surveillance and immediate surgery, or between remote-access and conventional thyroidectomy. Every comparison cited here is between patients who chose or were chosen.
5. The complication meta-analyses are not Korean, and the Korean surgical figures come from single high-volume institutions rather than from national data.