Pooled data on 304,352 thyroid operations puts nerve injury between 1 and 10 percent. The lowest figure belongs to the approach with the most selected patients

The argument about thyroid cancer overdiagnosis rests entirely on a premise that is rarely quantified in the same breath: that removing a thyroid costs the patient something. If it did not, finding harmless cancers would be a waste of money and nothing worse.

Two meta-analyses published in 2026 put figures on the two complications that matter most — injury to the nerve that moves the vocal cord, and loss of the parathyroid glands that control calcium. Both sets of figures are more useful and more easily misused than they first appear.

Nerve injury, by operation#

The recurrent laryngeal nerve runs immediately behind the thyroid and supplies the vocal cord. Injuring it changes the voice; injuring both sides can obstruct the airway. A review of 199 studies covering 304,352 patients pooled how often it happens.

Permanent injury was rarer and worst after a second operation on the same neck: 2 percent (1 to 3) following secondary thyroidectomy. Reoperating through scar tissue, where the nerve is no longer where the anatomy says it should be, is the recognised hard case and the pooled data agrees.

Definitions are the likely culprit. A unit that routinely examines every patient's vocal cords after surgery finds transient palsies that a unit relying on hoarseness never records. Pooling both produces a number that is arithmetically correct and cannot be read as a patient's risk.

The approach comparison, and why Korea should read it carefully#

The same review reports transient injury by how the surgeon reached the gland: 5 percent (3 to 6) for transoral and transcervical approaches, and 1 percent (1 to 2) for transaxillary.

Korean surgeons perform more remote-access thyroid surgery than anyone, and transaxillary robotic thyroidectomy is a Korean speciality. That makes this the number most likely to be quoted here, and the one that most needs its context attached.

Nobody randomised these patients. Remote-access operations are offered to people who are younger, slimmer, and have small tumours without extensive nodal disease — precisely the operations that were going to be easiest whichever way the surgeon got there. A 1 percent rate in a selected population is not evidence that the approach protects the nerve. It is evidence about who was selected.

The parathyroid glands#

Four glands the size of grains of rice sit on the back of the thyroid and regulate blood calcium. Removing or devascularising them causes hypoparathyroidism, which in its permanent form means lifelong calcium and vitamin D replacement.

A meta-analysis of 64 studies examining 19 candidate risk factors found the picture depends on how the complication is defined — by serum calcium or by parathyroid hormone. Under the calcium definition, associations were found with female sex, central and lateral neck dissection, malignant pathology, incidental removal of parathyroid tissue, and total rather than partial thyroidectomy. Under the hormone definition, only malignant pathology reached significance.

Its conclusion identifies the highest-risk combination plainly: thyroid cancer, total thyroidectomy, lateral neck dissection. Every element of that is a consequence of how far treatment escalates once a cancer is found.

Why this connects to who gets screened#

Korean national data records postoperative complications of thyroid surgery as more common among higher-income patients, and falling fastest in that group as screening declined. There is nothing surprising in that once it is stated plainly: complications follow operations, operations follow diagnoses, and diagnoses followed who was being scanned.

It is the closing of the loop. The harm of overdiagnosis is not the diagnosis. It is the percentages above, multiplied by the number of people who were operated on for something that would never have troubled them.

What these figures do not tell an individual#

1. Every rate here is pooled across studies with different definitions of injury and different intensity of postoperative examination. The confidence intervals are wide for that reason and should be quoted with the estimate, not after it.
2. Comparisons between operations and between approaches are comparisons between differently selected patients. None of these studies randomised anyone to an approach.
3. Neither meta-analysis is Korean. They describe the operation internationally; the only Korean figures in this article are the income-linked complication trends from national data.
4. Transient and permanent injury are different outcomes with different consequences, and a transient rate of 8 percent does not mean 8 percent of patients are permanently affected.
5. Surgeon and centre volume are among the strongest determinants of these complications and are not resolvable from pooled published series.