The main conclusion
In a large electronic-health-record analysis of adults with type 2 diabetes, documented COVID-19 was associated with more delayed diagnoses of Hashimoto thyroiditis and Graves disease. The relative associations were substantial, but new diagnoses were uncommon. The study cannot establish that SARS-CoV-2 caused thyroid autoimmunity and does not support routine thyroid-antibody testing after every infection.
More than one million people, yet only hundreds of events
The investigators used the TriNetX Global Collaborative Network. Adults with type 2 diabetes and no previous thyroid disease were eligible. After propensity-score matching, 590,346 people with documented COVID-19 were compared with 590,346 controls without documented infection.
The primary follow-up began 12 months after the index date. This landmark design was intended to reduce reverse causation, early misclassification and the intense surveillance that may follow an acute infection.
Hashimoto thyroiditis was recorded in 649 people in the COVID cohort and 511 controls. The hazard ratio was 1.56 (95% CI 1.39–1.75). Graves disease was recorded in 467 and 299 people respectively, giving an HR of 1.92 (95% CI 1.66–2.23).
Why the hazard ratio is not the whole story
A hazard ratio compares the rate at which events occur over time. An HR of 1.56 does not mean that 56% of people developed Hashimoto disease, nor that an individual patient has a 56% probability of disease. In the authors’ descriptive competing-risk analysis, six-year cumulative incidence was approximately 0.21% versus 0.17% for Hashimoto disease and 0.15% versus 0.09% for Graves disease.
A very large dataset can detect a statistically convincing relative difference even when the absolute event rate is low. Both facts belong in a responsible explanation.
Strengths and limitations
The sample was large and multicentre. Previous thyroid disease was excluded. The investigators balanced many recorded baseline characteristics and tested the robustness of the findings in sensitivity and subgroup analyses. The 12-month landmark is especially useful because it shifts attention away from acute thyroid abnormalities and early diagnostic activity.
Nevertheless, this remains an observational EHR study. Diagnoses were based on clinical coding and were not centrally adjudicated with antibodies, hormone results and ultrasound. Undocumented infections may have occurred among controls. People with known COVID may have interacted with health services more often, increasing the opportunity for thyroid disease to be detected. Propensity matching cannot remove differences that were not measured or reliably recorded, including details of infection severity, vaccination, access to care and health-seeking behaviour.
The findings apply to adults with type 2 diabetes. The numerical estimates should not automatically be transferred to people without diabetes, children, pregnant women or patients with established thyroid disease.
Should everyone have thyroid antibodies tested after COVID
No. This study supports clinical vigilance, not universal screening. Testing is most useful when it answers a clinical question raised by symptoms, an abnormal TSH result or a specific risk profile. Persistent palpitations, tremor, heat or cold intolerance, unexplained weight change, fatigue, altered bowel habits or menstrual disturbance warrant clinical assessment, although none is specific to thyroid disease.
In many situations the first step is a clinical review and TSH measurement. Free thyroid hormones, antibodies and imaging should then be selected according to the findings. A positive antibody result without thyroid dysfunction does not automatically require treatment.
A defensible way to communicate the result
It is accurate to say that delayed autoimmune-thyroid diagnoses were recorded more often after documented COVID in adults with type 2 diabetes. It is not accurate to say that COVID has been proven to cause Hashimoto disease or that every survivor needs a full antibody panel. The practical message is targeted evaluation when symptoms or biochemical abnormalities justify it.
Source: Hung KC, Weng HL, Lai YC, Chen IW. Frontiers in Endocrinology. 2026. https://doi.org/10.3389/fendo.2026.1916915
This educational information does not replace individual medical assessment.