Prevalence and Incidence of Thyroid Dysfunction in Type 1 Diabetes, Type 2 Diabetes and Latent Autoimmune Diabetes of Adults

The Fremantle Diabetes Study Phase II

Kirsten E. Peters; Stephen A. Paul Chubb; David G. Bruce; Wendy A. Davis; Timothy M. E. Davis


Clin Endocrinol. 2020;92(4):373-382. 

In This Article


In this well-characterized, community-based sample of people with diabetes, the overall baseline prevalence of any thyroid disease was 17.4%. In the subgroup with known or biochemically detected overt thyroid disease at baseline, 8.7% had not been diagnosed prior to recruitment to FDS2. In addition, 7.4% of those who developed known or overt thyroid disease over the next four years had not been diagnosed through usual care in the community during the period of FDS2 follow-up. A raised baseline TSH was a poor predictor of future hypothyroidism. These data suggest that contemporary guidelines for thyroid function testing in people with diabetes of any type should address the significant proportion of those with potentially treatable disease who remain undiagnosed.

The prevalence of known/undiagnosed thyroid disease in the present study (17.4%) was similar to that reported in other large studies. In studies of clinic-based samples of type 1 and type 2 diabetes patients, prevalent thyroid disease was found in 13.4% and 12.4%, respectively.[2,26] Among participants with type 1 and type 2 diabetes in the population-based HUNT2 and HUNT3 studies, the overall prevalences of known and undiagnosed thyroid disease were 16.1% and 14.9%, respectively.[7] General population-based surveys have found a prevalence of known and undiagnosed thyroid disease of between 4% and 14.0%.[27–30] Our point estimate of 17.4% is similar to that (14.0%) in the only general population study with a similar mean age to that of our participants.[28] In one study in which participants with and without diabetes from the same population could be compared, the age-adjusted prevalence of thyroid dysfunction was higher in those with type 1 diabetes than in those without diabetes and those with type 2 diabetes, but there was no difference between the latter two groups.[7] In a smaller clinic-based study, hypothyroidism was more common among participants with type 2 diabetes than in controls without diabetes.[5] Our results are consistent with the former study in that the age- and sex-adjusted prevalence of known and undiagnosed thyroid disease was higher among those with type 1 vs type 2 diabetes, while the prevalence in our participants with type 2 diabetes was similar to that in Australians without diabetes of comparable age.[28] Nevertheless, as there is evidence that thyroid dysfunction in type 1 and type 2 diabetes may worsen cardiometabolic risk factors, thyroid disease may have greater significance in people with diabetes.[6,31]

We found that further adjustment of the prevalences of known and undiagnosed thyroid disease for anti-TPO positivity rendered associations with diabetes type nonsignificant. This confirms that autoimmune disease drives the association,[32] as also seen in the significant relationship we observed between anti-TPO and anti-GAD positivity in our participants with type 1 diabetes. We and others have previously observed this association in participants classified as having type 2 diabetes on clinical grounds,[11,33] but this analysis was not possible in the present study as those with clinically defined type 2 diabetes, but positive GAD antibodies were reclassified as LADA.

The prevalence of known thyroid dysfunction in our female participants with type 2 diabetes was 22.2%, over double that in the women with type 2 diabetes in our previous analysis from Phase I of the FDS (8.8%; P < .001).[11] Given that these estimates were based on similar recruitment and data collection methodology, we conclude that thyroid disease is being increasingly detected in the community. However, we were unable to determine the basis on which the diagnoses were made, raising the possibility that thyroid disease might now be over-diagnosed. Recent evidence of thyroxine over-use in the UK and USA[34] may parallel the Australian situation since guidelines similar to those in the UK and US are in place in Australia.[35]

In the participants with a raised baseline TSH, we found persistent elevation in over half (56.5%) of those not commenced on thyroid replacement therapy during follow-up. Persistent subclinical hypothyroidism in 36%-56% and reversion to euthyroidism in 35%-54% of participants have been observed in studies carried out over 2–4.3 years in cohorts from the general population.[12–14] Our results are consistent with these observations. In a small study of women with type 2 diabetes and subclinical hypothyroidism, persistent elevation of TSH was seen in 43%,[11] but there are no other comparable data in patients with diabetes. General population studies have found incident overt hypothyroidism and/or thyroid replacement therapy in 7.1%-26.2% of subjects with subclinical hypothyroidism at baseline, with a higher incidence in those with baseline TSH above 8–15 mU/L.[12–14] Our results are in accord with these observations, although we had too few cases to stratify by baseline TSH. Our observations confirm that subclinical hypothyroidism is a frequently transient phenomenon in type 2 diabetes and that, while large population-based studies show that such people are at increased risk of requiring thyroid replacement,[10,13,14] serial thyroid function and clinical assessments are needed before the diagnosis can be made. A single TSH is not a good predictor of subsequent overt hypothyroidism.

Of participants who were euthyroid and not taking thyroid medication at baseline, 3.7% developed a raised TSH during follow-up and the incidence of TSH abnormality or new thyroid medication use was 7.0% after 4.3 years. This suggests that single normal TSH results do not exclude the possibility of thyroid abnormality manifesting in the future and that ongoing surveillance is needed to detect new cases.

We found cases of untreated overt hypothyroidism both at baseline (1.1%) and during follow-up (0.2%). In addition, of those eight participants who were overtly hypothyroid at baseline and returned for the Year 4 assessment, only 25% had commenced thyroxine therapy. None of the remainder showed progressive deterioration of thyroid function, their baseline thyroid function was only marginally abnormal, and three became euthyroid, suggesting regression to the mean. Guidelines usually recommend screening for thyroid disease in all people with type 1 diabetes.[18] However, our unadjusted analysis did not find a significantly different prevalence or incidence of thyroid dysfunction on the basis of diabetes type. This may be because there is a tendency towards a higher incidence of thyroid disorders among older people.[36] Therefore, an average aged patient with type 2 diabetes is as likely to have thyroid dysfunction as a typically younger patient with type 1 diabetes. Our data suggest that usual community-based care is effective at detecting significant thyroid disease in people with type 1 or 2 diabetes. In contrast to the current guidelines, periodic testing of all type 2 patients may be necessary to detect all affected patients, as is being done on an ad hoc basis in countries such as the UK.[22]

Our participants with LADA were of similar age to those with type 2 diabetes but were older than those with type 1 diabetes. Their anti-TPO positivity was intermediate between type 1 and type 2 diabetes, consistent with observations of higher prevalence of anti-TPO positivity in type 2 diabetic subjects with anti-GAD positivity versus negativity.[33] Nevertheless, the prevalence of any thyroid disease (known/unknown) in the LADA participants was similar to those with type 2 diabetes, and lower than among those with type 1 diabetes. Although there were a limited number of participants with LADA, our data relating to diagnosed thyroid disease suggest that they behave more like people with type 2 rather than type 1diabetes.

The present study had limitations. Identification of thyroid disease depended on the documentation in the WADLS and FDS2 databases, respectively. However, there was 85% agreement in the classification of participants based on the hospital morbidity database and self-report suggesting that these sources were acceptably accurate (data not shown). We were unable to ascertain pretreatment thyroid function in most participants with incident thyroid dysfunction requiring replacement therapy as these were done outside the study, and so cannot confirm whether the participants were being treated for overt or subclinical thyroid disease. We did not measure free T3 so are unable to comment on the prevalence of T3 thyrotoxicosis. The strengths of this study include its large community-based cohort, detailed prospective data collection and WADLS linkage.

Thyroid dysfunction, whether diagnosed or detected on biochemical screening, is common in diabetes regardless of type. Subclinical hypothyroidism is the commonest form (at around 1 in 25 people with diabetes both at baseline and during follow-up in FDS2), and it has a variable course. This latter observation alone makes an argument for periodic biochemical screening in all people with diabetes (not just type 1) as part of routine management.