In this large prospective study we did not find any association between thyroid function and the risk of THR or TKR due to osteoarthritis. Neither were changes in TSH over time, or overt hypo- or hyperthyroidism, associated with incidence of THR or TKR.
Few previous population studies have investigated the association of thyroid function with risk of osteoarthritis. In 1996, a cross-sectional study of 577 men and 798 women found no evidence of a significant association between current thyroid status and either chondrocalcinosis or osteoarthritis [
10]. However, that study only investigated prevalent osteoarthritis with a concurrent serum TSH concentration and could not take into account development in TSH or later treatment for abnormal thyroid function. Since our study could use data from both the second and third waves of the HUNT-survey, we were able to investigate the development of TSH over a median time of 11.2 years (SD 0.6). Change in TSH over time was however not associated with osteoarthritis development resulting in the need for joint replacement.
Previous or current thyroid disease at baseline was an exclusion criterion in our study. Nonetheless, all participants with TSH levels suggesting hypothyroid or hyperthyroid function may have received medical treatment for thyroid disease during the follow-up period, as participants with biochemical indication of pathological thyroid function were recommended to contact their general practitioners [
22]. This could have weakened any association between TSH and osteoarthritis. We therefore did a sub-analysis of persons that participated in both HUNT2 and HUNT3, and excluded participants that reported use of thyroid medication or thyroid disease in HUNT3. This did not significantly alter the results.
Our findings must be interpreted in relation to recent genetic studies on intracellular T3 availability in joint cartilage. There has been an increased interest in the effect of deiodinase polymorphisms on osteoarthritis [
5]. Iodothyronine deiodinases represent a family of proteins involved in local homeostasis of thyroxine (T4) and triiodothyronine (T3). Three deiodinases have been described and, of these, the deiodinase type 2 (D2) and deiodinase type 3 (D3) are detected in bone and cartilage. D2 plays a major role in conversion of T4 to biologically active T3 [
23] and thus upregulates local T3 levels. Deiodinase type 3 (D3) is the main T3-inactivating enzyme and consequently downregulates the local T3 levels. T3 is considered an important regulator of chondrocyte cell growth and differentiation in the endochondral growth plate [
24]. Local T3 availability, regulated by the opposite functions of D2 and D3, may be a determinant of osteoarthritis development. D2 has been reported to be upregulated in the cartilage of joints affected by osteoarthritis compared to joints unaffected by osteoarthritis [
25,
26]. However, it is not known if this is a result of the ongoing osteoarthritis process, or a reflection of the underlying disease pathway. Taken together, these findings suggest that deiodinase regulated local availability of T3 in chondrocytes is a possible factor in the pathophysiology of osteoarthritis [
27,
28]. Since our study did not find any association between circulating TSH, T3/T4 levels and osteoarthritis, it is conceivable that the serum thyroid hormone levels may be independent of local intracellular T3 levels in joints. Another possible explanation could be that polymorphism in the gene coding for D2 creates a predisposition for non-optimal bone shape [
29,
30], leading to increased risk of osteoarthritis independent of local thyroid hormone levels.
Strengths and limitations
Our study included over 37 000 persons without known thyroid disease at baseline, and in most cases thyroid function was measured many years prior to joint replacement. To the best of our knowledge, this is the first prospective population study addressing the association between thyroid function and joint replacement due to primary osteoarthritis. The prospective design and longitudinal data on TSH measurements are strengths of this study. By excluding participants with self-reported osteoarthritis at baseline it was possible to differentiate between risk of osteoarthritis development and progression.
Our study used joint replacement due to primary osteoarthritis as a surrogate measure of severe osteoarthritis. Validation of the osteoarthrosis diagnosis from the Norwegian Arthroplasty Register has not been done in an unselected population [
31]. However, the Danish hip Arthroplasty Registry has reported a positive predictive value of 85% regarding primary hip osteoarthritis diagnosis [
32], and it is likely that these results are comparable to the Norwegian Arthroplasty Register. The advantage of using joint replacement as a proxy for osteoarthritis is its unambiguous connection with disease burden of osteoarthritis compared to other osteoarthritis definitions, e.g., radiographic criteria, symptom criteria or osteoarthritis defined by self-reported diagnosis [
33]. However, this outcome measure is still limited in some respects, most importantly, that patients’ health status and potential comorbidities influence orthopaedic surgeons’ choices regarding operative treatment. Secondly, persons with moderate osteoarthritis who engage in demanding physical activities could be more motivated to have surgery than less active persons. Persons who are generally inactive may be less motivated to have surgery even if they have more severe osteoarthritis. This could give a healthy patient selection bias with corresponding underestimation of the effect of thyroid function.
Previous injuries increase the risk of knee osteoarthritis [
34,
35], but only joint replacements due to primary/idiopathic osteoarthritis were included in our study. We did not have direct information on previous injury, but we excluded all cases in which the operating surgeon reported that the knee joint replacement was due to sequela from fracture, ligament injury, meniscal injury, infection, rheumatoid arthritis or ankylosing spondylitis.
The interrelationship between BMI and thyroid function is complex and BMI could be treated as either a confounder or a mediator in our model [
36]. The reason we chose to define it as a confounder was that we wanted to investigate the direct effect of thyroid function on joint replacement, independent of BMI. Potential confounding by other unmeasured factors could not be excluded. But these factors should then be associated with both thyrotropin level and osteoarthritis. Therefore, we did not adjust for level of education in this study (
Appendix).
The participation rate in the HUNT-surveys was fairly high compared to most other surveys, but there is always a potential risk of selection bias that cannot be adjusted for in the statistical analysis [
37]. Blood samples were not drawn at a set time of the day, and it is known that other factors like exercise and sleep deprivation influence TSH levels [
38]. This might have led to non-differential classification bias, thus weakening any associations. We also only had data on fT4 and T3 in subpopulations. Therefore, the absolute numbers of participants with overt hypo- or hyperthyroidism were small, reducing the power to detect any association between them and euthyroid subject and should thus be interpreted with caution.
This study focused on the relationship between thyroid function and osteoarthritis. However, there may also be an association between autoimmune thyroid disease and osteoarthritis [
39]. As thyroid autoantibody not necessarily correlate with thyroid function, an association between thyroid function and osteoarthritis trough autoimmune factors could be missed in our study.
The exclusion of participants who reported hypo- or hyperthyroidism in their answers regarding use of treatment or medication might have caused misclassification since radioiodine is used in the treatment of cancer and T4 in the medical treatment of goitre. However, these treatments are infrequent, and it is unlikely that they substantially altered our results.
In the analysis comparing incidence rate of THR in people with and without self-reported thyroid disease we found a small, but significant, increased risk of THR in people reporting thyroid disease. And since we excluded participants with self-reported thyroid disease, this might have led to an underestimation of the effect on THR. We therefore did an additional analysis including those with self-reported thyroid disease: This showed no association between TSH levels and THR in a Cox-regression model (data not shown), and thus confirmed the findings from the primary analysis.