Subacute thyroiditis is a non-bacterial inflammatory disease of the thyroid gland. It comprises nearly 3%-6% of all thyroid lesions[5].The diagnosis is made by a combination of clinical manifestations, physical examinations and laboratory tests[6].Tissue diagnosis is not a routine,but only necessary in rare cases such as in differential diagnosis of thyroid cancer[7].Usually,the typical SAT has three stages:hyperthyroidism stage,hypothyroidism stage and normal thyroid function stage. However, patients may come to hospital at any stage, and the clinical manifestations may not show in a typical way, which makes the diagnosis more difficult. Furthermore,a study showed most laboratory results associated with thyrotoxicosis have reached abnormal levels within 3 weeks after onset.But longer time-lags could existed between the onset of clinical symptoms and the appearance of abnormal laboratory findings in patients with SAT[9].Therefore,hyperthyroidism and suppressed uptake of Technetium-99 m (99 m-Tc) or 131I at the same time play a significant role in diagnosing. However, the radioisotope scanning is not always available in every hospital and sometimes it is contraindicated in specific situations such as in pregnant or lactating women. So we developed and internally validated a diagnostic model for SAT without the need for radioisotope scanning.
Our model contained a limited number of signs and symptoms and 2 laboratory tests,which were easy to perform in primary care institutions.Thyroid tenderness will appear almost 100% due to the inflammatory destruction of thyrocytes.In our diagnostic model,the regression coefficient of this variable was 5.271 with a clinical score of 5,demonstrating the importance of thyroid tenderness in diagnosing SAT.Follicular epithelial cells and multinuclear giant cells against a dirty background is the pathological characteristics of SAT,resulting in feeling firm on thyroid palpation[10].The variable of “firm on palpation” obtained a clinical score of 2 according to the regression coefficient in our model. Different degrees of inflammatory cell infiltration can lead to different levels of firmness, and in some cases,the gland felt no firm on palpation.This partly explained the cause of the relatively low score. ESR is a sensitive indicator of acute inflammation and always elevated in SAT patients[4, 6].But ESR is not a special feature to some certain diseases,so it earned only 3 scores in our model.In the first stage of SAT,all patients are hyperthyroidism,which is defined by elevated FT3 and/or FT4 concentrations and suppressed TSH level in our study, whether they have symptoms or not.However,patients may be at other stages when they come to hospital.Their thyroid function may be normal or even decreased.So the variable of hyperthyroidism was assigned 3 scores in our model. The P value of Omnibus tests of our model was ≤ 0.001 and the Nagelkerke R Square was 0.894,demonstrating that our model has statistically significant differences and better goodness of fit.According to the ROC curve analysis,the optimal cut-off point was 5.5.At this point, our model has high sensitivity(93.33%) and specificity(97.53%) in our patients. The accuracy of our diagnostic model was also high(94.64%).To our knowledge, this diagnostic study is the first one to develop a clinical prediction model for the diagnosis of SAT.
A lot of studies have shown that the etiology of SAT was related to viral infection such as coxsackievirus, echovirus, mumps, measles, influenza and other viruses[11, 12] because there was a flu-like syndrome before the disease onset.In our study,only 14.4% patients had upper respiratory infection before SAT,maybe due to the blurry memory of patients and the inapparent infections.Espinoza et al[13] have compared the diagnostic value of radioactive iodine uptake, 99 m-Tc thyroid static imaging and thyroid ultrasonography.They found both radioisotope scanning had a better correlation with the clinical diagnosis of SAT than that with thyroid ultrasonography. In our study,we used the 99 m-Tc thyroid static imaging to evaluate the thyroid uptake function instead of radioactive iodine uptake because radioactive iodine uptake needs more time(24hours) and is more complicated to operate.As reported by Frates et al[14] ,the typical appearance of sonography of SAT was a patchy, poorly defined hypoechoic process that could affect a portion of one or both lobes, an entire lobe, or the entire gland.However,they also found that the sonographic findings of SAT could mimic a large nodule replacing the lobe,the changes of lymphocytic,Hashimoto thyroiditis,thyroid carcinoma or thyroid lymphoma, leading to the differential diagnosis became more difficult.Furthermore,sonography is a relatively subjective examination and requires experienced doctors to get analysable results.So we excluded sonography as a variable from our model. Fever is also a clinical presentation of SAT.Sometimes fever was the only clinical manifestation as reported by Dalugama[15].But this situation is rare and fever can present in many diseases.So fever, though there was significant difference between the two groups,was not included in our regression model. It is important to differentiate SAT from Graves’ disease(GD) because they have similar clinical features and thyroid hormone concentrations but have different treatments and prognosis.Some studies have demonstrated that a higher ratio of FT3 to FT4 supported a diagnosis of GD and a very low ratio supported a diagnosis of SAT[16, 17].However in our study,the ratio of FT3 to FT4 showed no significant difference between the two groups,mainly because GD only occupied a small portion in our patients.The sample size was not large enough to make statistical difference.
Our study has some limitations. First, it was a retrospective study. We could only collected data from the hospital system other than examining the patients by ourselves,which may cause the inconsistency.Second,we have excluded patients who did not perform radioisotope scanning or in pregnant or lactating women,which might lead to a selection bias.Third, our patients were all at their first attack and first visit to hospital.We were not sure if our diagnostic model was suitable for recurrent patients or treated patients. The last but not the least, the specificity and sensitivity of our diagnostic model were really high because we only did internal validation and obtained overoptimistic results.Therefore,external validation is needed before wider application.