This study included 99 patients diagnosed with Graves' disease at a tertiary hospital (Dankook University Hospital, Cheonan, Korea) between 2013 and 2015. All patients were newly diagnosed patients without previous medication for thyroid disease and received methimazole treatments. Graves' disease was diagnosed by thyrotoxic symptoms (palpitation, tremor, and weight loss), increased free thyroxine (T4)/decreased serum TSH levels and elevated thyrotropin binding inhibiting immunoglobulin (TBII) titers, or the presence of the appropriate clinical features such as diffuse goiter or thyroid-associated orbitopathy. Goiter was classified into three grades (none; small: less than 40 g; and medium to large: 40 g or more). We excluded patients without reliable information or initial TFT concentrations and TBII titers, fewer than three instances of laboratory testing, without sufficient medication compliance, and previous history of medication affecting thyroid function such as steroids or iodine-containing drugs. This study was approved by Institutional Review Board of Dankook University Hospital (number 2016-09-012).

The anti-thyroid drug therapy and follow-up examinations were performed using a uniform protocol using a dose titrating regimen, mostly with methimazole, according to the present guidelines throughout the study period. Some patients were shifted to propylthiouracil therapy owing to side effects associated with methimazole; these patients were excluded from this study. The starting dose of methimazole was approximately 20 to 30 mg and was gradually decreased according to the levels of free T4. When the serum levels of free T4 in patients were within the normal range, the methimazole dose was decreased. When the TSH level was within the normal range, the dose of methimazole was maintained at 5 to 10 mg. There was no formal strategy for determining the initial dose, but the doses tended to be lower in elderly patients or those with relatively less severe thyrotoxicosis.

Patients were followed up every 1 to 2 months till their thyroid function had stabilized with normal TFT levels, followed by regular visits at intervals of 3 to 6 months while the thyroid status remained stable. TBII titers were measured sequentially during every visit. The decision to stop the medication depended on normalized thyroid function and absence of TBII. After the treatment was completed, patients were followed up every 3 to 6 months to monitor for disease recurrence to 12 months.

The serum TSH concentrations (reference range, 0.25 to 4 µIU/mL; lower detection limit, 0.03 µIU/mL) were measured using an immunoradiometric assay (RIA-gnost rhTSH, Cisbio Bioassays, Codolet, France). The serum levels of free T4 (normal range, 10.0 to 25.0 pmol/L) were measured using the free T4 radioimmunoassay (RIA) kit (fT4-CTX, DiaSorin SpA, Saluggia, Italy) The total serum T3 levels (reference range, 1.2 to 3.1 nmol/L) were measured using an RIA kit (T3-CTK, DiaSorin SpA). TBII titers were determined using the BRAHMS. TRAK human RIA kit (BRAHMS GmbH, Oberhavel, Germany) according to the manufacturer's directions. TBII titers ≥1.5 IU/L were considered positive.

This study defined responsiveness to methimazole as decreasing rates of free T4 level per one tablet of methimazole (5 mg) and a month. Firstly, in each patient, we calculated changes of free T4 level per a day for every interval between visits at clinic. Supposed that each patient shows linear correlation between changes in free T4 level and prescribed methimazole doses, we considered these correlation coefficients as individual responsiveness to methimazole. Fig. 1 shows representative two cases with rapid response and slow response, respectively. By iterative linear regression, the methimazole responsiveness of all patients was obtained.

The patients were classified into two groups according to their responsiveness to methimazole: the rapid responder group included patients with scores in the upper 50th percentiles, while the slow responder group included those with scores in the lesser 50th percentiles. Because the distribution of TBII titer skewed to the right, we used log-transformed values of TBII titer in analyses. Continuous variables were represented by mean±standard deviation and analyzed by Student t test or Mann-Whitney U test, and analyses of dichotomous variables were performed by chi-square test or logistic regression. In logistic regression analysis for risk factors of responsiveness, all factors were entered at the same time (the enter method), because of a small number of predictor. To evaluate the optimal cut-off levels for prediction of responsiveness, receiver operating characteristic (ROC) analysis was performed. Iterative calculation of the correlation coefficients in the linear regression analyses and statistical analyses were performed using R version 3.3.2 (R Foundation for Statistical Computing, Vienna, Austria). All P values were two-tailed with P<0.05 considered significant.

The mean age of the patients was 44.0±13.5 years old and 40 (40%) were male. Among the patients, 69 patients (70.1%) showed goiter and 27 patients (27.3%) had current smoker. At diagnosis, the mean free T4 level was 48.9±21.9 pmol/L, and the mean TBII titer was 36.6±74.4 IU/L. The mean value of the initial methimazole dose was 24.5±6.0 mg (Table 1). Fig. 2 shows the serial changes in free T4 level in each patient from the time of initial treatment. Almost all patients showed normalization of the free T4 levels within about 2 months, regardless of initial free T4 level.

Calculated using linear regression model for each patient, the median value of responsiveness, decreasing rate of free T4 level per one tablet of methimazole during 1 month, was 4.33, which means that, if a patient takes one tablet of methimazole (5 mg) during 1 month, his free T4 level will decrease by 4.33 pmol/L 1 month later. Fig. 3 shows the distribution of calculated responsiveness, and the mean of responsiveness was 9.31±6.20 in patients with the upper 50th percentiles (rapid responder group) and 2.73±0.99 in patients with the lesser 50th percentiles (slow responder group).

Rapid responder group was not different from slow responder group in age, male proportion, and median duration for treatment (Table 1). Median duration of follow-up was longer slow responder group than rapid responder group. There were no differences of proportion of goiter, smoking rate, and initial dose of methimazole between groups. However, free T4 and TBII titer at initial diagnosis were higher in rapid responder group than slow responder group with statistical significance.

In univariate analyses, high level of initial free T4 and TBII titer were significant predicting factor for rapid responsiveness to methimazole; however, age, male sex, presence of goiter, and smoking did not show significances (Table 2). Multivariate analysis with adjustment for all factors showed also significances of initial free T4 level and TBII titer for prediction of rapid responsiveness. On the other hand, presence of medium to large goiter was a predictive factor for slow responsiveness.

To evaluate of optimal cut-off value for prediction of responsiveness, we performed ROC analyses of TBII titer and initial free T4 level. The optimal cut-off value of TBII for prediction of responsiveness was 17.95 IU/L with sensitivity of 60.0% and specificity of 77.6% in ROC analysis (Fig. 4A). The optimal cut-off value of initial free T4 level was 47.3 pmol/L with sensitivity of 70.0% and specificity of 79.6% (Fig. 4B).