We prospectively enrolled patients who visited our clinic between October 2020 and November 2021 (https://clinicaltrials.gov/ct2/show/NCT04879173). This study was approved by the Institutional Review Board of the Seoul St. Mary’s Hospital (Seoul, Korea) (KC19DCSI0744), and written informed consent was received from all patients. The thyrotoxic status of the enrolled subjects was confirmed with biochemical laboratory data of increased free thyroxine (fT4) above the reference range and decreased thyrotropin (TSH) below the reference range. Patients who are taking ATD within 3 months before ultrasonography or who received any surgical procedure to their thyroid were excluded. The blood tests for each study subject included TSH, fT4, triiodothyronine (T3), anti-microsomal-antibody (thyroid peroxidase antibody [TPO-Ab]), thyroglobulin antibody (Tg-Ab), TSH-R-Ab, and thyroid stimulating immunoglobulin (TSI) bioassay, if clinically needed. The normal ranges were as follows: TSH (0.55 to 4.78 μIU/mL), fT4 (0.89 to 1.76 ng/mL), T3 (0.6 to 1.81 ng/mL), TPO-Ab (<60 U/mL), Tg-Ab (<60 U/mL), TSH-R-Ab (<1.75 IU/L), and TSI bioassay (<140%).

GD was confirmed by TSH-R-Abs. The clinical history and thyroid scan results were also considered in the diagnosis of GD. Autoimmune thyroiditis was diagnosed by positive TPO-Ab or Tg-Ab levels and the unique US findings of Hashimoto’s thyroiditis [3,14]. The diagnosis of subacute thyroiditis was made based on painful goiter, systemic inflammatory signs (erythrocyte sedimentation rate and C-reactive protein), and irregular hypoechoic tender lesions on US [15]. Drug-induced thyroiditis or postpartum thyroiditis was diagnosed based on the patient’s clinical history. All patients were followed up at least 3 months to observe the clinical course.

For the generally healthy group, we included volunteers who had no history of thyroid disease. They visited the hospital for a general health checkup and a euthyroid state was confirmed by normal TSH and free T4 levels. Blood pressure and heart rate were measured with an autonomic blood pressure meter after sufficient rest before the sonography exam.

All gray US, CD, PD, and MVUS were performed with an RS85 equipped with MV-Flow manufactured by Samsung Medison Co. Ltd. (Seoul, Korea). The volume of each lobe was measured on grayscale US as π/6×length×width×depth. Afterward, CD and PD images were obtained in the middle of the right and left thyroid in the transverse plane. We classified and scored those images into four groups according to the semi-quantitative (vascularity score) blood flow patterns suggested by a previous study [14,16-18]: pattern 0, blood flow limited to the peripheral thyroid arteries, while parenchymal flow is absent; pattern 1, the presence of mildly increased parenchymal flow; pattern 2, clearly increased blood flow with diffuse homogenous distribution; and pattern 3, markedly increased blood flow with a homogenous distribution, including the so-called “thyroid inferno” (Fig. 1). The vascularity score for CD and PD was obtained from the right and left lobe in all study subjects and final values for CD and PD indicated the mean values of them.

MVUS was performed from the middle of each thyroid lobe in the transverse plane. Considering differences due to the cardiac cycle, the systolic cardiac phase in which the vascular codes were the brightest was caught in all cine images [19,20]. The vascularity index (VI) represents the percentage of color pixels in the total grayscale pixels in a defined region of interest (ROI). To identify the ROI, we traced the thyroid structure manually, excluding the peri-thyroidal vascular structures [19]. The VI automatically displays the number of pixels, area, and ratio (%) measurements within the ROI (Fig. 2). The VI was obtained twice from each lobe in all study subjects and finally MVUS-VI indicated the mean values of right and left VI. For the measurement of intra-observer variations, a second VI measurement was performed after completely removing the probe for at least 5 seconds and was made at almost the same plane as the first measurement. For the measurement of inter-observer variations, two observers performed VI measurements on the same patients sequentially (Fig. 3). Parameters for CD were Frequency; general, Gain 50, pulse-repetition frequency 0.21 kHz, and Power; 90. Parameters for PD or MVUS were same except pulse-repetition frequency (0.56 kHz for PD, 0.74 kHz for MVUS).

SPSS version 24 (IBM Co., Armonk, NY, USA) was used for almost statistical analyses. Receiver operating characteristic (ROC) curve analysis was performed to compare the diagnostic performance of MVUS for differentiating GD from DT. Through the ROC curve, the area under the curve (AUC), and optimal cutoff values for CD, PD, and VI were calculated and compared. The optimal cutoff value was defined as the number where the sum of sensitivity and 1-specificity was maximized. To compare each ROC curve to other ROC curve, we did non-inferiority test based on web-site based on R package (plotROC, pROC, rocNIT, R Project for Statistical Computing, Vienna, Austria) [21]. A P value of <0.05 was considered statistically significant.

For intra and inter-observer variation analysis, we did two statistical methods; Interclass correlation coefficient (ICC) and Cohen’s kappa method. ICC was used to analyze the reliability of MVUS. Cohen’s kappa analysis was used to concordance of final diagnosis according to VI between intra- and inter-observer. The degree of correlation of two methods was expected based on previous study [22,23].

During the study period, 109 patients were screened and examined by the MVUS. Among them, eight were excluded because of inadequate medication status (n=5), previous isthmectomy surgery (n=1), and hyperfunctioning nodules (n=2) (Fig. 3). Therefore, MVUS was examined for 93 thyrotoxicosis patients and eight normal healthy volunteers. Finally, 46 GD (10 males, 36 females) and 47 DT (10 males, 37 females) patients were enrolled in this study. Among 47 DT patients, 18 with subacute thyroiditis, 22 with autoimmune thyroiditis, two with postpartum thyroiditis, and five with drug-induced thyroiditis were included.

There were no significant differences in the age at diagnosis, gender distribution and body mass index between the GD and DT groups. However, the fT4, T3 and TSH-R-Ab levels were significantly higher in the GD patients. T3/fT4 was not different statistically between two groups (Table 1).

The semi-quantitative CD, PD, and MVUS scores were all significantly higher in GD patients than in DT, indicating the hypervascular status of GD. The mean value of the right and left VI of the GD patients was higher than that of the DT patients (46.90%±22.65% vs. 19.46%±16.83%, P<0.001).

Eight normal controls were included in the healthy group and there were no statistical differences in age (P=0.322) or sex (P=0.584) between the three groups. The mean age was 40.75±14.08 years, and three males and five females were included. The normal controls showed lower thyroid vascularity compared to the GD and DT groups (right 11.79%±6.40%, left 8.09%±7.72%, and mean 9.94%±6.11%; P<0.001).

The AUC for the mean of right and left CD, PD, MVUS, and MVUS-VI was 0.822, 0.844, 0.808, and 0.852 respectively. VIs of MVUS showed greater AUC values than the semi-quantitative MVUS scores. MVUS showed non-inferiority to CD (P=0.03) and PD (P=0.004). MVUS-VI also showed non-inferiority to CD (P=0.05) and PD (P=0.02). With an optimal cutoff VI of 24.95% for distinguishing GD and DT, the sensitivity and specificity of MVUS-VI were 87.0% and 80.9% respectively. The VI showed the highest negative predictive value (86.4%) compared to other Doppler tools (Table 2, Supplemental Fig. S1).

Six GD patients were in the false negative range (low VI values with GD) and nine DT patients in the false positive range (high VI values without GD) according to the designated cutoff value (24.95%) obtained by AUC analysis (Fig. 4A). Among the patients who showed false-negative results, two were older than 68, and one received steroid treatment for a neurology problem. All patients who showed false-positive results were TPO or Tg antibody-positive. One patient received an immune checkpoint inhibitor for progressive cervical cancer.

We further investigated the difference in the MVUS-VI between Graves’ patients, subacute thyroiditis patients, and autoimmune thyroiditis patients. The mean VI value in GD patients was 46.90±22.65 and the MVUS-VI among autoimmune thyroiditis and subacute thyroiditis patients was 27.34±20.87 and 12.14±5.64, respectively, indicating that GD patients showed the highest MVUS-VI, followed by autoimmune thyroiditis and subacute thyroiditis in that order (Fig. 4B).

To determine whether the MVUS-VI was closely associated with other clinical parameters, a correlation analysis was performed only in GD patients. From the correlation coefficients, we found a positive correlation between the MVUS-VI and TSH-R-Ab (r=0.554) (Fig. 4C). Despite of small number (n=14), the MVUS-VI also showed the best positive correlation with TSI bioassay value (r=0.841) (Fig. 4D). The MVUS-VI also showed a positive correlation with fT4 and T3 levels. But it was rarely associated with the age at diagnosis, T3/fT4 or heart rate (Table 3).

Between the intra-observers or inter-observer, there were no significant differences in reliability as assessed by ICC (0.98 in intra-observer, 0.93 in inter-observer). The kappa value was 0.821 in the intra-observer and 0.883 in the inter-observer. It means the concordance of two operators would be almost perfect (Table 4) [23].