Diagnostic Testing for Graves' or Non-Graves' Hyperthyroidism

A Comparison of Two Thyrotropin Receptor Antibody Immunoassays With Thyroid Scintigraphy and Ultrasonography

Lorenzo Scappaticcio; Pierpaolo Trimboli; Franco Keller; Mauro Imperiali; Arnoldo Piccardo; Luca Giovanella


Clin Endocrinol. 2020;92(2):169-178. 

In This Article


The diagnosis of Graves' hyperthyroidism based on clinical manifestations is unreliable,[17] and the choice of the initial tests is crucial in this setting.[18] Currently, diagnostic strategies widely vary in different centres due to different resources, expertise, preferences, availability and costs.[2,18,19]

Therefore, in order to optimize the diagnostic workup of thyrotoxicosis/hyperthyroidism, our paper was conceived to compare the performance of four tests (TRAb Kryptor® and Immulite® TSI assays, scintigraphy and ultrasonography) in thyrotoxic scenario.

In our cohort of patients with newly diagnosed and untreated thyrotoxicosis, the optimal thresholds to discriminate Graves' and non-Graves' hyperthyroidism were 0.7 IU/L and 0.1 IU/L for TRAb and TSI assays, respectively. The TRAb Kryptor® cut-off of 0.7 IU/L was not directly comparable to that obtained in other two studies,[20,21] since other immunoassays[22] and different selection criteria were used. The Immulite® TSI limit of 0.1 IU/L corresponded to the assay's functional sensitivity, and it was lower than that reported by Tozzoli et al[13] and Allelein et al,[23] in which patients with nonthyroid autoimmune diseases and/or Hashimoto's thyroiditis were included. Perharps, this was the reason for higher TSI thresholds (0.54 IU/L[13] and 0.55 IU/L[23]) found in the latter two studies, since it is well known that TSHR antibodies can be found in patients with nonthyroid autoimmune diseases[24,25] and in patients with Hashimoto's thyroiditis.[26] Both TRAb and TSI assays were proved to be highly accurate (accuracy >91%), with a discordance rate of 1.2% in patients with GD and 2.6% in patients with non-Graves' hyperthyroidism. In particular, five out of 86 patients with GD (5.8%) were both TRAb and TSI negative, all were characterized by mild thyrotoxicosis and absence of thyroid-associated orbitopathy; on the other hand, four old women primarily classified as having AFTN were both TRAb and TSI-positive (4/10, 40%), and they had subclinical hyperthyroidism. Similar findings were reported by Vos et al,[27] who demonstrated that the prevalence of TRAb-negative Graves' hyperthyroidism was 5.4%, and in the study by Pedersen et al[28] where the high occurrence of TRAb in multinodular toxic goitre (MTG) might in part reflect an overlap between Graves' disease and MTG in some patients. Moreover, there was a full concordance for positivity between the two assays, and the correlation among measurements mainly existed for relatively low TSHR antibody values. Indeed, in only two cases, our two IMAs recorded discordant results: one case corresponded to a patient with diagnosis of GD showing TRAb-negative result (0.3 IU/L) along with positivity for TSI (0.6 IU/L); the other patient was diagnosed as having painless thyroiditis, and she showed TRAb negativity along with TSI-positive result (0.7 IU/L) (one out of 12 cases with painless thyroiditis, 8.3%). Seropositivity for TRAb/TSI was documented by Morita et al[29] in ~15% of patients with painless thyroiditis, and it was emphasized how the activity of TSHR-antibody could modulate the thyroid function during the course of the disease.

We found that the optimal TcTU cut-off to best discriminate GD from non-Graves' hyperthyroidism was 1.3%, and it was associated to the highest diagnostic accuracy (95.6%), compared with the two IMAs and US. Four patients with GD (4 out of 86, 4.6%) had thyroid uptake <1.3%, and conversely, one patient with transient thyrotoxicosis showed TcTU values of 2.1%. We suppose that the four cases of GD with normal TcTU may belong to the same group of patients described by Ikekubo et al,[30] as they had normal thyroid volume, subclinical hyperthyroidism, TRAb and TSI <4 IU/l. Similar results regarding the optimal cut-off value of TcTU and accuracy were obtained in the study by Uchida et al.[31]

Among our cohort, the echo pattern 3 ('thyroid inferno'), as tool for identifying Graves' hyperthyroidism, showed the lowest accuracy (76.6%) and NPV (~57%). This meant that the echo patterns 0–2 (mainly US pattern 2) cannot reliably exclude GD; conversely, when 'thyroid inferno' was found, it was related to a high probability of GD (more than 95%). This reduced clinical effectiveness of US could be explained mostly by the use of a qualitative evaluation for the degree of vascularity (therefore an arbitrary assessment) and partly by a bias of interpretation of US images (albeit the reading of images was carried out by an experienced professional).

Our findings were in contrast with other studies: in two of them[15,16] the diagnostic utility of the three tests (TRAb, radionuclide scanning, Doppler ultrasonography) in distinguishing Graves' hyperthyroidism from destructive thyrotoxicosis was found to be comparable, while in the study by Ota et al[14] thyroid blood flow was superior to TRAb and radionuclide scanning. However, in these three studies[14–16] thyroid vascularity was assessed by a quantitative analysis, and the tests were performed with different equipments in different populations.

Lastly, TRAb and TSI titres similarly correlated to fT4 but not to fT3 concentrations. These result was in line with the findings by Carlé et al[32] according to which TRAb levels are main determinants for hormonal levels. The lack of a correlation between fT3 and antibodies levels could be due to a more pronounced variation in intrathyroid T3 production and/or peripheral T4 to T3 conversion in our patients.[33] As already demonstrated for TRAb,[34] we found that also TSI levels positively correlated with TcTU rates.

Most of the abovementioned past studies explored the diagnostic performance of TRAb assays, eventually comparing it with that of the imaging tests, but no study compared the TSI accuracy by the new 'bridge' assay Immulite® with radionuclide scanning and US in thyrotoxic scenario. Yet, only few studies[12,13,23,35] regarded the clinical use of Immulite® for TSI measurement, instead of the more cumbersome bioassays. These are the major strengths of our study, especially since the evaluation of both TRAb and TSI serum levels were carried out in the clinical setting of patients with untreated thyrotoxicosis (excluding other thyroid diseases and nonthyroid autoimmune diseases), where also thyroid scintigraphy and US are routinely performed.

This study has some limitations to be discussed. First, the number of patients with Graves' hyperthyroidism significantly exceeds that of non-GD ones; however, our GD 70%/non-GD 30% ratio is well in line with available clinical and epidemiological data.[1] Second, TRAb measurements were performed using a 2nd generation IMA (Kryptor® compact PLUS). However, 2nd and 3rd generation methods performed equally well, and both can be adopted for screening purposes in hyperthyroid patients.[9] Third, TSI levels could be more accurately assessed by a bioassay (instead of using a bridge immunoassay [Immulite®]), as recently demonstrated by Allelein et al.[35] However, we aimed to evaluate the only fully automated TSI assay, as bioassays remain not suitable for routine clinical use. The evaluation of thyroid vascularity by colour Doppler ultrasonography (US) could be better assessed in a quantitative manner, by means of thyroid blood flow, peak systolic velocity (PSV) measurements obtained from intrathyroidal and perithyroidal vasculature, or from inferior thyroid artery (ITA) and superior thyroidal artery (STA). Therefore, a fourth limitation could consist in not assessing thyroid vascularity by a quantitative methodology and in including nodular goitres, so that US accuracy could be underestimated. However, only two out of the 86 patients with Graves' hyperthyroidism had relevant multinodular goitres hampering US vascularity evaluation. Fifth, a selection bias were introduced because of the excluded thyrotoxic patients; however, this could minimally skew our results since only 18 patients (13 GD, three painless thyroiditis, two toxic multinodular goitre) missing of TSI measurement were not included in our cohort.

In conclusion, in the present cohort of patients with newly diagnosed and untreated thyrotoxicosis thyroid scintigraphy (quantitive evaluation by TcTU) showed the highest accuracy (with 1.3% as optimal cut-off) for the differential diagnosis between Graves' and non-Graves' hyperthyroidism. However, TRAb and TSI assays exhibited a high and similar clinical effectiveness, while the less reliable test to detect Graves' hyperthyroidism corresponded to the US vascularity assessment.

Therefore, TSI measurement by this recently developed 'bridge' assay (Immulite®) did not provide a significant added value in the diagnosis of Graves' hyperthyroidism compared with TRAb Kryptor®, with both TRAb and TSI titres predicating the degree of iodine uptake (as measured by TcTU) and fT4 biosynthesis and secretion. This finding imply that, at least in the diagnostic setting of thyrotoxicosis, TSI evaluation do not need to be ordered at all, since the performance of the TRAb assay is similar.

Thus, thyroid scintigraphy remains the most accurate method to differentiate causes of thyrotoxicosis. Basing on our data, however, TRAb assays can be alternatively adopted in this setting, limiting the use of thyroid scintigraphy (TcTU evaluation) to TRAb-negative patients. Thyoid US is less accurate than both TRAb/TSI and thyroid scintigraphy, but the 'thyroid inferno' pattern provides a high PPV for GD.