Diagnostic accuracy of [99mTc]Tc-Sestamibi in the assessment of thyroid nodules

[99mTc]Tc-Sestamibi (MIBI) is an increasingly used tool for evaluation of thyroid nodules. However, there is a lack of evidence about the accuracy of this method in the European population. The aim of this study was to assess the utility of MIBI for the differentiation of thyroid nodules in a large cohort. 161 patients underwent MIBI, followed by a thyroidectomy. We used a dual phase MIBI protocol. Interpretation of the images included a scoring system from 0 (absent) to 3 (increased); this was to provide a scale for the uptake of the thyroid nodule in comparison to the paranodular tissue. Additionally, we evaluated the tracer uptake trend in late images compared to early images. We used the final histopathology as the reference standard. Scores 0-1 in early images, scores 0-2 in late images, and an absence of increasing uptake in the thyroid nodule in late images, showed the best predictive values to exclude malignancy, respectively (negative predictive value (NPV) 89%). Highest sensitivity (91%) for malignant nodules was evident in early images with a score 1-3. Highest specificity (91%) was obtained when the negative was defined as an absence of uptake-increase, in the late images. This study confirms that the most valuable feature of MIBI is the high NPV. Thus, with the appropriate interpretation method, high sensitivity and specificity, and moderate PPV can be obtained.


INTRODUCTION
In Germany, more than 15 million people have thyroid nodules and about 90,000-100,000 patients undergo thyroid surgery each year; yet only approximately 6000 of these patients are diagnosed with thyroid cancer. This means that not even 1 in 1000 thyroid nodules is a malignant one [1][2][3][4]. To avoid such overtreatment, it is very important to improve the diagnostic approach to thyroid nodules.
Nodules with a diameter of > 1 cm require a thyroid scintigraphy to evaluate the functional activity [5]. There is an increased suspicion of malignancy (2-5%) in hypoand isofunctional nodules, and in these cases, further evaluation is needed [6].
[ 99m Tc]Tc-Sestamibi(MIBI) is increasingly used in practice for the investigation of thyroid nodules. The tracer accumulates in mitochondria-rich cells, common in hyperplasia, malignant tumors, or parathyroid adenomas [6]. This may help distinguish between benign and malignant nodules, which is important for the selection those patients who would benefit from surgery [7].
Fine needle aspiration biopsy (FNAB) is a safe and minimally invasive method for the evaluation of thyroid nodules [8]. However, FNAB can be problematic, especially in patients with multinodular goiter or in those with difficult to access lesions. Furthermore, this technique is highly examiner-dependent and the rate of non-diagnostic results varies between 2-32%, requiring a re-biopsy [9][10][11][12][13][14][15].
Another issue is that there is no standardized MIBIimaging protocol; there can be either a single-phase protocol with late images (1-2 h post injection) or a dualphase protocol with early (15-30 min post injection) and late images. The imaging can include planar images and/ or single-photon emission computed tomography (SPECT) images [6].
The aim of this study was to examine the accuracy of MIBI-scintigraphy for the differentiation of thyroid nodules in the German population. The secondary objective was to determine the best acquisition time for the images. To our knowledge, this is the largest study, to date, evaluating the utility of MIBI in thyroid nodules.

Histopathological results
The histopathological examination revealed 139 (86%) benign and 22 (14%) malignant nodules. From the malignant tumors, 15 (68%) were papillary, four (18%) were follicular, and three (14%) were papillary cancer with follicular differentiation. There were no cases of anaplastic cancer. Medullary carcinomas were not observed, because all cases of elevated tumor marker calcitonin underwent a calcium stimulation test for further evaluation and did not receive a MIBI-scintigraphy.
From the histopathologically benign nodules, 48 had normal cytology results and one case had pathologic cytology. There were 19 patients with intermediate cytology, meaning that the results were suspicious but unconducive, for example cellular atypia or follicular neoplasia. In 26 cases there was inadequate aspiration material.
From the malignant nodules, four had normal results, four had intermediate/suspect results, and two had pathological results in the cytology. Five aspirates were non-diagnostic.
The obtained p-value of the utility of FNAB to differentiate between benign and malignant nodules was 0.058 (not significant). Statistical analyses showed a good NPV of 92%, the PPV was 23%. Sensitivity and specificity were 60% and 71%, respectively (Table 3). Table 4 presents the utility of MIBI-scintigraphy to differentiate between benign and malignant nodules. Flow charts of different imaging interpretation methods are depicted in Figures 1-3. Tables 5-7 show the diagnostic values (predictivity, sensitivity, specificity, and accuracy) of the different imaging interpretation methods. PPV and NPV of early and late images are also shown in Figure 4. Highest sensitivity (91%) for the detection of benign nodules was observed in the early images, when only score 0 findings were defined as negative. The highest specificity (91%) was obtained in the interpretation of the trend of the tracer uptake in late images; normal finding = decreased or constant uptake, abnormal finding = increased uptake. www.impactjournals.com/oncotarget

Early images
Among the benign nodules the score was 0 in 13 nodules, while 46 nodules had a low MIBI-uptake (score 1), in comparison to the paranodular tissue. Another 43 nodules had a score 2, and 37 lesions had score 3.
Among the malignant nodules, the score was 0 in two nodules. Five lesions had a score 1, eight were isointense (score 2), and seven nodules showed increased uptake (score 3) in the scintigraphy.
The ability of the early images to differentiate between benign and malignant nodules was not significant.

Late images
In the late images, among the benign nodules, 39 had no uptake (score 0), 41 had a score 1, and 29 had a score 2. 30 benign lesions had a score 3.
Among the malignant nodules, the score in the late images was 0 in 10 lesions, 1 in three nodules, and 2 in one nodule. Eight nodules showed high uptake (score 3).    In summary, the late images could reliably differentiate between benign and malignant nodules (p = 0.044, according to chi-square-test).

Washout/tracer retention in the late images
Among the benign lesions, tracer uptake in the late images, compared with the early images, remained The p-value from the early images was not significant, the late images achieved p = 0.044, and the washout/tracer retention method was p = 0.034. constant in 48 cases. There was an uptake decline in 78 of the lesions and an uptake increase in 13 lesions. Among the malignant nodules, there was a sufficient washout in 12 cases, in four nodules the uptake was constant, and six nodules showed an increased retention of the tracer over time.
The washout/tracer retention method showed good reliability (p = 0.034) in terms of differentiation between benign and malignant nodules.

DISCUSSION
There are several publications reporting the diagnostic value of MIBI-scintigraphy for the evaluation of thyroid nodules. Most impressing is MIBI's high negative predictive value, of up to 100%; mean > 97% [9,[16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31]. Only Kresnik et. al. reported a much lower NPV of 84%; however, the study included only 62 patients [21]. Most of the studies have a small patient cohort (range 25-    [6,20]. In comparison, the prevalence in Germany is only 11% [6]. It is the opinion of the authors, that image interpretation should be adapted to the prevalence of thyroid carcinomas in the studied population. In the current study, the best NPV (89%) of thyroid nodules was observed in lesions with a score of 0-1 in the early images, and a score of 0-2 and in the late images (86% NPV). Good NPV (89%) was also obtained in the interpretation of the trend of the uptake over time; negative finding = decreased or constant uptake in the late images. This method also showed the best specificity of 91%. The reliable NPV of MIBI-scintigraphy can justify a "wait-and-see" strategy.
False-negative findings are rarely reported in the literature. Examples are two patients with undifferentiated / anaplastic carcinoma from the study of Kresnik et al    and Foldes et al [19,21]. However, there were also small number of patients with differentiated cancer (papillary n=3 / follicular n=1), who had a negative MIBI-scan [26,28,29]. In our patient group, there were two cases of score 0 malignant nodules in the early images, and 10 false negative results in the late images. The false negatives may indicate that a continuous follow-up with ultrasound of the neck and if necessary repeated FNAB are needed in these patients.
We used the dual-phase study protocol with planar images for performing MIBI. The late images could reliably differentiate between benign and malignant nodules, p = 0.044. Good reliability for the differentiation (p = 0.034) was also revealed in the interpretation of the tracer uptake trend (washout or tracer retention) in the late images. However, better diagnostic values could have been achieved with SPECT imaging. For example, the study of Schenke et. al. compared SPECT with planar MIBI-imaging, achieving a much better NPV; 100% versus 91.7% [27].
There are some limitations and biases in the current study which need to be mentioned. Firstly, not all patients underwent a FNAB. From the 109 patients who had an ultrasound-guided FNAB, 31 (28%) had nondiagnostic biopsies. A better pre-operative diagnostic procedure would have decreased the benign nodules in this study. Secondly, we informed patients who did not undergo a FNAB, or who had an inadequate biopsy, or who had a FNAB with no evidence of malignancy, that the only way to exclude a malignancy is to undergo surgery. This may have raised the number of cases of operated patients. Thirdly, the majority of the malignant nodules were papillary carcinoma (68%); therefore, it is uncertain if the utility of MIBI-scintigraphy is the same for all thyroid cancer entities. Furthermore, there were no cases of anaplastic cancer in this study. This may be because anaplastic cancer is often clinically classified as suspect (fast growing and irregular in the ultrasound) and directly referred for surgical evaluation. Finally, there were patients who received MIBI-scintigraphy despite suspicious FNAB findings. The reason for this is that these patients received both a FNAB and MIBI on the same day, and we were not aware of the cytopathology results. To avoid such bias in the future, and to confirm the reliability of MIBI, prospective studies are needed.
In conclusion, the current study confirms that the most valuable feature of MIBI-scintigraphy is its NPV. Using a dual-phase protocol, 89% NPV could be achieved either in early images of lesions with a score 0-1, or in late images with a score 0-2. Furthermore, with the appropriate image interpretation method, a high sensitivity (91%), high specificity (91%) and moderate PPV (32%) can be obtained. The current study confirms that the most valuable feature of MIBI is the NPV. Thus, with the appropriate image interpretation method, high sensitivity and specificity, and moderate PPV can be obtained.

Patients
In total, 603 patients underwent a MIBI-scintigraphy at our department between October 2005 and August 2014. All patients had at least one thyroid nodule, which was cold/indifferent in the [ 99m Tc]Tc-pertechnetate (TPT)scintigraphy. At the time this study was conducted, MIBI was not an established diagnostic tool. For that reason, we informed patients who did not undergo a FNAB, or who had an inadequate biopsy, or who underwent a FNAB with no evidence of malignancy, that the only way to exclude a malignancy is surgery. Patients with elevated tumor marker calcitonin and suspicion of medullary cancer did not receive a MIBI-scintigraphy. In these cases we used the calcium stimulation test for further evaluation. From the 603 studied patients, 161 received a histopathological diagnosis after thyroid surgery, and were included for further analysis. The local ethics committee of our university approved this retrospective study; informed consent was obtained from all participants.

Imaging and interpretation
All patients underwent TPT-scintigraphy followed by MIBI-scintigraphy. After intravenous injection of MIBI (standard dose 370 MBq) we used a dual-phase protocol with planar images approximately 20 min post-injection (p.i.) and 2 h p.i.
The uptake in the examined nodule was compared with the paranodular thyroid tissue. The findings were classified visually; accumulation of the tracer in the thyroid nodule was classified as absent (score 0), low (score 1), isointense (score 2), or increased (score 3). Examples of this scoring system are presented in Figure 1. We also considered the washout or the tracer-retention in the late images, compared to the early images.
A correlation with the TPT-scintigraphy showed either a "match" (i.e., concordant decreased uptake of both tracers) or a "mismatch" (i.e., cold nodule with increased MIBI-uptake).

Fine needle aspiration biopsy (FNAB)
109 patients underwent a FNAB. The biopsy was performed through guided ultrasound using a 20-gauge needle attached to a 20 ml Cameco syringe-pistol. Smears were made and air-dried slides were stained with hematoxylin-eosin. The remaining aspirate material in the syringe was rinsed with 0.9% sodium chloride. The fluid material was centrifuged, the sediment was smeared and slides were prepared. Adequacy of the aspirates was assessed on the basis of the guidelines of the Papanicolaou Society [32].

Histopathology
The gold standard for histologic diagnosis served as a reference. From the 603 patients, 161 patients underwent surgery for further histopathological examination of the thyroid nodules. Routine staining with hematoxylin-eosin, Elastica van-Giesson and immunochemistry, if necessary, was performed.

Statistical analyses
Statistical analyses were performed with IBM SPSS software version 22. Different image interpretation methods were compared for their utility to differentiate between benign and malignant nodules. We used the chi-square test to determine the statistical significance of the results. The significance level was accepted at 5%. Additionally, to obtain the validity of the imaging, we classified the findings into true positive (TP), true negative (TN), false positive (FP), and false negative (FN). Sensitivity (SEN) was calculated as TP/ (TP + FN), specificity (SPE) as TN/(TN + FP), and accuracy (ACC) as (TP + TN)/(TP + TN + FP + FN). The positive predictive value (PPV) was defined as TP/ (TP + FP), the negative predictive value (NPV) as TN/ (TN + FN).