^99mTc-MIBI Imaging in the Presurgical Characterization of Thyroid Follicular Neoplasms: Relationship to Multidrug Resistance Protein Expression

Patients

This study prospectively included 51 consecutive patients (11 men and 40 women; age range, 28–72 y; median, 55 y) with ^99mTc-pertechnetate–cold solitary (or prevalent) thyroid nodules at least 10 mm in diameter (median, 17 mm; range, 10–80 mm), cytologically diagnosed as Thy3 (nononcocytic or oncocytic follicular neoplasm) according to the British Thyroid Association guidelines (Table 1) (36). All patients were diagnosed and treated at San Luigi Hospital, Orbassano (Turin, Italy), between 2006 and 2007.

In all patients, serum levels of free thyroxine, free triiodothyronine, thyroid-stimulating hormone, antithyroglobulin and antithyroid peroxidase antibodies, parathyroid hormone, and calcitonin were determined. All patients underwent a thyroid ultrasound examination followed by ^99mTc-MIBI imaging and were subsequently operated on to confirm or exclude a malignant lesion.

The study was approved by the San Luigi Hospital review board, and written informed consent was obtained from each patient.

Fine-Needle Aspiration Biopsy

Preoperative FNAB was performed with a 22-gauge needle attached to a 30-mL plastic syringe. The aspirated fluid was in part expelled and smeared onto charged slides and was fixed and stained with a rapid hematoxylin-and-eosin method for adequacy evaluation or processed for Giemsa staining. The remaining material was used to prepare alcohol-fixed cell blocks and hematoxylin- and eosin-stained sections (9).

Oncocytic follicular neoplasms were defined by the presence of oncocytic cells—large cells characterized by abundant deeply eosinophilic and granular cytoplasm, with hyperchromatic nuclei having prominent nucleoli (37)—in at least 50% of the total follicular cells present on smears or cell-block sections.

Cytologic and Histologic Samples

The histologic diagnoses of the 36 nononcocytic follicular neoplasms were microfollicular goiter (9 cases), follicular adenoma (12 cases), minimally invasive follicular carcinoma (2 cases), follicular variants of papillary carcinoma (12 cases), and poorly differentiated carcinoma (1 case) (Table 1). The histologic diagnoses of the 15 oncocytic follicular neoplasms were oncocytic microfollicular goiter (5 cases), oncocytic follicular adenoma (6 cases), oncocytic insular carcinoma (1 case), and oncocytic follicular variants of papillary carcinoma (3 cases) (Table 1).

Thyroid tumors were classified according to the most recent criteria of the World Health Organization (37). Oncocytic variants of follicular adenoma, follicular variants of papillary carcinoma, insular carcinoma, and microfollicular goiter were defined by the presence of oncocytic changes in at least 75% of the lesion (37).

Thyroid ^99mTc-MIBI Scintigraphy

^99mTc-MIBI scintigraphy was performed on each patient using a large-field-of-view γ-camera (Axis; Philips Netherland) equipped with a high-resolution parallel-hole low-energy collimator. Static images of the neck were acquired in the anterior view with a 128 × 128 matrix, a 1.33 zoom, and a pixel size of 1.75 mm and stored on a Philips Odyssey computer. Images were obtained at 10 min (early image) and 120 min (late image) after intravenous injection of 400 MBq of ^99mTc-MIBI, according to American guidelines (38). Labeling efficiency was assessed by thin-layer chromatography and was found always to exceed 94%. To ensure correct positioning of the patients, we always included both salivary glands and activity from the myocardium in the ^99mTc-MIBI images.

The images were independently assessed by 2 experienced nuclear physicians before surgery. They used both a visual scoring method and a semiquantitative technique and were not aware of the clinical data or histologic results. If they disagreed, they discussed the case and reached a consensus.

Visual Analysis of Images

The early and delayed ^99mTc-MIBI thyroid images of a given patient were placed side by side for comparison. The visual assessment used the following scoring system developed in our institution: pattern 0, no increased uptake of the nodule in either early or delayed images; pattern 1, increased uptake of the nodule only in early images; pattern 2, increased uptake of the nodule both in early and in delayed images; and pattern 3, increased uptake of the nodule only in delayed images.

Semiquantitative Analysis of Images

Both early and delayed images were displayed on the same screen. The pertechnetate image of the patient was also displayed on the same screen when definite anatomic landmarks were needed. On each early image, an exact region of interest was drawn over the perimeter of the thyroid nodule and then moved to the opposite normal lobe. These regions of interest were then copied to the delayed images. If high radiotracer uptake in the nodule made it difficult to see the contralateral normal lobe, the intensity of the image was increased until the normal lobe was clearly visualized and the pertechnetate image was used as an anatomic landmark. For background, early-image rectangular regions of interest were drawn on the right superior region of the patient's thorax and then copied to the delayed images. The mean count in each region of interest was determined, and early ratio (ER) and delayed ratio (DR) were calculated by dividing nodule counts by normal-tissue counts after area correction for background activity. The retention index (RI) was then found using the formula RI = (DR – ER) × 100/ER.

Antibodies

Two primary monoclonal antibodies were used in this study: clone C219 diluted 1/150 (Zymed Laboratories) to P-gp and clone QCRL-1 diluted 1/400 (Sigma-Aldrich) to MRP1. These antibodies were selected from a preliminary pilot immunohistochemical study involving a panel of commercial standardized antibodies.

Immunoperoxidase Technique

Paraffin blocks containing surgical specimens were cut at a 4-μm thickness, collected onto poly-l-lysine–coated slides, dewaxed in xylene, rehydrated in decreasing ethanol concentrations, and incubated for 10 min in phosphate-buffered saline (PBS) (pH 7.4). A heat-induced antigen retrieval procedure was performed for all antibodies by placing slides in 0.01 M ethylene-diamine-tetraacetate buffer (pH 8.0) in a microwave oven set at high power for 3 consecutive cycles of 5 min each. The slides were left to cool for 20 min at room temperature and rinsed in PBS. Endogenous peroxidase activity was quenched with methanol–hydrogen peroxide 3% for 15 min at room temperature. Slides were then washed twice in PBS for 5 min, and tissue sections were incubated first in the blocking solution (ChemMate Buffer Kit; DakoCytomation) and then with the primary antibodies in a humidified chamber at room temperature for 45 min. After a prolonged wash in PBS, the sections were incubated with an enzyme-labeled polymer preconjugated with antimouse secondary antibodies (Envision System; DakoCytomation) at room temperature for 30 min. The sections were finally washed 3 times in PBS and incubated with 3′-3′-diaminobenzidine-tetrahydrochloride for 10 min. Slides were subsequently rinsed in tap water, counterstained with Mayer hemalum solution, mounted in Entellan (Merck), and examined with a DM-RBE photomicroscope (Leica Microsystems AG). Positive controls were human liver, kidney, and colon tissues. Negative controls were obtained by omitting the primary antibody.

P-gp and MRP1 immunostaining was evaluated by 3 independent observers without knowledge of clinical and pathologic data. MRP1 topographic expression (cytoplasmic or apical membranous immunoreactivity) was analyzed using the following semiquantitative scale: − (no reactivity), + (focal reactivity, < 25% of positive follicular cells), ++ (moderate reactivity, 25%−50% of positive follicular cells), and +++ (diffuse reactivity, > 50% of positive follicular cells).

Statistical Analysis

Visual scintigraphic patterns 2 or 3 and 0 or 1 were considered positive and negative, respectively.

The intra- and interobserver variability (reproducibility) and repeatability coefficients (39) were computed for both ER and DR measures.

The Mann–Whitney U test (Statistica software, version 6.1; Statsoft Italia s.r.l.) was used to determine the differences in ER, DR, and RI indices between benign and malignant nononcocytic lesions, as well as between benign and malignant oncocytic lesions. Results were considered significant when the P value was less than 0.05. The histologic diagnosis on the surgical specimens was regarded as the gold standard.

The correlation between nodule size and RI value was expressed using the Spearman rank correlation coefficient (s) with 95% confidence interval (CI) and P value (level of significance, P < 0.05).

Receiver operating characteristic analysis (MedCalc software, version 10.0.2.0; Frank Shoonjans) was performed to determine the RI threshold above which malignant nononcocytic thyroid nodules could be detected.

Sensitivity, specificity, positive and negative predictive values, likelihood ratio, and diagnostic accuracy of visual and semiquantitative ^99mTc-MIBI analyses were computed for benign versus malignant nononcocytic and oncocytic lesions, separately.

For the purpose of statistical analysis, MRP1 was considered positive when apical membrane immunoreactivity was observed in more than 25% of tumor cells.

MRP1 immunohistochemistry was associated and correlated to ^99mTc-MIBI findings by means of nonparametric 2-tailed Fisher exact testing and Kruskal–Wallis 1-way ANOVA (level of significance, P < 0.01).

Visual Analysis

Eleven of 15 nononcocytic carcinomas showed a significantly increased ^99mTc-MIBI uptake at 120 min (scintigraphic pattern 2), whereas 4 follicular variants of papillary carcinoma presented a total radiotracer washout at 120 min (Table 1). Among nononcocytic benign lesions, 17 of 21 samples displayed a ^99mTc-MIBI pattern 0 or 1, whereas 4 cases (2 follicular adenomas and 2 micofollicular goiters) showed a pattern 2 (Table 1).

The sensitivity and specificity of ^99mTc-MIBI visual analysis in differentiating benign from malignant nononcocytic tumors were 73.33% (CI, 44.91–92.05) and 80.95% (CI, 58.08–94.44), respectively. Positive and negative predictive values were 73.33% (CI, 44.91–92.05) and 80.95% (CI, 58.08–94.44), respectively. The positive likelihood ratio was 3.85 (CI, 1.51–9.79). The diagnostic accuracy of ^99mTc-MIBI visual analysis was 77.78% (Table 2).

All but 1 oncocytic lesion exhibited a significantly increased ^99mTc-MIBI uptake at 120 min (scintigraphic pattern 2 or 3) (Table 1).

The sensitivity and specificity of ^99mTc-MIBI visual analysis in differentiating benign from malignant oncocytic tumors were 100% (CI, 40.23–100) and 9.09% (CI, 1.51–41.33), respectively. Positive and negative predictive values were 28.57% (CI, 8.57–58.08) and 100% (CI, 16.55–100), respectively. The positive likelihood ratio was 1.10 (CI, 0.91–1.33). The diagnostic accuracy of ^99mTc-MIBI visual analysis was 33.33%.

Semiquantitative Analysis

The results of semiquantitative ^99mTc-MIBI analysis are reported in Table 3.

The intraobserver and interobserver correlation coefficients (r) of ER values, with their related regression equations, were r = 0.996 (y = 0.9661x + 0.0501) and r = 0.983 (y = 1.0684x − 0.0536), respectively, whereas those of the DR values were r = 0.984 (y = 0.9822x + 0.1695) and r = 0.956 (y = 0.9597x + 0.2107), respectively. The repeatability coefficients for early and delayed images were 0.283 and 0.523, respectively.

No significant difference in ER values was observed between malignant and benign nononcocytic lesions, whereas significant differences were found for both DR and RI values (P < 0.05 and P < 0.001, respectively) in the same group.

No significant differences in the ER, DR, and RI indices were observed between malignant and benign oncocytic thyroid nodules.

No correlation was found between nodule size and RI value (s = −0.098, CI = −0.364 to 0.182, P = 0,486).

The RI cutoff corresponding to the highest accuracy for discriminating between benign and malignant nononcocytic thyroid lesions was −11.94. Eleven of 15 nononcocytic nodules with an RI above this cutoff had a histologic result consistent with malignancy (Figs. 1A and 1B), whereas 17 of 21 lesions with a RI equal to or below this cutoff was histologically benign (Figs. 1C and 1D). For this threshold, the sensitivity and specificity were 100% (CI, 78.03–100) and 90.48% (CI, 69.58–98.55), respectively. The positive and negative predictive values were 88.24% (CI, 63.52–98.20) and 100% (CI, 82.20–100), respectively. The diagnostic accuracy of the semiquantitative ^99mTc-MIBI assay was 94.44%, and the positive likelihood ratio was 10.50 (CI, 2.81–39.24) (Table 2). The area under the receiver operating characteristic curve was 0.943 (CI, 0.811–0.991, P = 0.0001), and SE was 0.044.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

Thyroid ^99mTc-MIBI scan. (A and B) Representative positive case (RI, 71.76) of solitary cold thyroid nodule in right lobe histologically diagnosed as nononcocytic solid/trabecular poorly differentiated carcinoma. (A) Early image with tracer accumulation in nodule (ER, 2.62). (B) Delayed image with tracer retention in lesion (DR, 4.50). (C and D) Representative negative case (RI, −40.64) of solitary cold thyroid nodule in left lobe that proved to be microfollicular goiter. (C) Early image with tracer increase uptake in nodule (ER, 2.81). (D) Delayed image with nearly complete tracer washout from nodule (DR, 1.67).

Multidrug Resistance Protein Expression

No P-gp immunoreactivity was found in the follicular cells of either nononcocytic or oncocytic histologic specimens (Fig. 2A).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

MDR1 expression in thyroid nodules with cytologic result of follicular neoplasm. (A) No P-gp immunoreactivity of follicular cells in nononcocytic emblematic case (follicular adenoma). (B and C) MRP1 membranous apical immunoexpression in case of nononcocytic microfollicular goiter (B) and in case of nononcocytic follicular variant of papillary carcinoma (C), both with negative RI indices. Cytoplasmic immunoreactivity with granular pattern is seen. (D) No MRP1 membranous apical labeling in case of nononcocytic solid/trabecular papillary carcinoma with positive RI value. Cytoplasmic immunoreactivity with granular pattern is seen. Arrows show linear membranous apical immunoreactivity. Immunoperoxidase with Mayer solution counterstain was used (magnification, ×160; insets: magnification, ×630).

All but 3 cases (2 nononcocytic follicular adenomas and 1 nononcocytic follicular variant of papillary carcinoma) had a moderate or diffuse cytoplasmic immunoreaction for MRP1, with a granular pattern of staining (Figs. 2B, 2C, and 2D).

Fourteen of 21 (67%) nononcocytic benign lesions showed apical membranous MRP1 staining in more then 25% of follicular cells (Fig. 2B), whereas 7 cases were either negative (4 samples) or only focally (≤25% of cells) positive (3 samples; 14%) (Table 1).

All but 3 (80%) (Fig. 2C) nononcocytic carcinomas (1 minimally invasive follicular carcinoma and 2 follicular variants of papillary carcinomas) were negative for MRP1 or showed only occasional cells with apical immunoreactivity (Table 1; Fig. 2D).

Only 4 cases (3 follicular adenomas and 1 microfollicular goiter; 27%) of 15 oncocytic lesions had apical MRP1 expression in 25%−50% of follicular cells, whereas the remaining 11 cases (73%) were negative or showed only occasional cells with apical immunoreactivity (Table 1).

Normal thyroid tissue surrounding tumor nodules was MRP1-negative or contained occasional positive cells. All nonfollicular cells were completely nonreactive.

^99mTc-MIBI and MRP1 Membranous Apical Expression

Eighteen of 23 cases (78%) with a negative RI value at semiquantitative ^99mTc-MIBI analysis had apical membranous MRP1 staining in more then 25% of follicular cells (Figs. 2B and 2C), whereas 3 cases (13%) were only focally (≤25% of cells) positive. The remaining 2 cases (9%) were negative at the apical level.

All but 3 cases (89%) with a positive RI value at semiquantitative ^99mTc-MIBI analysis did not show apical MRP1 staining or showed only occasional cells with apical immunoreactivity (Fig. 2D). The 3 positive cases (11%) presented MRP1 apical immunoreactivity in 25%−50% of follicular cells.

MRP1 membranous apical expression in follicular cells was significantly associated with negative ^99mTc-MIBI RI values (P < 0.001).

A significant inverse correlation between the percentage of apical MRP1–positive follicular cells and negative values for RI was also detected (P < 0.001, 3 degrees of freedom).