Diagnostic accuracy of follicular variant of papillary thyroid carcinoma in fine-needle aspirates processed by ultrafast Papanicolaou stain

Histologic follow-up of 125 cases


  • Presented, in part, at the 34th Annual Meeting of the Australian Society of Cytology, Brisbane, Australia, October 16–18, 2004.



Detection of follicular variant (FV) of papillary carcinoma (PC) of the thyroid is considered difficult in cytology. Various ancillary studies have been used to increase the sensitivity in the detection of FVPC in fine-needle aspiration (FNA). We previously reported that the clear nuclei of PC became conspicuous in Ultrafast Papanicolaou (UFP)-stained smears. This study reports our experience on using this UFP-induced artifact in the detection of FVPC.


Over an 11-year period, 5637 ultrasound-guided thyroid FNAs were performed. All samples were smeared and air-dried. Diff-Quik stain was used to assess the colloid and UFP stain was used to study the nuclei.


Histologic follow-up was available from 125 cases. Of the 107 aspirates with diffuse “grape-like” watery clear nuclei, histologic follow-up showed 94 PCs, 8 follicular adenomas containing atypical nuclei, and 5 nonneoplastic lesions. Of the 18 aspirates with focal clear nuclei, histologic follow-up showed 6 PCs, 8 follicular adenomas, and 4 nonneoplastic nodules. As reported previously, histologic follow-up of 147 FNAs contained follicles with normal nuclei showed 107 true follicular neoplasms, 32 hyperplastic nodules, and 8 FVPC. The sensitivity, specificity, positive predictive value, negative predictive value, and diagnostic accuracy in the detection of PC were 93.6%, 94.9%, 93.6%, 94.9%, and 94.3%, respectively, when follicular adenomas containing atypical nuclei were counted as encapsulated FVPC; and 92.6%, 85.6%, 80%, 94.9%, and 88.3%, respectively, when follicular adenomas containing atypical nuclei were counted as benign.


UFP stain is one of the options to increase the sensitivity of detection of FVPC in preoperative FNA, and to triage microfollicular nodules into those that require surgery no matter how small, and those which can be followed when small. Cancer (Cancer Cytopathol) 2006. © 2006 American Cancer Society.

Diagnosis of follicular variant (FV) of papillary carcinoma (PC) of the thyroid is considered difficult on cytology. There are several studies on the cytologic clues improving the diagnostic accuracy.1–8 In Diff-Quik (Baxter Diagnostics, McGaw Park, IL)-stained smears of 63 cases of FVPC, Gallagher et al.1 reported “tumor cellularity,” arrangement in empty follicles, rosettes, and tubules. Nuclei were twice the size of RBCs, had smooth contours, were hyperchromatic, and varied in shape. Nuclear overlapping was common. Some nuclei resemble “arrowheads.” Pink-stained colloid was frequent. In Papanicolaou-stained smears, Logani et al.2 reported that nuclear grooves, powdery chromatin, and intranuclear pseudoinclusions were important in a study of 52 cases of FVPC. Wu et al.4 reported the statistically significant more frequent presence of colloid in FVPC than true follicular neoplasms in a study of 29 cases. Yan et al.5 reported that a branching sheet arrangement was a low-power clue for FVPC in a study of 23 cases. The observation of colloid and branching sheets at low power, combined with high-power observation of grooved nuclei, intranuclear pseudoinclusions, and fine chromatin in Papanicolaou stain are the currently used cytologic clues in diagnosing FVPC.

In recent years, ancillary studies have shown great promise in improving the diagnostic accuracy in FVPC in FNA, including the presence of Ret/PTC gene rearrangements by reverse transcriptase polymerase chain reaction (RT-PCR) and Southern hybridization,9 analysis of BRAF point mutation,10 and immunohistochemistry11 using HBME-1,12 cytokeratin 19,13 and galactin-3.14

Ultrafast Papanicolaou stain15 (UFP) is a 90-second Papanicolaou stain, using rehydrated air-dried cells to obtain flattened larger cells to maximize the resolution of cellular details, alcoholic formalin to accelerate the fixation, and Richard-Allan Scientific (Kalamazoo, MI) Hematoxylin II and Cytostain to reduce the steps. In air-dried preparations the flattened cells appear larger, because the observed nuclear area in air-dried cells reflects the nuclear volume.16 Rehydration in normal saline, discovered by Chan and Kung in 1988,17 permits water to enter the air-dried cells through osmosis, restoring transparency. As the result of the alteration of nuclear envelopes and chromatin structure,18, 19 the nuclei of UFP-processed PC cells would be distended with water, flattening many nuclear grooves and becoming “grape-like” watery clear in smears (Fig. 1), correlating to “Orphan Annie-eyed” clear nuclei, a 2D cartoon drawing seen in paraffin sections. The “grape-like” watery clear nuclei are conspicuous20, 21 and can be easily seen at ×10 objective lens (Fig. 2B, left). Basolo et al.22 reported 98% correlation of final diagnosis using intraoperative cytology prepared with the UFP method as compared with 71% using frozen sections alone. Paessler et al.23 reported that UFP-stained intraoperative cytology was helpful in avoiding 2-step completion thyroidectomy. Most of these cases aided by UFP-stained intraoperative cytology represented FVPC on final diagnosis.22, 23

Figure 1.

Grape-like nuclei of papillary thyroid carcinoma were caused by the artifact of air-dry rehydration with saline and alcoholic formalin fixation used in Ultrafast Papanicolaou stain. Except for 1 grooved nucleus (arrow), the remaining nuclear grooves have been smoothed out by the water distention. (A) Chinese watercolor painting of a cluster of watery oval grapes (painting by Sharon Chia-wei Hsu). (B) Grape-like nuclei arranged in a cluster in papillary thyroid carcinoma.

Figure 2.

The microfollicles in thyroid FNA smears, processed by Ultrafast Papanicolaou stain, were triaged into (A) normal with round, dark, nuclei or (B) grape-like with enlarged, oval, watery clear nuclei. (C) Rare cases showed mixed normal and grape-like nuclei. The histologic follow-up showed (A) follicular adenoma, (B) follicular variant of papillary carcinoma, and (C) well-differentiated carcinoma with mixed papillary and follicular features (capsular invasion). Original magnification ×100 (left); ×400 (middle); ×1000 (right).

This study documents our experience with preoperative FNA diagnosis of FVPC detected by the grape-like artifact of UFP stain and followed prospectively with surgical pathology diagnosis.


From November 1994 to December 2005, 5637 ultrasound-guided FNA of the thyroid from 3920 patients were performed at a radiology office in New York City, 1 day a week, average 15 minutes per patient. Patients were instructed not to take aspirin or other blood-thinning agents at least 5 days before biopsy to minimize blood dilution of the samples. Consent for follow-up was obtained from patients before the biopsy. As reported previously,21 using the ATL HDI 3000 system (Advanced Technology Laboratories, Bothell, WA), the thyroid nodules were localized by a 10 to 5 Hz compact linear array probe, guided by 1 of 2 registered medical diagnostic sonographers with 15-16 years' experience. The nodules were aspirated free-hand by 1 of 2 radiologists using a 27-G needle attached to 10-mL disposable syringes with 1-2 mL suction. All samples were directly smeared and air-dried. A pair of smears was made from each pass, and an average of 2 passes were made from each nodule. The smears were inspected macroscopically while waiting to be dried: honey-like, thin, watery, or bloody. Abundant honey-like colloid reflects light, thus can be recognized. The first smear was stained with Diff-Quik (Baxter Diagnostics) to assess the background, i.e., abundant colloid or exceedingly bloody. If the first smear was consistent with follicular neoplasm or suspicious for malignancy, the remaining smears were stained with UFP (Richard Allen Scientific) to assess the nuclear features. At ×10 objective, the nuclei are then triaged into normal (Fig. 2A) (spherical and granular nuclei) or “grape-like” (oval, watery clear nuclei) (Fig. 2B). Infrequently, we encountered follicles with mixed normal and grape-like watery clear nuclei, which were focally distributed (Fig. 2C). As soon as the “grape-like” nuclei were detected at ×4 objective, and confirmed at ×10 objective, the patients were told “suspicious for cancer, advise surgery.”

During the study period, we encountered 152 (2.7%) thyroid nodules with “grape-like” watery clear nuclei, diffusely distributed in 118 nodules and focally distributed in 34 nodules. The former group was reported as “suspicious for FVPC” and the latter group was reported as “atypical, cannot excluded FVPC.” The histologic follow-ups were obtained from various hospitals in New York City.


One hundred seven histologic follow-ups (90.7%) were obtained from 118 nodules with diffuse “grape-like” watery clear nuclei and 18 histologic follow-ups (52.9%) were obtained from 34 nodules with focal “grape-like” watery clear nuclei. Of the 125 patients with histologic follow-up, 103 were female and 12 were male, age range 22-82 years (mean, 46). The size of the nodules ranged from 0.4-4.9 cm (mean, 1.7 cm). Out of 107 FNAs reported as suspicious for FVPC, histologic follow-up showed 94 (87.8%) PCs, 8 (7.5%) follicular adenomas, and 5 (4.7%) nonneoplastic nodules. The frozen section for the PC cases was most frequently reported as “follicular neoplasm, defer to permanent section.” In the 9 follicular adenoma cases, they were usually accompanied by comments stating that atypical nuclei were present in the follicular adenoma. In the 5 nonneoplastic cases, 4 were hyperplastic nodules, and in 1 case the small nodule (measured 0.4 cm in vivo) could not be found in the lobectomy specimen, even though the second opinion of the cytopathologist at the hospital of surgery concurred with the original FNA diagnosis. Out of the 34 FNAs reported as “atypical, FVPC cannot be excluded,” histologic follow-up was available from 18 cases and showed 6 (33.3%) PCs, 8 (44.4%) follicular adenomas, and 4 (22%) hyperplastic nodules. As reported previously,24 using UFP-stained FNA, histologic follow-up of 147 FNAs containing microfollicles with normal nuclei had surgery and showed 107 true follicular neoplasms (92 follicular adenomas and 10 minimally invasive follicular carcinomas, 5 widely invasive follicular carcinomas), 32 hyperplastic nodules, and 8 FVPC. The above data are summarized in Table 1.

Table 1. Histologic Follow-up of Microfollicles Triaged into “Grape-Like” Watery Clear Nuclei and Normal Nuclei in Preoperative FNA Smears Processed with Ultrafast Papanicolaou Stain
Histology“Grape-like” nuclei (125)HistologyNormal nuclei (147)
  1. FVPC: follicular variant of papillary carcinoma; FA: follicular adenoma; FC: follicular carcinoma.

Follicular adenoma with atypical nuclei16FA + FC117

The accuracy of using “grape-like” watery clear nuclei in detecting FVPC in FNA smears is shown in Table 2. If follicular adenomas containing atypical nuclei were counted as encapsulated FVPC, the grape-like UFP-stained artifact had a sensitivity of 93.6%, specificity of 94.9%, positive predictive value of 93.6%, negative predictive value of 94.9%, and diagnostic accuracy of 94.3%. If follicular adenomas containing atypical nuclei were counted as follicular adenomas, the sensitivity was 92.6%, specificity 85.6%, positive predictive value 80%, negative predictive value 94.6%, and diagnostic accuracy 88.3%. Most of the FVPC cases exhibited nuclear features diagnostic for PC in Diff-Quik stain, but required a ×40 objective and much longer time of examination to distinguish from the nuclei of true follicular neoplasms. The watery clear grape-like nuclei of FVPC can be immediately distinguished from the dark nuclei of true follicular neoplasms using a ×4 objective and then confirmation using a ×10 objective (Fig. 2., left). A comparison of the nuclear features of a tissue-proven follicular adenoma and a tissue-proven FVPC in Diff-Quik stain and UFP stain is illustrated in Figure 3.

Table 2. Accuracy of Using “Grape-Like” Water Clear Nuclei in Detecting Follicular Variant of Papillary Carcinoma (FVPC) in FNA Smears
 Follicular adenoma with atypical nuclei counted as encapsulated FVPC, %Follicular adenoma with atypical nuclei counted as benign, %
  1. PV: predictive values.

(+) PV93.6%80.0%
(−) PV94.9%94.9%
Figure 3.

Comparison of nuclear features of a tissue-proven follicular adenoma and a tissue-proven follicular variant of papillary carcinoma in Diff-Quik stain and Ultrafast Papanicolaou (UFP) stain. (A) Round nuclei without nuclear grooves seen in follicular adenoma. Diff-Quik stain. (B) Dark, round nuclei with granular chromatin seen in follicular adenoma. UFP stain. (C) Oval nuclei with nuclear grooves seen in FVPC. Diff-Quik stain. (D) Watery clear grape-like nuclei seen in FVPC. UFP stain. Original magnification ×400, (A–D).


The susceptibility of the nuclei of PC to the “grape-like” UFP artifact is due to an intrinsic chromatin defect in the nuclei of PC. Fischer et al.18, 19 used retrovirus as a vector to insert RET/PTC oncogene into primary cultures of human thyroid epithelial cells, transforming them into cells with the nuclear features of PC. Fusco et al.25 analyzed 46 noninvasive thyroid nodules that exhibit borderline morphologic features of PC by RET activation by immunohistochemistry and, in selected cases, by RT-PCR performed on RNA extracted after laser capture microdissection of the tumor foci with and without PC features and positive RET immunoreactivity. RET immunoreactivity was identified in 65.2% of the nodules, where it closely paralleled the morphologic changes. RET/PTC1 or RET/PTC3 were detected only in microscopic foci with PC features in most of the thyroid nodules. They concluded that RET activation closely parallels the nuclear features of PC and that it is restricted to foci with PC nuclear features.

The reason why “suspicious” is added to our preoperative FNA diagnosis of FVPC is because our patients are referred by physicians all over New York City. The histologic criteria for diagnosing FVPC has not yet reached a consensus among surgical pathologists.26 Baloch et al.3 diagnosed encapsulated follicular pattern lesion with focal nuclear features of PC as encapsulated FVPC. Their opinion is based on bone metastasis on long-term clinical follow-up,27 whereas Chan28 and other expert surgical pathologists use stricter criteria and felt justified to err on the benign side and reported this type of tumor as follicular adenoma to avoid overtreatment of a cancer with excellent prognosis. The Chernobyl Pathologists group,29 with the notable dissent of LiVolsi, advocated the term “well-differentiated tumor of uncertain malignant potential” for encapsulated follicular pattern lesion with “questionable” nuclear features of PC and without capsular invasion. Because we cannot predict if the nodule containing “grape-like” nuclei is encapsulated, which of the 14 hospitals the patient will undergo surgery, and the philosophy of the surgical pathologist who would make the final diagnosis, we simply add a “suspicious” to all FVPC in our FNA report in order to avoid any scenario of total thyroidectomy with a final diagnosis of “follicular adenoma.”

Using the UFP protocol, follicular neoplasm of microfollicular cell type showing normal nuclei (Fig. 2A) can be monitored when the nodules are smaller than 2.1 cm, because the probability for follicular carcinoma was only 2.28% to 4.25% based on statistical analysis of 397 thyroidectomies.30

In our experience, the “grape-like” watery clear nucleus is a reliable criteria to diagnosis FVPC in cytology because this phenomenon parallels the “Orphan Annie-eyed” clear nucleus in paraffin section, which is the criteria used by surgical pathologists in diagnosing FVPC. The intrinsic nuclear envelope and chromatin defect of PC is the cell biologic basis for both artifacts. Relying on this UFP-induced artifact to distinguish FVPC (which needs surgery no matter how small) from true follicular neoplasms (which can be followed when small), the number of thyroid FNA referred to this practice increased 7-fold from 138/yr in 1995 to 1005/yr in 2005.


The authors thank Sharon Chia-wei Hsu for the Chinese watercolor painting of grapes in Figure 1.