Management of thyroid tumors diagnosed cytologically as follicular neoplasms in a high-volume center: utility of a scoring system using serum thyroglobulin level, tumor size, ultrasound testing, and cytological diagnosis

Abstract

Managing thyroid nodules diagnosed cytologically as follicular neoplasms (FN) is challenging for patients and clinicians. Gene panel testing was recently introduced to determine the management strategy for FN; however, it is unavailable in Japan. In this study, we assessed FN management. This study included 2,144 FNs from 2,067 patients diagnosed between 2012 and 2018. Of these, 952 (44.5%) tumors underwent active surveillance, and 1,188 (55.6%) underwent immediate surgery (IS). Tumors of young patients (<55 years), male patients, and patients with serum thyroglobulin (Tg) ≥500 ng/mL, ultrasound diagnoses as FN or malignancy, large tumors (>4 cm), non-oxyphilic cytology, and cytological findings favoring malignancy and multiplicity underwent IS more frequently. Of the 1,412 tumors that underwent surgery, 279 (19.8%) and 1,133 (80.2%) were pathologically diagnosed as malignant and benign tumors or low-risk neoplasms, respectively. High Tg levels, non-benign ultrasound findings, cytological findings favoring malignancy, non-oncocytic cytology, and large tumor size were related to malignant pathology; however, tumor enlargement was not. The former three were independent predictors of malignancy in the multivariate logistic analysis. After assigning scores of 2 and 1 for cytological findings favoring malignancy and others, respectively, a receiver operating characteristic curve analysis indicated a score of 3 as the optimal cutoff for predicting malignant diagnosis; however, the area under the curve remained low, at 0.642. Accurately predicting the malignant pathology of FNs is challenging, and inducing gene panel testing will be helpful for managing FN tumors. Our scoring system would also be useful in estimating the risk of malignancy.

Introduction

A cytological examination is important for deciding the management strategies for thyroid nodules. Surgical treatment is generally recommended if a nodule is diagnosed as malignant upon cytological assessment, except for papillary thyroid microcarcinoma, for which active surveillance (AS) is often suitable. If diagnosed as benign, AS is generally recommended as a first-line management strategy unless patients prefer surgery because of the symptoms and cosmetic problems. However, managing nodules cytologically diagnosed as follicular neoplasms (FNs), based on the Bethesda System [1], is often challenging for patients and attending clinicians. Gene panel testing has recently been used to diagnose whether nodules are malignant or benign; the estimated ranges of sensitivity, specificity, and negative and positive predictive values (NPV and PPV) are 91–93%, 62–82%, 95–97%, and 17–66%, respectively [2, 3]. These tests are particularly useful for selecting nodules suitable for AS because the sensitivity and negative predictive values are higher than the specificity and positive predictive values. Further, Patel et al. (2018) validated the sequencing classifier for thyroid nodules with Bethesda III or IV and showed that under a 23.7% malignancy prevalence, it had a sensitivity of 91%, a specific of 68%, an NPV of 96.1%, and a PPV of 47.1% [4]. In Japan, however, these testing systems are not covered by national insurance, and no laboratories are available for testing; thus, they are not practically feasible.

In Japan, not all patients can undergo surgery immediately after their nodules are cytologically diagnosed as FN. Kihara et al. reported the direct association between ultrasound findings and malignant pathology and concluded that ultrasound is crucial for considering surgical indications for FNs [5]. However, other clinical and cytological factors should also be considered when deciding on management strategies in the real world. Therefore, in this study, we investigated the management of FNs and their clinical outcomes in our institution.

Materials and Methods

Patients

Specifically, 2,144 tumors from 2,067 patients (measuring 5–138 mm), diagnosed cytologically as FN between 2012 and 2018 at Kuma Hospital (Kobe, Japan) by a single pathologist (M.H.), were included. This study was approved by the Ethics Committee of Kuma Hospital (No. 20200709-1). We checked whether each FN had cytological findings favoring malignancy (see section Cytological Findings). All except eight cases underwent ultrasound evaluation; the tumors were diagnosed as benign, follicular tumor (FT), or malignant based on our classification system [6]. For serum thyroglobulin (Tg) levels, the values nearest to the date of cytology were used. However, cases of Tg antibody (TgAb)-positivity (>40.0 IU/mL) or lack of Tg data within 1 year before and after the date of cytology were considered missing values. Patients who had other tumors diagnosed as thyroid carcinoma by cytological assessment and those who underwent AS ≤1 year or underwent ultrasound examination only ≤2 times and did not undergo surgery were excluded. Patients with clinical lymph node metastasis (N1) or distant metastasis (M1) were also excluded.

Cytological findings

The guidelines of the Japan Thyroid Association subdivided FN cytology into favoring benign, borderline, and favoring malignant categories. FNs with none, one, or two or more of the five cytological findings suggestive of malignancy, such as dense follicles, three-dimensional microfollicules, trabecular arrangement, hyperchromasia, and enlarged nuclei, were classified as benign, borderline, or malignant, respectively [7]. A pathologist in our hospital (M.H.) then decided to adopt “favoring malignancy” as one of the cytological findings. After that, the rate of malignancy of tumors with favoring malignancy was reported to be significantly higher than that of borderline and favoring benign tumors [8].

Management strategies

The management of each tumor (surgery or AS) was at the attending clinicians’ discretion. Factors for considering IS include large size, high serum Tg level, cytology favoring malignancy, and non-benign ultrasound findings; however, no established surgical indications common to all clinicians are available. A few patients underwent conversion surgery (CS) after undergoing AS >1 year for various reasons such as tumor enlargement and strengthened symptoms; however, no indications of CS shared by all clinicians were available.

AS and evaluation of tumor enlargement

AS was performed using ultrasonography (at least once a year) to detect changes in ultrasound findings, including tumor size. In this study, we evaluated tumor enlargement by calculating the tumor volume-doubling rate (TV-DR) for tumors that underwent size evaluation ≥3 times; the details of this calculation have been previously described [9]. We set the cutoff of TV-DR at 0.5 (/year) to determine the high growth activity of the tumors.

Surgery

Surgery was performed for 1,412 tumors (1,353 patients); the extent of thyroidectomy was total thyroidectomy for 216 patients and subtotal thyroidectomy, hemithyroidectomy, or isthmectomy for the remaining 1,137 patients. Eighty-three patients underwent central node dissection, three of whom underwent lateral node dissection because their tumors were considered possibly malignant by the attending clinicians. The remaining patients did not undergo lymph node dissection.

Pathological examination

A pathologist (M.H.) examined all surgical specimens and pathologically diagnosed the FNs as benign, low-risk neoplasm, or malignant, based on the latest World Health Organization classification [10].

Postoperative follow-up

Patients who underwent surgery were requested to visit our hospital regularly, once or twice per year, for blood examinations and imaging studies, such as ultrasonography. We considered cases of recurrence only when structurally recurrent tumors were detected in imaging studies. Questionnaires were sent annually to patients referred to other hospitals to assess their physical status.

Statistical analysis

StatFlex (Artec, Osaka, Japan) was used for univariate and multivariate logistic analyses. Statistical significance was set at p < 0.05. We selected factors with p-values <0.20 for multivariate logistic analysis. Receiver operating characteristic (ROC) curve analysis was performed using the R software.

Results

Comparison between patients who underwent AS and IS

In our series, 956 FNs underwent AS for >1 year, of which 224 underwent CS for various reasons. Female patients (p = 0.044) and older patients (≥55 years) (p < 0.001) more frequently underwent AS (Table 1). Furthermore, AS was more likely to be considered for tumors showing oncocytic cytology (p = 0.001), low serum Tg levels (p < 0.001), a benign finding on ultrasonography (p < 0.001), no cytological findings favoring malignancy (p < 0.001), small size (p < 0.001), and no multiplicity (multiple tumors were diagnosed as FN on cytology) (p = 0.012). Among the tumors/patients who underwent AS, CS was performed more frequently on male patients (p = 0.031), younger patients (p < 0.001), and tumors with high serum Tg levels (p < 0.001), ultrasound diagnosis of FT or malignant tumors (p < 0.001), large size (p < 0.001), and multiplicity (p = 0.011).

Table 1 Comparison between backgrounds and clinical features of patients who underwent AS, CS (after AS), and IS

Variables	AS only	CS (after AS)	IS	Total	p-values (AS vs. CS)	p-values (AS, CS vs. IS)
Sex
Male	109 (14.9%)	47 (21.0%)	234 (19.7%)	390 (18.2%)	0.031	0.044
Female	623 (85.1%)	177 (79.0%)	954 (80.3%)	1,754 (81.8%)
Age
≥55 years	432 (59.0%)	99 (44.2%)	537 (45.2%)	1,068 (49.8%)	<0.001	<0.001
<55 years	300 (41.0%)	125 (55.8%)	651 (54.8%)	1,076 (50.2%)
Oncocytic
Yes	249 (34.0%)	68 (30.4%)	317 (26.7%)	634 (29.6%)	0.309	0.001
No	483 (66.0%)	156 (69.6%)	871 (73.3%)	1,510 (70.4%)
Serum Tg level
≥500 ng/mL	28 (3.7%)	24 (8.3%)	253 (10.3%)	305 (8.0%)
<500 ng/mL	519 (11.2%)	168 (17.7%)	736 (20.7%)	1,423 (17.4%)	<0.001	<0.001
Ultrasound findings
Benign	331 (45.6%)	72 (32.1%)	252 (21.2%)	655 (30.7%)	<0.001	<0.001
Non-benign	395 (54.4%)	152 (67.9%)	934 (78.8%)	1,481 (69.3%)
Cytological findings favoring malignancy
Yes	11 (1.5%)	5 (2.2%)	171 (14.4%)	187 (8.7%)	0.550	<0.001
No	721 (98.5%)	219 (97.8%)	1,017 (85.6%)	1,957 (91.3%)
Tumor size
>4 cm	26 (3.6%)	25 (11.2%)	466 (39.2%)	517 (24.1%)
>2 and ≤4 cm	210 (28.7%)	124 (55.4%)	545 (45.9%)	879 (41.0%)	<0.001	<0.001
≤2 cm	496 (67.8%)	75 (33.4%)	177 (14.9%)	748 (34.9%)
Multiplicity*
Multiple	33 (4.5%)	20 (8.9%)	99 (8.3%)	152 (7.1%)	0.011	0.012
Solitary	699 (95.5%)	204 (91.1%)	1,089 (91.7%)	1,992 (92.9%)
Total	732 (100%)	224 (100%)	1,188 (100%)	2,144 (100%)

FT, follicular tumor; AS, active surveillance; CS, conversion surgery; IS, immediate surgery; Tg, thyroglobulin;

Tg-Ab, anti-thyroglobulin antibody

*Multiple tumors were diagnosed as FN on cytology.

Evaluation of tumor enlargement during AS

We calculated the TV-DR of the 956 tumors that underwent AS, and TV-DR ≥0.5 (/year) was regarded as high growth activity. Tumors with oncocytic cytology (p = 0.002) showed more rapid enlargement on univariate analysis (Table 2). In a multivariate logistic analysis for factors with p-values <0.20 in the univariate analysis, young age (<55 years, p = 0.018) and oncocytic cytology (p = 0.002) were independently associated with high enlargement activity (Table 3).

Table 2 Relationship between tumor growth activity and patients’ backgrounds and clinical features

Variables	TV-DR ≥0.5 (/year)	TV-DR <0.5 (/year)	Total	p-values
Sex
Male	19 (21.8%)	137 (15.8%)	156 (16.3%)	0.144
Female	68 (79.2%)	732 (84.2%)	800 (83.7%)
Age
≥55 years	40 (46.0%)	491 (56.5%)	531 (55.5%)	0.060
<55 years	47 (54.0%)	378 (43.5%)	425 (44.5%)
Oncocytic
Yes	42 (48.3%)	275 (31.6%)	317 (32.2%)	0.002
No	45 (51.7%)	594 (68.4%)	639 (66.8%)
Serum Tg level
≥1,000 ng/mL	1 (1.4%)	15 (2.2%)	16 (2.2%)
≥500 and <1,000 ng/mL	5 (6.9%)	31 (4.6%)	36 (4.9%)
≥200 and <500 ng/mL	8 (11.1%)	87 (13.0%)	95 (12.9%)	0.773
<200 ng/mL	58 (80.6%)	534 (80.1%)	592 (78.8%)
Tg-Ab positive (≥40.0 IU/mL) or Tg not measured	15	202	217
Ultrasound findings
Benign	30 (34.5%)	373 (43.2%)	403 (42.4%)
FT or malignant	57 (65.5%)	490 (56.8%)	547 (57.6%)	0.116
Unknown	0	6	6
Cytological findings favoring malignancy
Yes	2 (2.3%)	14 (1.6%)	16 (1.7%)
No	85 (97.7%)	855 (98.4%)	940 (98.3%)	0.650
Tumor size
>4 cm	3 (3.4%)	48 (5.5%)	51 (5.3%)
>2 and ≤4 cm	34 (39.1%)	300 (34.5%)	334 (35.0%)	0.551
≤2 cm	50 (57.5%)	521 (55.5%)	571 (59.7%)
Multiplicity*
Multiple	5 (5.7%)	48 (5.5%)	53 (5.5%)	0.809
Solitary	82 (94.3%)	821 (94.5%)	903 (94.5%)
Total	87 (100%)	869 (100%)	956 (100%)

FT, follicular tumor; LI, labeling index; Tg, thyroglobulin; Tg-Ab, anti-thyroglobulin antibody

*Multiple tumors were diagnosed as FN on cytology.

Table 3 Multivariate logistic analysis of evaluable factors before active surveillance for tumor volume-doubling rate ≥0.5/years for the entire series

Variables	Odds ratio	95% CI	p-values
Sex Male	1.56	0.90–2.71	0.111
Age <55 years	1.73	1.10–2.73	0.018
Oncocytic	2.40	1.51–3.80	0.002
FT or malignant ultrasound findings	1.59	0.99–2.56	0.055

CI, confidence interval; FT, follicular tumor

Multivariate analysis was performed for factors with p < 0.20 on univariate analysis.

Relationship between pathological findings and clinicopathological features

To date, 1,412 tumors have been surgically dissected and pathologically examined; 884 tumors were diagnosed as benign (391 follicular adenomas and 493 other benign nodules), 249 were classified as low-risk neoplasms (185 FTs with uncertain malignant potential and 64 well-differentiated tumors with uncertain malignant potential), and 279 were classified as malignant (231 follicular thyroid carcinomas [FTCs], 19 papillary thyroid carcinomas [PTCs], 14 poorly differentiated carcinomas, and 15 well-differentiated carcinomas not otherwise specified) tumors. Table 4 shows the relationship between various clinicopathological features and pathological diagnoses. Malignant tumors were more likely to be non-oncocytic (p = 0.010), have higher serum Tg levels (p = 0.001), be diagnosed as FT or malignant on ultrasonography (p < 0.001), exhibit cytological findings favoring malignancy (p < 0.001), and be large-sized (p = 0.002). TV-DR was measured for 336 tumors that underwent CS after AS >1 year with ultrasound examination ≥3 times; however, it did not significantly differ between benign tumors or low-risk neoplasms and malignant tumors.

Table 4 Comparison between backgrounds and clinicopathological features of tumors diagnosed as benign or low-grade neoplasm and malignant

Variables	Benign or low-risk neoplasm	Malignant	Total	p-values (malignant vs. non-malignant)
Sex
Male	229 (81.5%)	52 (18.5%)	281 (100%)	0.555
Female	904 (79.9%)	227 (20.1%)	1,131 (100%)
Age
≥55 years	515 (81.0%)	121 (19.0%)	636 (100%)	0.531
<55 years	618 (79.7%)	158 (20.3%)	776 (100%)
Oncocytic
Yes	326 (84.7%)	59 (15.3%)	385 (100%)	0.010
No	807 (78.6%)	220 (21.4%)	1,027 (100%)
Serum Tg level
≥1,000 ng/mL	113 (71.1%)	46 (28.9%)	159 (100%)
≥500 and <1,000 ng/mL	92 (78.0%)	26 (22.0%)	118 (100%)	0.001
≥200 and <500 ng/mL	188 (78.7%)	51 (21.3%)	239 (100%)
<200 ng/mL	559 (84.1%)	106 (15.9%)	665 (100%)
Tg-Ab positive (≥40.0 IU/mL) or Tg not measured	181	50	231
Ultrasound findings
Benign	286 (88.3%)	38 (11.7%)	324 (100%)
FT or malignant	846 (77.9%)	240 (22.1%)	1,086 (100%)	<0.001
Unknown	1	1	2
Cytological findings favoring malignancy
Yes	105 (59.7%)	71 (40.3%)	176 (100%)	<0.001
No	1,028 (83.2%)	208 (16.8%)	1,236 (100%)
Tumor size
>4 cm	371 (75.6%)	120 (24.4%)	491 (100%)
>2 and ≤4 cm	546 (81.6%)	123 (18.4%)	669 (100%)	0.002
≤2 cm	216 (85.7%)	36 (14.3%)	252 (100%)
Multiplicity*
Multiple	102 (85.7%)	17 (14.3%)	119 (100%)	0.117
Solitary	1,031 (79.7%)	262 (20.3%)	1,293 (100%)
TV-DR
≥0.5/year	79 (79.8%)	20 (20.2%)	99 (100%)
<0.5/year	188 (79.3%)	49 (20.7%)	237 (100%)	0.922
Unknown	866	210	1,076
Total	1,133 (80.2%)	279 (19.8%)	1,412 (100%)

LI, labeling index; FT, follicular tumor; Tg, thyroglobulin; Tg-Ab, anti-thyroglobulin antibody

*Multiple tumors were diagnosed as FN on cytology.

Table 5 shows the results of a multivariate logistic analysis of factors evaluated before surgery for malignant pathology; high serum Tg levels (≥500 ng/mL) (p = 0.031), FT or malignant ultrasound findings (p = 0.024), and cytological findings favoring malignancy (p < 0.001) were independent predictors of malignant pathology.

Table 5 Multivariate logistic analysis of evaluable factors before surgery for malignant pathology of tumors in the entire series (Analysis for 1,179 tumors of which all the below factors are available)

Variables	Odds ratio	95% CI	p-values
Oncocytic	0.71	0.50–1.03	0.069
Serum Tg level
≥500 ng/mL	1.45	1.03–2.04	0.031
FT or malignant ultrasound findings	1.60	1.06–2.40	0.024
Cytological findings favoring malignancy	2.78	1.89–4.10	<0.001
Tumor size >4 cm	1.32	0.96–1.82	0.090
Multiplicity*	0.65	0.35–1.21	0.170

CI, confidence interval; LI, labeling index; FT, follicular tumor

Multivariate analysis was performed for factors with p < 0.20 on univariate analysis. *Multiple tumors were diagnosed as FN on cytology.

Scoring of factors suggesting malignancy and the ROC curve for diagnosing malignancy

We selected the five factors with p < 0.10 on multivariate logistic analysis in Table 5. We first assigned a score of 1 for four of them; non-oncocytic cytology, high Tg level (≥500 ng/mL), FT or malignant ultrasound findings, and tumor size >4 cm because the odds ratio of these factors on multivariate analysis were <2 (Table 5) [11]. We analyzed 1,179 tumors for which all these scores were available. Table 6 shows the incidence of malignancy and other diseases according to the scores 0–4. The incidence of malignancy increased from 9.4% to 34.0% with increasing scores. Fig. 1 indicates the ROC curve for these four scores. The AUC was not high (0.609 [95% confidence interval, 0.571–0.647]). The optimal cutoff value was 2 (specificity, 0.318; sensitivity, 0.833).

Table 6 Relationship between scores of malignancy for four factors and pathological findings

Pathology	Scores
	0	1	2	3	4	Total
Benign or low-risk neoplasm	58 (90.6%)	244 (88.4%)	378 (80.3%)	205 (76.5%)	66 (66.0%)	951 (80.7%)
Malignant	6 (9.4%)	32 (11.6%)	93 (19.7%)	63 (23.5%)	34 (34.0%)	228 (19.3%)
Total	64 (100%)	276 (100%)	471 (100%)	268 (100%)	100 (100%)	1,179 (100%)

Fig. 1  A receiver operating characteristic curve with the four scores for diagnosing pathological malignancy

Next, we assigned a score of 2 for the remaining factor, cytology favoring malignancy, because its odds ratio in the multivariate analysis (Table 5) was higher (>2) [11]. Table 7 shows the incidence of malignancy and other diseases according to the scores of 0–6. The incidence of malignancy increased from 6.5% to 59.0% as the score increased from 0 to 6. Fig. 2 shows the ROC curve of the six scores for diagnosing pathological malignancy. The AUC became slightly higher at 0.642 (95% confidence interval, 0.604–0.680). The optimal cutoff value for the score predicting the risk of a malignant diagnosis was 3 (specificity, 0.658; sensitivity, 0.535); however, the incidence of malignant pathology in patients with a score of 3 remained low at 21.9% (Table 7).

Table 7 Relationship between scores of five factors of malignancy and pathological findings

Pathology	Scores
	0	1	2	3	4	5	6	Total
Benign or low-risk neoplasm	58 (93.5%)	233 (89.6%)	335 (81.7%)	198 (78.1%)	100 (70.9%)	23 (62.2%)	9 (41.0%)	951 (80.7%)
Malignant	4 (6.5%)	27 (10.4%)	75 (18.3%)	54 (21.9%)	41 (29.1%)	14 (37.8%)	13 (59.0%)	228 (19.3%)
Total	62 (100%)	260 (100%)	410 (100%)	247 (100%)	141 (100%)	37 (100%)	22 (100%)	1,179 (100%)

Fig. 2  A receiver operating characteristic curve with the six scores (including cytology favoring malignancy) for diagnosing pathological malignancy

Relationship between the scores and the AS, CS, and IS groups

We investigated score distribution in patients who underwent AS, CS, and IS. We enrolled 544, 192, and 987 patients who underwent AS, CS, and IS, respectively (Table 8), for whom all scores were available. Analysis for 4 scores indicated that the scores of patients who underwent IS were significantly higher (p < 0.001) than those of patients who underwent AS or CS. In analysis for 6 scores, the scores of patients who underwent IS were significantly higher (p < 0.001) than those of patients who underwent AS or CS; the scores of patients who underwent CS were significantly higher (p < 0.001) than those of patients who underwent AS without CS.

Table 8 Relationship between the scores and active surveillance (AS), conversion surgery (CS), and immediate surgery (IS) groups

	AS group	CS group	IS group	Total	p-values (AS vs. CS)	p-values (AS and CS vs. IS)
Receiver operating characteristic (ROC) score (4 scores)
≥3	23 (4.2%)	22 (11.5%)	346 (35.0%)	391 (22.7%)	0.090	<0.001
≤2	521 (95.8%)	170 (88.5%)	641 (65.0%)	1,332 (77.3%)
ROC score (6 scores)
≥4	6 (1.1%)	6 (3.1%)	194 (19.7%)	206 (12.0%)	<0.001	<0.001
≤3	538 (98.9%)	186 (96.9%)	793 (80.3%)	1,517 (88.0%)
Total	544 (100%)	192 (100%)	987 (100%)	1,723 (100%)

Postoperative follow-up and prognosis of patients

The median postoperative follow-up periods of benign tumors, low-risk neoplasms, and malignant tumors were 4.57 years (0.07–11.22 years), 5.08 years (0.10–10.95 years), and 5.93 years (0.23–11.18 years), respectively. The follow-up period of malignant tumors was significantly longer (p < 0.001) than that of other tumors. No patient showed distant metastases on imaging studies during AS. After surgery, four patients with tumors diagnosed as low-risk neoplasms or benign tumors and 20 patients with tumors diagnosed as malignant showed recurrence in various organs (six local and 18 distant organs). Only one patient with a malignant tumor (diagnosed as PTC on pathology) died of thyroid carcinoma 18 months after surgery.

Discussion

This study investigated whether and how nodules cytologically diagnosed as FN based on existing clinico-cytological factors can be effectively and safely managed. In our institution, clinicians recommend surgery for FNs with conventional predictors such as large size, non-benign ultrasonographic findings, and high Tg levels. Furthermore, the incidence of tumors with cytology favoring malignancy was significantly higher in the IS group, which was reasonable considering the high incidence of malignancy in tumors with cytology favoring malignancy, as described below [8]. Other than those, tumors of young patients were more likely to undergo resection, possibly because the attending clinicians considered patients’ long-life expectancy. Female patients may prefer IS to alleviate anxiety. Tumors subjected to CS after AS had characteristics similar to those of IS, indicating that CS was performed based on similar concerns that warrant IS in patients and clinicians.

In this study, we evaluated tumor enlargement activity using TV-DR and observed that FNs in young patients had greater enlargement activity (Tables 2, 3). Papillary thyroid microcarcinomas in young patients were more likely to enlarge [12], indicating that thyroid tumors in young patients generally have higher enlargement activity than those in older patients, according to the natural history of tumors. However, the tumor size at diagnosis was not related to enlargement activity; therefore, we cannot predict the future change in size based on the present tumor size.

From the standpoint of clinical and cytological examinations, high Tg levels, FT or malignant findings on ultrasonography, and cytological findings favoring malignancy were independently related to malignant pathology in our series. Preoperative serum Tg levels can help in the preoperative differential diagnosis of FNs [13-15]. In our study, the incidence of high Tg levels was significantly higher in tumors diagnosed as malignant than in the other tumors (Table 4), indicating that Tg levels can help determine surgical indications for FNs. Recently, some studies have considered the usefulness of ultrasonographic findings for diagnosing FTC using ultrasound risk stratification systems, such as TI-RADS [16-20]. Kihara et al. reported that of the 137 cases in which FNs were surgically removed, the incidence of pathologically diagnosed malignant tumors was 9%, 19%, and 53% for nodules with ultrasonographic findings as benign, borderline, and malignant, respectively, based on our classification system [5, 6]. In our series, tumors with ultrasonographic findings of FT or malignancy were more often diagnosed as malignant on pathology, indicating that ultrasonographic findings can help decide surgical indications for FNs.

Cytological findings favoring malignancy were considered important for predicting the malignant pathology of FNs [7, 8]. Indeed, tumors with these cytological findings more often underwent IS (Table 1) and were diagnosed as malignant tumors on pathology with high incidence (Table 4), indicating that these findings effectively encourage clinicians to perform surgery for suspicious tumors. However, these characteristics may be difficult to detect in institutions where no pathologist specializes in thyroid carcinoma. Furthermore, the diagnostic accuracy for malignancy was high; however, deciding surgery based only on these findings should not be appropriate (the positive predictive value was 40.3%).

Tumors cytologically diagnosed as oncocytic showed a higher enlargement activity (Table 3); however, the tumors were less often diagnosed as malignant tumors on pathology (Table 4). The reasons for such difference remain unclear; however, two studies from Japan showed that the prognosis of oncocytic carcinoma did not differ from or was even better than that of non-oncocytic carcinoma [21, 22], indicating that oncocytic FNs can be managed similarly to non-oncocytic FNs.

We attempted to predict malignant pathology in combination with five characteristics. First, we adopted four characteristics, including non-oncocytic cytology, high Tg level (≥500 ng/mL), FT or malignant ultrasound findings, and tumor size >4 cm, and assigned a score of 1 to each. We performed an ROC curve analysis for these scores. However, the AUC was low (0.609) (Fig. 1). Cytological findings favoring malignancy were regarded important for predicting malignant pathology of FNs [7, 8], although these characteristics may be challenging to detect in institutions where no pathologist specializes in thyroid carcinoma; an odds ratio obtained during the multivariate analysis was 2.78, which was much higher than the others. Then, we assigned a score of 2 for this and repeated the ROC curve analysis. The resulting AUC was improved to 0.642; however, this result indicates that, although cytological findings are combined, an accurate diagnosis of follicular carcinoma remains difficult before surgery.

We investigated the scores for each case belonging to the AS, CS, and IS groups (Table 8). Scores in the IS group were significantly higher than those in the AS + CS groups, and scores of the CS group were also significantly higher than or tended to be (p = 0.090) higher than those of the AS group. Surgery for FNs can be recommended for other reasons (including cosmetic problems and patients’ symptoms) than those factors indicating malignancy; even so, the scores were significantly higher in the order of the IS, CS, and AS groups, suggesting that our management strategies for FNs were appropriate within the compass of existing markers discriminating malignancy from benign tumors.

This study has some limitations. First, it was a retrospective study, and the management of FNs (AS or IS, and when CS is recommended) was at the discretion of the attending clinicians. Second, the study might have encountered a verification bias. As written in the first paragraph of the Discussion, patients having factors suggestive of malignancy such as large size, non-benign ultrasound findings, high Tg levels, and cytology favoring malignancy were more likely to undergo diagnostic thyroidectomy than those having test results suggesting benignity, which should lead spuriously high sensitivity and low specificity. Third, the unit of analysis was uncertain. Some factors, such as tumor size, cytological, and ultrasound findings were unique for tumors, but Tg level may fluctuate because of, for example, the coexistence of adenomatous nodule/multinodular goiter. Fourth, sex, age, and Tg levels were characteristics applied to each patient, but not to each tumor, which affects the data of patients having multiple FNs. Therefore, our data in the tables may not be straightforwardly interpreted. Finally, TV-DR can only be evaluated for tumors that undergo AS, and the tumor enlargement activity is unknown for tumors immediately resected after diagnosis.

As shown in the Graphical Abstract, there is a limitation (low AUC) to accurately predicting the pathological malignancy of FN tumors even though cytological findings favoring malignancy were included. Therefore, the introduction of gene panel testing should be helpful for more accurate diagnosis. Indeed, Patel et al. showed a high NPV (95%) for their series, when the prevalence of malignancy was 22.4% [4]. However, in our series, tumors with high scores (e.g., Scores 5 and 6, in Table 7) showed a much higher probability of malignancy, at 37.8% and 59.0%, respectively, and in these circumstances, if the same sensitivity (88%) and same specificity (64%) as Patel’s series were set, the NPV would become much lower than 95%. Although gene panel testing could be a useful tool for the diagnosis of malignancy, it was not perfect in the differential diagnosis, and our scores could be helpful in estimating the probabilities of malignancy and in deciding the surgical indication of tumors.

Graphical Abstract 

In conclusion, our diagnosis scores suffer from low AUC, but help predict the probabilities of malignancy to decide the surgical indication, even when gene panel testing is induced for thyroid nodules cytologically diagnosed as FN.

Author Contributions

Yasuhiro Ito: Conception and design of the work, drafting the article, and final approval of the version to be published. Makoto Kawakami, Mitsuyoshi Hirokawa, Masashi Yamamoto; Data collection, data analysis, and interpretation. Minoru Kihara, Naoyoshi Onoda, Akihiro Miya, Akira Miyauchi, Takashi Akamizu; Critical revision of the manuscript.

Disclosure

Yasuhiro Ito is a member of Endocrine Journal’s Editorial Board.

References

• 1  Ali SZ, Vanderlaan PA (eds) (2023) The Bethesda System for Reporting Thyroid Cytopathology (3rd). Springer Nature, Cham, Switzerland.
• 2   Nishino  M,  Krane  JF (2020) Role of ancillary techniques in thyroid cytology specimens. Acta Cytol 64: 40–51.
• 3   Nishino  M,  Krane  JF (2018) Updates in thyroid cytology. Surg Pathol Clin 11: 467–487.
• 4   Patel  KN,  Angeli  TE,  Babiarz  J,  Barth  NM,  Blevins  T, et al. (2018) Performance of a genomic sequence classifier for the preoperative diagnosis of cytologically indeterminate thyroid nodules. JAMA Surg 153: 817–824.
• 5   Kihara  M,  Ito  Y,  Hirokawa  M,  Masuoka  H,  Yabuta  T, et al. (2011) Role of ultrasonography in patients with cytologically follicular thyroid tumor. Auris Nasus Larynx 37: 508–511.
• 6   Yokozawa  T,  Fukata  S,  Kuma  K,  Mastuzuka  F,  Kobayashi  A, et al. (1996) Thyroid cancer detected by ultrasound-guided fine-needle aspiration biopsy. World J Surg 20: 848–853.
• 7   Kakudo  K,  Kameyama  K,  Miyauchi  A,  Nakamura  H (2014) Introducing the reporting system for thyroid fine-needle aspiration cytology according to the new guidelines of the Japan Thyroid Association. Endocr J 61: 539–552.
• 8   Hirokawa  M,  Suzuki  A,  Kawakami  M,  Kudo  T,  Miyauchi  A (2022) Criteria for follow-up of thyroid nodules diagnosed as follicular neoplasm without molecular testing—The experience of high-volume thyroid centre in Japan. Diagn Cytopathol 50: 223–229.
• 9   Miyauchi  A,  Kudo  T,  Ito  Y,  Oda  H,  Yamamoto  M, et al. (2019) Natural history of papillary thyroid microcarcinoma: kinetic analyses on tumor volume during active surveillance and before presentation. Surgery 165: 25–30.
• 10   Lloyd  RV,  Osamura  RY,  Kloppel  G,  Rosai  J (eds) (2017) WHO classification of tumours of endocrine organs. International Agency for Research on Cancer, Lyon, France.
• 11   Dreyfus  J,  Audureau  E,  Bohbot  Y,  Coisne  A,  Lavie-Bandie  Y, et al. (2022) TRI-SCORE: a new risk score for in-hospital mortality prediction after isolated tricuspid valve surgery. Eur Heart J 43: 654–662.
• 12   Ito  Y,  Miyauchi  A,  Fujishima  M,  Noda  T,  Sano  T, et al. (2023) Thyroid-stimulating hormone, age, and tumor size are risk factors for progression during active surveillance of low-risk papillary thyroid microcarcinoma in adults. World J Surg 47: 392–401.
• 13   Chen  Z,  Lin  Y,  Lai  S,  Wang  P,  Li  J, et al. (2022) The utility of serum anti-thyroglobulin antibody and thyroglobulin in the preoperative differential diagnosis of thyroid follicular neoplasms. Endocrine 76: 369–376.
• 14   Okamoto  T,  Karbe  M,  Iihara  M,  Yamazaki  K,  Okamoto  J, et al. (1997) Measuring serum thyroglobulin in patients with follicular thyroid nodule: its diagnostic implications. Endocr J 44: 187–193.
• 15   Besic  N,  Sesek  M,  Peric  B,  Zgajnar  J,  Hocevar  M (2008) Predictive factors of carcinoma in 327 patients with follicular neoplasm of the thyroid. Med Sci Monit 14: CR459–CR467.
• 16   Liu  BJ,  Zhang  YF,  Zhao  CK,  Wang  HX,  Li  MX, et al. (2020) Conventional ultrasound characteristics, TI-RADS category and shear wave speed measurement between follicular adenoma and follicular thyroid carcinoma. Clin Hemorheol Microcirc 75: 291–301.
• 17   Wu  SJ,  Tan  L,  Ruan  JL,  Qiu  Y,  Hao  SY, et al. (2022) ACR TI-RADS classification combined with number of nodules, halo features optimizes diagnosis and prediction of follicular thyroid cancer. Clin Hemorheol Microcirc 85: 323–334.
• 18   Lin  Y,  Lai  S,  Wang  P,  Li  J,  Chen  A, et al. (2022) Performance of current ultrasound-based malignancy risk stratification systems for thyroid nodules in patients with follicular neoplasms. Eur Radiol 32: 3617–3630.
• 19   Li  HJ,  Yang  YP,  Liang  X,  Zhang  XZ,  Xu  XH (2023) Comparison of the diagnostic performance of three ultrasound thyroid nodule risk stratification systems for follicular thyroid neoplasm: K-TIRADS, ACR-TIRADS and C-TIRADS. Clin Hemorheol Microcirc 85: 395–406.
• 20   Cho  YY,  Ahn  SH,  Lee  EK,  Park  YJ,  Choi  D, et al. (2024) Malignancy risk of follicular neoplasm (Bethesda IV) with variable cutoffs of tumor size: a systemic review and meta-analysis. J Clin Endocrinol Metab 109: 1383–1392.
• 21   Sugino  K,  Kameyama  K,  Ito  K,  Nagahama  M,  Kitagawa  W, et al. (2013) Does Hurthle cell carcinoma of the thyroid have a poorer prognosis than ordinary follicular thyroid carcinoma? Ann Surg Oncol 20: 2944–2950.
• 22   Ito  Y,  Hirokawa  M,  Masuoka  H,  Higashiyama  T,  Kihara  M, et al. (2022) Prognostic factors for follicular thyroid carcinoma: the importance of vascular invasion. Endocr J 69: 1149–1156.

Abbreviations

Follicular neoplasms

FN

Active surveillance

AS

Immediate surgery

IS

Follicular tumor

FT

Tumor volume-doubling rate

TV-DR

Thyroglobulin

Tg

Conversion surgery

CS

Receiver operating characteristic

ROC

Follicular thyroid carcinomas

FTCs

Papillary thyroid carcinomas

PTC

Anti-thyroglobulin antibody

Tg-Ab