Prediction of cytology–histology discrepancy when Bethesda cytology reports benign results for thyroid nodules in women: with special emphasis on pregnancy

ABSTRACT

Objectives: Benign category of Bethesda classification is generally well known to carry a false-negative rate of 0–3%. The current study was designed to investigate the rate of false-negative cytology in patients who underwent thyroidectomy for presumably benign thyroid diseases. Predictive risk factors for false results and malignancy were evaluated along with cytology–histology discrepant cases.

Materials and methods: Females who underwent thyroidectomy between May 2014 and December 2022 were included. Demographics, ultrasound (US) features, fine-needle aspiration (FNA) diagnosis, surgical indications and outcomes, final histology reports, risk factors, and malignancy rate were recorded. Cytology–histology discrepant cases were further evaluated for interpretation errors and risk factors. Statistical analyses were performed using Fisher’s exact and Mann–Whitney U tests.

Results: Of 581 women with a benign thyroid disease who underwent thyroidectomy, 91 was diagnosed as incidental carcinoma (15.6%) and most was T1a (4.9 ± 2.7 mm, 95.6%). Final histology reports revealed mostly papillary carcinoma (93.4%). Predictors of malignancy such as age, family history, previous radiation exposure, and iodine-deficient diet did not help in risk stratification (p > 0.05, for each). However, FNA taken during pregnancy was determined as a risk factor (n = 7, 7.6%, p < 0.05) since it may cause a delay in diagnosis. Cytology–histology discrepant cases were seen to be mostly due to sampling errors (45%, p < 0.05), followed by misinterpretations (37.3%, p < 0.05). There was no reason for discrepancy in 17.5%, and this was linked to inherent nature of thyroid nodule with overlapping cytologic features. Best identifiable risk factor for misinterpretation was pregnancy as well (n = 5, 14.7%, p < 0.05).

Conclusions: Risk of malignancy in a presumably benign thyroid disease should not be ignored. Radiology–cytology correlation by an experienced dedicated team may help in decreasing sampling errors. Physiologic changes caused by pregnancy may shade malignant transformation in thyrocytes, and it would be appropriate to be cautious about benign FNA taken during this period.

KEYWORDS:

• Fine-needle aspiration (FNA)
• Bethesda cytology
• false negative
• cytology–histology discrepancy
• pregnancy

1. Introduction

The Bethesda System for Reporting Thyroid Cyto-pathology helps to avoid unnecessary surgery in almost 70% of thyroid fine-needle aspirations (FNAs) when a benign interpretation is reported [1,2]. The false-negative rate (FNR) of a benign interpretation is 0–3%, but patients are nevertheless followed up with repeated assessment by ultrasound (US) at regular intervals [2–4]. If nodule shows significant growth or suspicious sonographic changes, a repeat FNA is considered. Furthermore, some recent surgical series with histology follow-ups have suggested that the FNR might be significantly higher in this category of patients [2,5–7].

The inherent nature of thyroid nodules due to overlapping cytologic criteria, over- or underinterpretation of nondiagnostic specimens, misinterpretation of cytopathologists, and variable sampling skills among the institutions and operators may contribute to FNR [8]. To reduce the impact of FNR, complete known risk factor assessment is suggested, including lifestyle (obesity and iodine-deficient diet) and medical history (radiation and genetic predisposition) and a further risk stratification of thyroid nodules with Thyroid Imaging Reporting and Data System (TI-RADS) based on US features [8–12].

In the present study, our purpose was to detect FNR in patients operated for presumably benign thyroid disease and to discuss the feasibility of applying a strategy to decrease this rate. Cytology–histology discrepant cases were also further evaluated to detect the reasons of sampling and interpretation errors and to find possible solutions to decrease FNR. Current risk factors for malignancy were also investigated.

2. Materials and methods

The research protocol was approved by our institution’s Ethics’ Committe (TKH.4.34.H.GP.0.0 l-2023/111). Patients were provided with informed written consent to have data from their medical records to use in this study. Between May 2014 and December 2022, the pathology database of 581 females who underwent surgery for benign thyroid disorders were interrogated retrospectively. All had benign FNA cytology, and standard thyroidectomy was performed by the same surgical oncologist. Demographics, surgical indications, cytology–histology correlations, malignancy rates, and histopathologic subtypes of carcinomas were recorded at Excel file (Office 2017, Microsoft, U.S.).

Cytology–histology discrepant cases were further evaluated for sampling or interpretation errors by a senior cytopathologist. Preoperative US findings were re-evaluated by an online TI-RADS calculator (https://radiogyan.com/tirads-calculator/#tirads-calculator) which is designed by American College of Radiology [13]. Five main criteria (composition, echogenicity, shape, margin, and echogenic foci) were reviewed on the previous images and recorded by a senior radiologist.

All patients were also reevaluated for any overlooked risk factors for thyroid carcinoma, and their charts were reviewed in detail once again for previous radiation exposure, family history, and genetic disorders. US-guided FNAs were performed at our Interventional Radiology Department, by using a 20- to 23-gauge needle with mult-directional passes through the nodule, and the cytological results were evaluated as per the Bethesda classification system. Routine diagnostic criteria for interpretation of the FNA results were used as previously cited [14,15]. Male patients and the patients with Bethesda III to VI categories in their FNA reports were excluded from the study.

Statistical analyses were performed using IBM SPSS Statistics for Windows (Version 19, Armonk, NY). Categorical Fisher’s exact and continuous variable (Mann–Whitney U) tests were used to evaluate associations between predictive variables and thyroid malignancies. P < 0.05 value was accepted as statistically significant.

3. Results

Of 581 patients who underwent thyroidectomy for benign thyroid disorders, 91 was diagnosed as incidental thyroid carcinoma (15.6%) and four of them were above 1 cm in maximum diameter (mean 1.2 ± 0.2 cm, 4.3%). The mean diameter of the remaining micro-carcinomas (T1a) was 4.9 ± 2.7 mm (95.6%, p < 0.05). Twenty-one of 91 carcinomas were multifocal (23%), and in 4 of them, the tumors were in both lobes (19%). In two cases, tumor was seen to invade capsule (2.1%). Final histology reports were papillary (n = 85, 93.4%; classic vs. follicular variants; n = 76 vs. 9; 89.4% vs. 10.5%, respectively) and follicular or medullary carcinomas (n = 6, 6.5%).

Demographics, indications for surgery, and the predictor factors of malignancy were shown in Table 1. Sex, age, and pre-existing thyroid diseases did not differ between patients with and without thyroid carcinomas (p > 0.05, for each). The rate of other carcinoma predictors, such as family history and previous radiation exposure, were also very low (p > 0.05, for each). There was no genetic disorder. Other possible factors including body mass index (BMI), iodine-deficient diet, and diabetes mellitus did not reveal a significant difference, as well (p > 0.05, for each). However, most of the pregnant women who underwent FNA were included in carcinoma group, finally (n = 7 vs. 2, p < 0.05, Table 1).

Table 1. Demographics, indications for surgery, and the predictor factors of malignancy.

	Total	Final benign histology	Incidental thyroid carcinoma	P value
Demographics
Women (n, %)	581 (100)	490 (84.3)	91 (15.6)	-
Age (years, range)	48 (17–78)	47.8 (17–78)	49 (22–65)	0.80
Surgical indications (n, %)
Multinodular goitre	509 (87.6)	430 (87.7)	79 (86.8)	0.69
Giant dominant nodule	32 (5.5)	26 (5.3)	6 (6.5)	0.75
Toxic solitary nodule	22 (3.7)	18 (3.6)	4 (4.3)	0.56
Hyperthyroidism	18 (3)	16 (3.2)	2 (2.1)	0.54
Predictors
Family history (n, %)	12 (2)	10 (2)	2 (2.1)	0.80
Previous radiation exposure (n, %)	3 (0.5)	2 (0.4)	1 (1)	0.75
Genetical disorder	0	0	0	1.00
Body mass index (BMI, kg/m^2)	27.4 ± 3.1	28 ± 4	27 ± 3	0.95
Iodine-deficient diet (n, %)	14 (2.4)	11 (2.2)	3 (3.2)	0.95
Diabetes mellitus (DM) (n, %)	8 (1.3)	6 (1.2)	2 (2.1)	0.75
Pregnancy** (n, %)	9 (1.5)	2 (0.4)	7 (7.6)*	0.02*

*p < 0.05: statistical significance, **fine needle aspiration (FNA) done during pregnancy period.

Re-evaluation of cytology–histology discrepant cases revealed that 41 cases were sampling errors (45%). Re-assessment of preoperative US findings, which were fortified with TIRADS, proved that 25 of these cases were taken from the less suspicious nodules in multinodular glands (27.4%). Others (n = 16) were missed because of missing the core of the lesion while passing the needle through the nodule (17.5%).

Misinterpretation by the cytopathologists was the other common reason for the discrepancies (n = 34, 37.3%). Best identifiable risk factor for misinterpretation was pregnancy as well (n = 5, 14.7%, p < 0.05). In the remaining 16 cases (17.5%), we could not find a reason for discrepancy, and this was linked to the inherent nature of the thyroid nodule biopsied.

Mean TI-RADS scores did not differ between patients with and without thyroid carcinomas (3 vs. 3, respectively, p > 0.05, Table 2). When compared with preoperative US findings, only two patients were upgraded to a more suspicious level (TI-RADS4a, 2.1%) and both had normal histology reports.

Table 2. Thyroid Imaging Reporting and Data System (TI-RADS) scores.

TI-RADS	Definitions	Total	Final benign histology	Incidental malign histology	p value
		(n = 581)	(n = 490)	(n = 91)
1	Normal thyroid gland	0	0	0	1.00
2	Benign nodule	385 (66.2%)	298 (60.8%)	61 (67%)	0.35
3	Highly probable benign nodule	194 (33.3%)	190 (38.7%)	30 (32.9%)	0.75
4a	Low suspicion for malignancy	2 (0.3%)	2 (0.4%)	0	0.45
4b	High suspicion for malignancy	0	0	0	1.00
5	Malignant nodule with more than two criteria of high suspicion	0	0	0	1.00
Mean score		3	3	3	1.00

p < 0.05: statistical significance.

4. Discussion

The risk of malignancy for a benign diagnosis in a thyroid FNA is roughly around 0–3% [1,2]. However, several recent surgical studies have reported substantially higher FNR, ranging from 8% to 14% [2,5–7]. Surgery is often performed in these patients for reasons other than their FNA findings, and these reasons usually include multinodular goitre, nodular enlargement, symptomatic hyperthyroidism, etc. In a study by Richmond et al., lobectomy and near total and total thyroidectomy operations were performed in patients with symptomatic or multinodular goitre and giant solitary nodule [2]. Twenty-four of their 180 patients were found to have a thyroid malignancy on final pathology (13.4%). The authors reported the rates of incidental papillary microcarcinomas three to four times the rate found in classical textbooks. Similarly, Rodrigues et al. correlated between FNA biopsy cytology and pathological examination in 1,093 cases, and they found the prevalence of incidental thyroid carcinoma as 15.1% [16]. The higher FNR in surgical studies suggest that a higher risk of malignancy might be overlooked in these patients with presumably benign thyroid diseases. In the present study, of 581 patients who underwent thyroidectomy for benign thyroid disorders, 91 were diagnosed as thyroid carcinoma (15.6%). This rate obtained from our study was not surprising since the detection of thyroid nodules and cancers has become more common with the widespread use of US in a geography close to Chernobyl exclusion zone [17–20]. Including our findings, significant increase in FNR of Bethesda cytology that is proved by the rise in incidental thyroid carcinomas shown in last two decades’ surgical series made it mandatory to take precautions to avoid delay in cancer diagnosis and treatment.

Since FNR is inevitable to some extent due to the inherent nature of thyroid nodules with overlapping cytologic criteria (17.5% in our series), the only way left as a precaution is getting higher-quality images and reducing sampling errors. The American College of Radiology has recommended a point system for the assessment of imaging thyroid nodules [21–23]. Points are assigned based on five ultrasound features, and the sum determines the TI-RADS classification of the nodule. The 2015 American Thyroid Association (ATA) guidelines classified the nodules into five patterns, from the least (1) to the most (5) suspicious lesions, according to the US features [24]. However, in our study, mean TI-RADS scores did not differ between patients with and without incidental thyroid carcinomas (3 vs. 3, respectively, p > 0.05) and did not help us identify risky nodules in false-negative cases. When compared with preoperative US findings, only two patients were upgraded to a more suspicious level (TI-RADS 4a, 0.79%), and both had normal histology reports. On the other hand, fortifying preoperative US findings with TI-RADS helped us in detecting sampling errors.

Interventional radiology and pathology correlation after biopsy has been found to be useful in validating cytologic results in breast lesions, and any probable discordance has been suggested as an indication for an excision biopsy to guarantee the final diagnosis [25]. Similarly, in our opinion, the correlation of TI-RADS-verified US features with cytologic results after FNA could be an alternative in determining whether the nodule should be rebiopsied to confirm the diagnosis. In our series, TI-RADS study proved that some of the FNAs were taken from the less suspicious nodules in multinodular glands (27.4%). Missing the core of the lesion while passing the needle through nodule was the second reason for sampling errors (17.5%). Reevaluation of cytology–histology discrepant cases also revealed that misinterpretation of FNAs by cytopathologists constituted a significant proportion, following sampling errors (37% and 45% of the cases, respectively). Since the risk of malignancy in a presumably benign thyroid disease is much more than expected, radiology–cytology correlation by an experienced dedicated team may help in decreasing both sampling errors and misinterpretations. Surgeons’ most important contribution to this team work can be warning ahead about the risk factors for malignancy since the patients are first evaluated in outpatient clinics.

Radiation fallout from power plant accidents or childhood exposure to ionizing radiation has been fully recognized as a risk factor for thyroid cancer [19,20]. Advanced age, family history or genetic factors, obesity, diabetes mellitus, and iodine deficient diet are also among the predictive factors for malignancy [26,27]. If any of these factors is positive, clinicians could be more careful about a thyroid nodule or may approach a benign FNA cytology with suspicion. However, detailed investigation of these familial, life style, or environmental risk factors of thyroid carcinoma did not help us explain higher FNR in our series since there was no statistical significance between the patients with and without incidental thyroid carcinomas (p > 0.05, for each). However, FNA taken during pregnancy was determined as a risk factor both in misinterpretation of the cytology (p < 0.05) and in getting a higher FNR (p < 0.05).

Actually, thyroid cancer is the second most common carcinoma diagnosed during pregnancy after breast cancer [28]. The role of female sex hormones as an etiologic factor has been investigated with no clear association [29]. Pregnancy can cause an increase in the size of a previously existed thyroid nodule through the structural similarity between TSH and beta-HCG and the normally expressed estrogen receptors on thyrocytes [28,30]. The effect of pregnancy on development and prognosis of thyroid carcinoma has also been studied [31]. The prognosis of thyroid cancer is not worse in patients diagnosed during pregnancy or those who got pregnant after curative treatment [31,32]. In our opinion, pregnancy by itself is not a predictive risk factor for thyroid malignancy, but it may cause a delay in diagnosis as physiologic changes caused by pregnancy may shade malign transformation in thyrocytes. Therefore, it would be appropriate to be cautious about benign FNAs taken during this period.

This study has three main limitations. The first limitation is the small number of pregnant patients although it meets the criteria for statistical significance. Further multicentric studies with larger sample sizes should be conducted in elucidating possible hormonal effects of pregnancy on FNR with specific focus on misinterpretations and in taking precautions to reduce sampling errors during procedure. The second limitation is the method of sampling causing a selection bias since it includes only one gender. The third limitation is retrospective design and the observational nature leaving the possibility of residual confounding.

Author contributions

Conceptualization: EU AF

Data curation: EU

Formal analysis: AF

Investigation: AF, EU

Methodology: EU

Project administration: EU

Supervision: EU

Validation: AF

Writing – original draft: EU

Writing – review & editing: EU

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

Data from this study are available upon request since there are legal restrictions (by the Ministry of Health) on sharing data publicly. However, we anonymized the patients’ identities and protocol numbers on the system and saved whole data in Excel form; corresponding author will send them by email ([email protected]) on request.

Additional information

Funding

The authors received no specific funding for this work.