Diagnostic accuracy between fine needle aspiration biopsy (FNA) and pathological examination of the thyroid

Laboratorial ISSN: 1676-2444 jbpml@sbpc.org.br,adagmar.andriolo@g mail.com Sociedade Brasileira de Patologia Clínica/Medicina Laboratorial Brasil Menegassi, Jerso; Bordin, Vanessa; Santos Oliveira, Saimon Diagnostic accuracy between fine needle aspiration biopsy (FNA) and pathological examination of the thyroid Jornal Brasileiro de Patologia e Medicina Laboratorial, vol. 49, núm. 4, agosto, 2013, pp. 288-293 Sociedade Brasileira de Patologia Clínica/Medicina Laboratorial Rio de Janeiro, Brasil


introDuCtion
The nodular thyroid disease is characterized by an abnormal growth of a nodule clinically evident in the gland, causing structural and/or functional alterations. The etiology of the disease is multifactorial, comprising a spectrum from a small nodule found incidentally to a large intrathoracic multinodular goiter (7) .
Thyroid nodules, the main clinical manifestation of a number of thyroid diseases, are commonly benign and frequent lesions. Studies show the prevalence of approximately 4% of palpable thyroid nodules in the general population, of which only 5% are malignant (15) .
Currently, there is some research into new ways to reach an accurate diagnosis. Despite the use of common procedures such as gland palpation and modern diagnostic imaging, a new sensitive and specific method is still required in order to ensure cost effectiveness and avoid patient discomfort.
The fine needle aspiration biopsy (FNA) is one of the diagnostic methods of thyroid lesions, consisting of the cytopathological study of morphological alterations in isolated cells obtained by the introduction of a needle into the lesion. FNA guided by ultrasound scan (USS) is the initial method for the assessment of patients with thyroid nodules, with the objective to identify malignant nodules, differentiating them from benign lesions (14) .
The use of FNA has reduced the number of thyroidectomies by approximately 50%, while it increased the diagnosis of thyroid cancer (8) . Nevertheless, FNA cannot characterize malignant or benign follicular lesions, inasmuch as it is not possible to analyze the capsule of the lesion, which is a determinant factor in the characterization of malignancy.
At present, in cases of suspected malignancy or malignancy assessed by FNA, surgery is recommenced. Intraoperative pathological examinations may be performed during surgery through macroscopic and microscopic study of the surgical specimen with the aim to establish a final diagnosis. The pathological exam plays a pivotal role in the definition of risk after surgery (9) .

oBJECtivE
The objective of this study was to determine the diagnostic accuracy of FNA in relation to the pathological examination of thyroid nodules, both procedures performed in a pathology laboratory in Chapecó, Santa Catarina, Brazil.

MEthoDS
We conducted a retrospective study in a reference center for diagnostic pathology in Chapecó-SC by comparing cytology and histopathology (gold standard) from FNA. Data were obtained from the internal information system of procedures performed at Instituto de Patologia do Oeste. This paper was submitted to the analysis of the Committee of Ethics in Research-Unochapecó on August 12, 2010, meeting the criteria defined in Resolution 196/96 as well as its complements from the National Health Council of the Ministry of Health and being registered under the number 115/10.
We included in the study those patients who had undergone both FNA and pathological examination of the thyroid at Instituto de Patologia do Oeste. The selected cases were supposed to have cytopathological and pathological reports on file. Patients who had undergone only one of the procedures were excluded from the sample.  It was observed that 85% of patients were female, with a sex ratio (female: male) of 5.46:1. The mean age was 47.75 years (standard deviation [SD] = 14.93 years), with median age of 48 years (and variation of 2-83 years). The age group with predominant diagnosis was 50-60 years (27.9%).
The FNA findings were the following: benign (43.77%), suspected of malignancy (23.77%), follicular neoplasm (25.28%) and atypical lesions of undetermined significance (1.51%). The sample was unsatisfactory in 1.51% of analyzed biopsies ( Table 1). By correlating the types of surgical resection that were performed among the patients, it was noted that total thyroidectomy was the most frequent type of surgery (53%), followed by partial thyroidectomy (45%).
By analyzing the pathological examinations, it was observed that 69% of the findings were benign, of which the colloid goiter was the most frequent pattern, accounting for 57% of benign lesions, followed by follicular adenoma (33%), normal parenchyma and thyroiditis (3% each), hyperplasia and Hurthle cell adenoma (2% each) ( Table 2). The most prevalent malignant pattern was the classic papillary carcinoma, accounting for 60% of total malignant lesions, followed by follicular variant of papillary carcinoma and follicular carcinoma (11% each), medullary carcinoma (9%), other types of papillary carcinoma (5%), anaplastic carcinoma and a poorly differentiated carcinoma (2% each) ( Table 3).
Diagnostic accuracy between fine needle aspiration biopsy (FNA) and pathological examination of the thyroid  Considering only patients with satisfactory samples, it was possible to verify that the diagnostic accuracy was 89.66%,

DiSCuSSion
FNA is a well established diagnostic technique in the selection of patients with thyroid nodules for surgical treatment, reducing the number of unnecessary surgical procedures in cases of nonmalignant neoplasms (1,2,6) .
Herein, we deployed the cytological classification recommended by the National Cancer Institute of the United States (2008), therefore the FNA findings were included into six different categories. Thus, it was observed that most of these nodules were benign (43.77%), which corroborates literature findings, followed by follicular neoplasia (25.28%), suspected malignancy (23.77%), atypical lesions of undetermined significance and unsatisfactory sample (1.51% each). In a study conducted in the United States in 2007, which used six different possible classifications for FNA results, 64.6% of the findings were benign, followed by sample with undetermined significance (11.6%), unsatisfactory sample (10, 4%), malignant (7.6%), atypical lesions (3.2%) and suspected malignancy (2.6%). The follicular lesions were placed in the "atypical lesions" category. Correlating these data, we observed a higher prevalence of malignant findings and suspected malignancy in the present study by contrasting the respective prevalences detected in the cited study (16) .
In a Brazilian study from 2003, 44.6% of the findings were considered benign, 16.2% malignant, 29.2% undetermined, and 10% unsatisfactory sample. Herein, the follicular lesion was classified along with undetermined sample. By adding data on follicular neoplasm and atypical lesion of undetermined significance, a total of 26.76% was achieved (2) . This difference in classification in various studies may be due to the difficulty in characterizing benign or malignant follicular lesion by FNA. Consequently, some authors prefer to consider it undetermined sample, others apply the follicular neoplasia classification or even classify it as suspected malignant lesion.
The differential diagnosis between follicular adenoma and follicular thyroid carcinoma is hindered owing to the cytological similarity of both lesions and the impossibility of verifying vessel and/or capsule invasion through FNA, which are ultimately the criteria for determining malignancy. The use of the terms "follicular lesion" or "follicular standard" is recommended for these nodules, thus the classification "follicular neoplasia" was adopted herein (4) .
As to the pathological findings, there was a prevalence of benign lesions (69%) and malignant lesions (31%) in the present study, consistent with the yielded results from previously mentioned studies. Regarding benign pathological findings, the colloid goiter was the most prevalent diagnosis, accounting for 57% of 182 patients belonging to this group, followed by follicular adenoma (33%), hyperplasia, thyroiditis and normal parenchyma, which accounted for 8% of patients. These results were similar to those yielded at a study carried out in 2001 in a hospital in Porto Alegre-RS (12) .
It is known that the papillary carcinoma is the most common form of thyroid cancer, accounting for about 70% of cases. Moreover, it has the best prognosis in comparison with other thyroid malignant neoplasias. In our study, the classic papillary carcinoma appeared as the most prevalent, corresponding to 60% of total malignant pathological findings, followed by the follicular variant of papillary carcinoma and follicular carcinoma (11% each) as well as other types of papillary carcinoma (5%) (3) .
As to diagnostic values, the diagnostic accuracy was 89.66% and this value was situated within the variation window in comparison with the results obtained in previous studies. Based on a 2002 study, sensitivity values for FNA ranged from 65% to 99% and specificity from 72% to 100%. Herein, sensitivity and specificity for detecting neoplasias by FNA corresponded to 82.14% and 93.22%, respectively (10) .
Furthermore, it was observed that the specificity value was higher than the sensitivity value, hence FNB is a reliable predictor of benign thyroid pathology. It is known that the specificity demonstrates the ability of a test to distinguish true negatives in relation to the total number of patients (5) .
The way pathologists categorize suspicious lesions and how they define the false-positives and negatives account for this wide variation in sensitivity and specificity results. Although some authors include follicular lesions in the malignancy category, some characterize them in the negative group and others even exclude follicular lesions from the general calculation (1) .
A variable that may have contributed to the accuracy results is the participation of both a pathologist conducting FNA and a radiologist guiding the procedure via USS. Although there are few studies comparing the performance of USS-guided FNA and conventional approach (guided by palpation), they concur that the association with USS improves overall accuracy (13) .
Regarding NPV and PPV, an individual had 91.67% probability of not having malignant thyroid lesions when the cytopathological examination indicated benign alteration and 75.19% probability of presenting a malignant lesion when the cytological examination revealed malignant alteration.
It is claimed that the accuracy of a diagnostic test is contingent on the reliable distinction between benign and malignant findings. A widespread tool to assess the performance of diagnostic tests is the analysis of Receiver Operating Characteristic curve (ROC), a direct visual representation of a test performance. The accuracy is measured by the area under the ROC curve. An area equal to 1 represents a perfect test, that is to say that all diagnoses have to be accurate. An area equal to 0.5 represents an unreliable test, inasmuch as the simple flip of a coin would lead to a curve close to identity line. Therefore, the higher the value (0.5 >), the better the test classification according to its performance. The curve demonstrated in this investigation ( Figure) is consistent with previous studies, comprising an area of 0.876 (11) .  Taking into account the importance of the interpretation of ROC curves, it is possible to choose the best clinical and surgical procedures based on cutoff points, providing greater safety for both doctor and patient. Nevertheless, even with the availability of these complementary diagnostic tools, it is imperative to address the need for further studies that will add more specific criteria for determining malignant nodules, hence improving the accuracy of FNA, which is deemed to be a cost-effective option.