Shah MD, Conrad A, Ahmed A, Eski S, MacMillan C, Freeman JL. Decision Making for the Extent of Thyroidectomy in the Patient With Atypical Cytologic Results. Arch Otolaryngol Head Neck Surg. 2010;136(12):1177-1180. doi:10.1001/archoto.2010.207
Copyright 2010 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2010
To identify additional preoperative factors that could reliably be used to aid in determining the appropriate extent of thyroidectomy.
Retrospective chart review.
Tertiary care academic hospital.
Two hundred consecutively treated patients who underwent thyroid surgery after having a fine-needle aspiration biopsy procedure yielding a specimen that met the criteria for atypical cytologic features.
Main Outcome Measure
Final histopathologic diagnosis of malignant vs benign disease.
The final diagnosis was benign in 42.5% of patients and malignant in 57.5%. The presence of microcalcifications within the nodule on ultrasonography (US) was significantly associated with a higher risk of malignant disease (relative risk = 1.31, P = .04). When examined individually, age, sex, family history of thyroid malignant disease, exposure to head and neck irradiation, nodule size, rim enhancement on US, and intranodular vascularity on US were not significantly associated with an increased risk of malignant disease. Mulivariate stepwise logistic regression modeling was used to identify a model that could reliably predict a higher probability of malignant disease. The final model determined that patients with microcalcifications on US and a nodule of 2.0 cm or larger had a 74.3% risk of malignant disease vs a 47.5% risk in patients with no microcalcifications and a nodule smaller than 2.0 cm. This difference was statistically significant. When the predicted probabilities of malignant disease were compared with the observed probabilities, the goodness-of-fit test revealed no significant difference (P = .95).
Microcalcifications and nodule size can be used to risk-stratify patients with an atypical fine-needle aspiration biopsy result and aid in determining the appropriate extent of thyroidectomy.
A US study1 has suggested that the observed increasing incidence of well-differentiated thyroid carcinoma is largely secondary to improved and more widespread use of thyroid imaging techniques. The increasing number of patients diagnosed with thyroid nodules creates a challenge for physicians to determine the most appropriate management of these patients, given that most of their nodules exhibit benign disease.2 The management of thyroid nodules is guided by numerous factors including clinical criteria, findings on ultrasonography (US), and the results of fine-needle aspiration biopsy (FNAB), which is particularly useful in determining when surgery is required and the extent of surgery that is indicated.
No universally accepted classification scheme for FNAB results exists, to our knowledge, although several have been proposed.3 Classically, FNAB results would be classified as benign, indeterminate, suspicious for malignancy, or malignant.2 In our institution, FNAB results are classified according to the number of cytologic features present that are characteristic of papillary thyroid carcinoma (PTC).4 The presence of 2 to 3 characteristic features is reported as “atypical,”5 a diagnosis that would traditionally be classified as indeterminate. In a previous study from our institution, the overall incidence of PTC with an atypical cytologic diagnosis was 56%.5 This creates a management dilemma when deciding on the extent of thyroidectomy to be offered to patients because nearly half will have nonmalignant disease. The objective of this study was to identify additional preoperative factors that reliably could be used to aid in determining the appropriate extent of thyroidectomy.
The Department of Pathology and Laboratory Medicine at Mount Sinai Hospital reviewed 14 477 thyroid FNAB specimens between January 1, 2005, and May 30, 2009. Of these, 6216 (42.9%) were insufficient or unsatisfactory for diagnosis. The remaining 8261 specimens were given the following diagnoses: benign neoplasm/inflammatory (7189 [87.0%]); normal (99 [1.2%]); atypical (583 [7.1%]); suspicious for malignancy (165 [2.0%]); and malignant (225 [2.7%]). A retrospective chart review was performed on consecutively treated patients whose FNAB specimen had been given a diagnosis of being atypical. Approval was obtained from the institutional ethics review board at Mount Sinai Hospital prior to proceeding with data collection or analysis.
The FNAB specimens were categorized as atypical if they met 2 to 3 of the criteria described by Kini et al6 for the diagnosis of PTC. These criteria include syncytial-type tissue fragments, enlarged nuclei with powdery chromatin, multiple micronucleoli and/or macronucleoli, intranuclear cytoplasmic inclusions, and linear chromatin ridges or nuclear grooves.6 The FNAB samples also could be categorized as atypical if they met fewer than 3 criteria but had other types of cytologic atypia. Patients were excluded if they did not undergo surgery at Mount Sinai Hospital or if their chart contained incomplete or insufficient data. The final cohort consisted of 200 patients. The following data were collected: patient age and sex, nodule size, characteristics on US, family history of thyroid cancer, radiation exposure, extent of thyroidectomy, and final histopathologic diagnosis. The specific findings recorded on US were microcalcification, internal hypervascularity, and rim enhancement of nodules. The final diagnosis was classified as benign or malignant.
Each of the preoperative factors examined was investigated for a potential association with final diagnosis of malignant disease. Univariable analysis was conducted using the χ2 test with a significance level set at P < .05. Multivariable analysis was then completed using stepwise logistic regression techniques. Backward and forward stepwise techniques were used and similar results were obtained. All variables were tested in the analysis to derive a model that best fit the data set. The significance level for exclusion or inclusion of variables was set at P < .20 to account for the limitations of stepwise selection procedures in regression modeling.7 Overall model significance was determined using log-likelihood ratio tests, with the significance level set at P < .05. Using this model, we were able to calculate the probability of a diagnosis of malignant disease in a nodule with an atypical FNAB result for different values of the factors in the model. Probabilities were calculated according to equation: probability = e u /(1+e u), where u = [constant + rx(x)+ ry(y)+ . . . ] and r represents the values of the regression coefficients. To determine how well the final model described the observed data, a goodness-of-fit test was conducted; the predicted probabilities were compared with the observed probabilities using a χ2 test. A nonsignificant difference between the observed and predicted data (P > .05) indicates good model fit to the observed data. All statistical analyses were performed using Stata statistical software, version 8.2 (StataCorp LP, College Station, Texas).
Of the 200 patients in the study, 115 (57.5%) had malignant disease, while the remaining 85 patients (42.5%) had benign disease. All the malignant disease was PTC. The mean age of patients was 52 years (range, 22-87) and most (80.0%) were women.
The results of the univariable analysis are summarized in Table 1. Various nodule size categories were investigated. Although there was a trend favoring a higher incidence of malignant disease in larger nodules, it was not statistically significant. The only factor significantly associated with malignancy was microcalcifications observed on US (relative risk, 1.31; P = .04).
The multivariable analysis found that 2 factors were significantly associated with an increased incidence of malignant disease: the presence of microcalcifications on US (regression coefficient, 0.722; odds ratio, 2.06; P = .04) and nodule size of 2.0 cm or larger (0.439; 1.55; P = .12). A multivariable model including these 2 variables best fit the data. Using this model, we were able to calculate the probability of a diagnosis of malignant disease in a nodule with an atypical FNAB result given the presence or absence of either of these factors (Table 2). The risk of malignancy varies between 47.5% if both factors are absent and 74.3% if both are present. When the predicted probabilities of malignant disease were compared with the observed probabilities, the goodness-of-fit test revealed no significant difference (P = .95).
The incidence of malignant disease in this series of patients was 57.5%, which is almost identical to the findings from the previous study5 conducted at our institution examining patients with atypical cytologic results on FNAB testing. This finding creates a dilemma when deciding the extent of surgery to offer such patients. Total thyroidectomy in the setting of benign disease exposes patients to unnecessary intervention and potential morbidity, such as recurrent laryngeal nerve injury and hypoparathyroidism. Conversely, performing a hemithyroidectomy for malignant disease, in most instances, creates the undesirable requirement for a second operation to remove the residual thyroid lobe. This study aimed to identify factors that could assist surgeons and patients in their decision-making processes by identifying those patients who were at highest risk of harboring a malignant nodule.
No universally agreed-upon classification of FNAB results exists, to our knowledge, although several organizations and institutions such as the National Cancer Institute3 have suggested tiered classification schemes. At our institution, meeting all 5 cytologic criteria suggested by Kini et al6 yields a diagnosis of malignant disease. The presence of 4 criteria corresponds to a diagnosis of “suspicious for malignant disease,” while 2 to 3 criteria results in “atypical.”4 More recently, we have adopted the tiered classification scheme recommended by the National Institutes of Health.3 The “atypical” cytologic diagnosis corresponds to the National Institutes of Health category “Follicular lesion/Atypia of undetermined significance.” This category includes cases that are not clearly benign lesions, yet the degree of cellular or architectural atypia is not sufficient to make a diagnosis of follicular neoplasm or “suspicious for malignancy.”3 The incidence of malignant disease in this category was determined to be 5% to 10% on review of the literature,3 which is significantly lower than the incidence of malignant disease in the patients with an atypical FNAB result at our institution. This is likely because the National Institutes of Health category captures a larger group of cytologic diagnoses that do not meet our relatively strict criteria for atypia.
Limited literature exists that examines the incidence of malignant disease in atypical FNAB results. Yang et al8 defined an atypical FNAB specimen as having sheets of follicular cells with microfollicles or macrofollicles or having features of cellular atypia, such as enlarged nuclei or powdery chromatin. The incidence of malignant disease in their series was 13.5%. Kim et al9 used various combinations of nuclear pleomorphism, nuclear grooves, hyperchromatism, powdery chromatin, and intranuclear inclusions to define atypia and found that 32.3% of patients had a diagnosis of malignant disease. Miller et al10 found a much higher incidence of malignant disease, namely, 71%. However, their definition of atypia was not clearly stated. Their study illustrates the difficulty with any comparisons among reports in the literature—varying definitions of “atypical” likely explain much of the wide difference in the reported incidence of malignant disease.
The strong association between microcalcifications on ultrasonographic examination of a thyroid nodule and malignant disease is well recognized.11,12 Thus, it was no surprise that microcalcifications helped to stratify patients with atypical FNAB results. The association between nodule size and the risk of malignant disease is more controversial. Our univariate results suggested a trend toward increasing incidence of malignant disease with increasing nodule size, but this was not significant. However, with the multivariable modeling, the addition of nodule size did improve the statistical fit of the model and helped in the risk stratification of patients.
The results of this study have changed our decision-making processes and the ways we counsel patients with respect to the extent of thyroidectomy that is most appropriate in the setting of an atypical FNAB result. Patients with nodules larger than 2.0 cm and ultrasonographic evidence of microcalcifications are typically recommended to undergo a total thyroidectomy, while a recommendation of hemithyroidectomy is given to patients without these factors. When 1 factor is present, patients are informed that they are at an intermediate risk of malignant disease, and their personal preference often influences the final decision. The results of our study must be interpreted with certain limitations in mind. This series of patients represents only those with an atypical FNAB result who underwent surgery. Thus, by being unable to include those patients, our results may be affected by selection bias. Furthermore, all FNAB results were reviewed at our institution, which manages a very high volume of thyroid disease and uses strict definitions for FNAB results. This may limit the generalizability of our results, and validation at another institution may be useful.
In conclusion, we have determined that the presence of microcalcifications on US and nodule size of 2 cm or larger can be used to stratify patients with an atypical FNAB result. This will help physicians and patients in the decision-making process with respect to the appropriate extent of thyroidectomy.
Correspondence: Jeremy L. Freeman, MD, FRCSC, Department of Otolaryngology–Head and Neck Surgery, Mount Sinai Hospital, 600 University Ave, Ste 401, Toronto, ON M5G 1X5, Canada (JFreeman@mtsinai.on.ca).
Submitted for Publication: March 31, 2010; final revision received May 26, 2010; accepted June 21, 2010.
Author Contributions: Drs Shah and Freeman had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Shah and Freeman. Acquisition of data: Shah, Conrad, Ahmed, Eski, and MacMillan. Analysis and interpretation of data: Shah and MacMillan. Drafting of the manuscript: Shah, Conrad, Ahmed, Eski, and MacMillan. Critical revision of the manuscript for important intellectual content: Freeman. Statistical analysis: Shah. Administrative, technical, and material support: Conrad and Ahmed. Study supervision: Eski and Freeman.
Previous Presentation: This study was presented at the annual meeting of the American Head and Neck Society; April 29, 2010; Las Vegas, Nevada.
Financial Disclosure: None reported.