Ann Surg Oncol DOI 10.1245/s10434-014-4190-8
ORIGINAL ARTICLE – ENDOCRINE TUMORS
A Risk Model to Determine Surgical Treatment in Patients with Thyroid Nodules with Indeterminate Cytology Carlos Aitor Macias, MD, MPH1, Dena Arumugam, MD2, Renee L. Arlow, MD2, Oliver S. Eng, MD2, Shou-En Lu, PhD3, Parisa Javidian, MD3, Tomer Davidov, MD, FACS2, and Stanley Z. Trooskin, MD, FACS2 Department of Surgery, Lenoir Memorial Hospital, Kinston, NC; 2Department of Surgery, Rutgers-Robert Wood Johnson Medical School, New Brunswick, NJ; 3Departments of Biostatistics and Pathology, Rutgers-Robert Wood Johnson Medical School, New Brunswick, NJ 1
ABSTRACT Background. Thyroid nodules are present in 19–67 % of the population and have a 5–10 % risk of malignancy. Fine needle aspiration biopsies are indeterminate in 20–30 % of patients, often necessitating thyroid surgery for diagnosis. We hypothesized that developing a risk model incorporating factors associated with malignancy could help predict the risk of malignancy in patients with indeterminate thyroid nodules. Methods. We identified 151 patients with a cytologic diagnosis of follicular neoplasm (Bethesda IV) who progressed to surgery. We retrospectively analyzed demographic, clinical, sonographic, and cytological variables in relation to thyroid carcinoma. Results. Of 151 patients, 51 (33.8 %) had a final diagnosis of thyroid carcinoma. Papillary carcinoma was diagnosed in 34 patients (66.7 %), follicular carcinoma in 15 (29.4 %), and Hu¨rthle cell carcinoma in 2 (3.9 %). On univariate analysis, younger age, male gender, tobacco use, larger nodule size, and calcifications on ultrasound, nuclear atypia on cytology, and suspicious frozen section were associated with the presence of malignancy. When determining odds ratios, four factors were most predictive of malignancy: nodule calcification [odds ratio (OR) 6.37, 95 % confidence interval (CI) 1.62–25.1, p \ 0.01] and nodule size (OR 1.75, 95 % CI 1.19–2.57, p \ 0.01) on ultrasound, nuclear atypia on cytology (OR 4.91, 95 % CI 1.90–12.66, p \ 0.01), and tobacco use (OR 4.59, 95 % CI
Ó Society of Surgical Oncology 2014 First Received: 15 July 2014 O. S. Eng, MD e-mail: [email protected]
1.30–16.27, p \ 0.02). A multivariable model based on these four factors resulted in a c-statistic of 0.82. Conclusions. A multivariable model based on calcification, nodule size, nuclear atypia, and tobacco use may predict the risk of thyroid cancer requiring a total thyroidectomy in patients with thyroid nodules of indeterminate cytology.
Thyroid nodules are present in 19–67 % of the population and have a 5–10 % risk of malignancy.1 Fine needle aspiration (FNA) is the accepted method for evaluating thyroid nodules.2 Unfortunately, FNA biopsies are indeterminate in 20–30 % of patients, often necessitating thyroid surgery for diagnosis. Several methods have been used to improve the diagnosis of indeterminate thyroid nodules. Some investigators have shown that routine second opinion review of cytopathology results in a different diagnosis in over one-third of FNAs.3,4 Several molecular markers have been investigated for distinguishing benign from malignant thyroid nodules with indeterminate cytology.5 Recently, a commercially available DNA microarray-based genomic classifier assay was shown to diagnose malignancy of indeterminate thyroid nodules with a positive predictive value of 47 % and a negative predictive value of 93 %. The c-statistic of this 142-gene probe set was 0.92.6 While this genomic classifier may gain more widespread acceptance and use, it is an expensive assay, often requires a second FNA, and is currently only available as a send-out lab. Many authors have reported clinical, sonographic, and cytological factors that help stratify high- from low-risk patients with indeterminate cytology.7,8 Male patients or those at the extremes of age, with larger or calcified nodules, patients with a personal history of radiation exposure
C. A. Macias et al.
or a family history of benign thyroid disease, and those with nuclear features suggestive of thyroid carcinoma upon examination of FNAs are considered to be at higher risk.4,7–12 Some authors have tried unsuccessfully to use logistic regression models to predict cancer in patients with indeterminate FNA cytology.11 Others have developed such models using some clinical and cytological parameters.13 The scope of our study was to analyze all the variables available to surgeons treating a patient with a thyroid nodule and follicular neoplasm FNA cytology diagnosis to predict thyroid cancer and thus determine definitive surgical treatment. We used demographic, clinical, sonographic, and cytological features, as well as frozen section analysis, to develop a mathematical model that can be used in surgical planning and patient counseling. METHODS This Institutional Review Board-approved retrospective study was conducted at a Robert Wood Johnson University Hospital (New Brunswick, NJ). Patients with no prior history of thyroid cancer and FNA cytology of follicular neoplasm (Bethesda IV) who underwent thyroidectomy (either lobectomy or total thyroidectomy) between June 2003 and December 2012 were included in the study. We reviewed each patient’s medical records, which included hospital, outpatient clinic, and operating room reports. Patient data collected included age, gender, radiation exposure, history of benign thyroid disease, and family history of thyroid cancer. We also gathered detailed information on tobacco, alcohol, and recreational drug use. Noted sonographic findings included the solid nature and size of nodules, as well as presence of calcifications. In addition to the cytopathologist’s final diagnosis, we also noted the following specific cytological characteristics described in the FNA cytology report: presence of a microfollicular pattern, presence of nuclear atypia, presence of hemosiderin-laden macrophages, predominance of Hu¨rtle cells, and the amount of colloid in the specimen. Frozen-section diagnosis was also included in the model. We routinely offer frozen section to patients with a diagnosis of follicular neoplasm without a contralateral thyroid nodule [1 cm. If the frozen section returns suspicious for cancer, a total thyroidectomy is performed at the initial operation. In our practice, patients with a diagnosis of follicular neoplasm and a contralateral thyroid nodule [1 cm undergo total thyroidectomy. Frozen section was defined as suspicious for either papillary thyroid carcinoma (PTC) by nuclear features (nuclear grooves, pseudonuclear inclusions, nuclear atypia) or suspicious for follicular thyroid carcinoma (FTC) by evidence of capsular or
vascular invasion. In cases where surgical pathology was in question, a second opinion was obtained from Dr. Juan Rosai (Istituto Nazionale Tumori, Milan, Italy). Descriptive statistics were calculated and compared between two groups: patients with thyroid cancer greater than 1 cm, and patients without thyroid cancer greater than 1 cm. v2 tests were used for categorical discrete variables and two-sample t tests for continuous variables. Bivariate analyses using logistic regressions were performed to assess the association of individual risk factors with thyroid cancer. Variables demonstrating a significant association with thyroid cancer (p \ 0.10) were used to build our multivariable logistic regression models based on the back elimination principle. Significant risk factors in the multivariable model were defined by p \ 0.05. Internal validation of the model was performed to evaluate discrimination and calibration. Discrimination measured how well the model separated the two groups. Calibration measured how closely the predicted risk of thyroid cancer was to the observed risk. Specifically, we used the c-statistic, the area under the receiver operating characteristic (ROC) curve, to determine discrimination of the model. In general, a c-statistic [0.7 indicates good discrimination, whereas a value of 0.5 suggests no better discrimination than chance. The calibration of the model was assessed using the Hosmer–Lemeshow goodness-of-fit test.14 A high, nonsignificant p value for this test indicates acceptable calibration. Bootstrap method was used to obtain the optimism statistic to correct for overfitting.15,16 Stata and R packages, RMS and pROC (available on CRAN at http://cran.r-project.org) were used to perform these statistical analyses.
RESULTS We identified 155 patients older than age 18 years who underwent thyroid lobectomy or total thyroidectomy after a fine needle aspiration biopsy with a diagnosis of follicular neoplasm (Bethesda IV cytology). The charts of four patients could not be located and therefore were excluded from the study. In our series of 151 patients, 51 (33.8 %) had a final diagnosis of thyroid cancer in the surgical specimen. Among these 51 patients, FTC was identified in 15 (29.4 %) and Hu¨rthle cell carcinoma was found in 2 (3.9 %). Papillary thyroid carcinoma (PTC) was found in 34 patients (66.7 %), 11 of whom had a follicular variant. The 16 patients who had papillary microcarcinoma (\1 cm in diameter) did not have thyroid cancer warranting total thyroidectomy and thus were included in the noncancer group. We therefore had 35 patients with thyroid
Risk Model Indeterminate Thyroid Nodules TABLE 1 Demographic, lifestyle, and other factors associated with thyroid cancer in all patients, those with thyroid cancer greater than 1 cm and the noncancer group All
Thyroid cancer [1 cm
52.2 ± 14.5 111 (73.5)
47.2 ± 16.3 22 (62.9)
53.7 ± 13.6 89 (76.7)
Demographics Age (years), mean ± SD Female gender, n (%) Ethnicity, n (%)a
\0.02 0.10 0.60
Lifestyle factors Tobacco use, n (%)
Alcohol use, n (%)
Recreational drugs use, n (%)
Radiation exposure, n (%)
Thyroid disease, n (%)
Factors related to thyroid cancer
Family history, n (%) Thyroid cancer
Missing data in 43 patients
TABLE 2 Sonographic and cytological characteristics and frozen-section analysis of thyroid nodules in patients with malignant and benign disease
Thyroid cancer [ 1 cm (n = 35)
Noncancer (n = 116)
Sonographic characteristics Nodule size (mm), mean ± SD
18.1 ± 3.08
12.7 ± 2.00
Solid nature, n (%)
Calcifications, n (%)
Colloid fluid, n (%)
Microfollicular pattern, n (%) Macrophages, n (%) Hu¨rthle cells, n (%)
12 (34.3) 5 (14.3)
52 (44.8) 17 (14.7)
Nuclear atypia, n (%)
Analyses limited to a subset of patients (n = 87) with intraoperative frozen-section analysis (n = 21 vs. n = 66 for thyroid cancer [1 cm vs. noncancer)
Frozen section analysisa Positive, n (%)
carcinoma 1 cm or larger and 116 patients in the noncancer group. Patients with thyroid cancer greater than 1 cm were younger compared with the noncancer group (47.2 ± 16.3 vs. 53.7 ± 13.6, p \ 0.02). In both groups, patients were predominantly females (62.9 vs. 76.7 %, p = 0.10) and Caucasian (80.0 vs. 87.5 %, p = 0.60). Alcohol and recreational drug use were similar in both groups, but patients with thyroid cancer were associated with a trend towards being current or former smokers (22.9 vs. 11.2 %, p = 0.06).
In both groups, patients had a similar history of exposure to radiation, and similar personal and familial history of nonmalignant thyroid disease (Table 1). Patients with thyroid cancer greater than 1 cm were significantly associated with larger nodules (18.1 ± 3.08 vs. 12.7 ± 2.0 mm, p = 0.01) and calcifications (20.0 vs. 6.0 %, p = 0.01) on thyroid ultrasound (Table 2). The presence of nuclear atypia upon FNA cytology was strongly associated with malignant thyroid nodules (51.4 vs. 17.2 %, p \ 0.01). Other cytological descriptors, including
C. A. Macias et al. TABLE 3 Adjusted odds ratios and 95 % CIs of the independent variables in best model for thyroid cancer greater than 1 cm with FNA cytology of follicular neoplasm OR
95 % CI
p value 0.06
Nodule size Calcification
Nuclear atypia Positive frozen sectiona
Analyses limited to a subset of patients (n = 87) with intraoperative frozen section analysis
1 - Specificity Area under ROC curve = 0.8164
FIG. 1 Area under receiver operator curve for the model, including the following variables: age, tobacco use, nodule size, and calcifications on sonographic examination, nuclear atypia on FNA cytology
amount of colloid and the presence of a microfollicular patterns, were not found to be predictive of malignancy. Similarly, the absence of macrophages or predominance of Hu¨rthle was not associated with malignancy (Table 2). Frozen-section analysis was available in 87 patients (57.6 % of patients included in the study). For patients who had intraoperative frozen-section analysis, results were suspicious in 8 (38.1 %) of the 21 patients with thyroid carcinoma, and only in 1 (1.5 %) of 66 patients in the noncancer group. Positive frozen section was associated with an increased risk of thyroid carcinoma greater than 1 cm (p \ 0.01; Table 2). Bivariate analyses suggested that younger age, male gender, tobacco use, lack of family history of benign thyroid disease, greater nodule size and calcifications on ultrasound, as well as the presence of atypical cells on cytology were associated with increased odds of thyroid carcinoma greater than 1 cm. Frozen-section analysis suspicious for cancer also was associated with higher likelihood of cancer on the final specimen.
In the multivariable model, the factors that were statistically significant predictors of malignancy were tobacco use [odds ratio (OR) 4.59, 95 % confidence interval (CI) 1.30–16.27, p \ 0.01], larger nodule size (OR 1.75, 95 % CI 1.19–2.57, p \ 0.01), calcifications on ultrasound (OR 6.37, 95 % CI 1.62–25.1, p \ 0.01), and atypia on FNA (OR 4.91, 95 % CI 1.90–12.66, p \ 0.01). Age was included in the model, because it had marginal significance in our analyses and is a known predictor of thyroid cancer in the literature (OR 0.97, 95 % CI 0.94–1.00).8,9 Adjusted odds ratios are presented in Table 3. The Hosmer–Lemeshow test produced a p value of 0.17, indicating that the multivariable model fits and calibrates well for this patient population. The c-statistic for this multivariable model was 0.82 (Bootstrap corrected c-statistic = 0.80, optimism = 0.02), suggesting very good discrimination (Fig. 1). Including frozen section analysis in our model yielded a c-statistic of 0.89 (bootstrap corrected c-statistic = 0.86, optimism = 0.03), which is higher than that of the model without frozen-section analysis. However, because of missing data, such an increase was not statistically significant (p = 0.142). The following mathematical equation was produced from the logistic regression analysis to predict the presence of thyroid carcinoma requiring thyroidectomy in patients with follicular neoplasm cytology. The probability is denoted by p: logitðpÞ ¼ 1:83 0:029 age ðin yearsÞ þ 1:52 tobacco smoking ð1 if present; 0 if notÞ þ 0:56 largest nodule (sonographic nodule size in centimeters) þ1:85 calcifications on ultrasound ð1 if present; 0 if notÞ þ 1:59 atypical cells in FNA ð1 if present; 0 if notÞ DISCUSSION In the current Bethesda Classification of thyroid cytology, 5–42 % of thyroid FNAs are indeterminate and often require a repeat FNA or diagnostic thyroid surgery.17 Indeterminate FNAs comprise two Bethesda Classification categories: follicular lesions (Bethesda III), which account for 3–6 % of thyroid FNAs and carry a 5–15 % risk of malignancy, and follicular neoplasms (Bethesda IV), which account for 11–20 % of thyroid FNAs with a 20–30 % malignancy risk. In this study, we focused only on the more uniform, prevalent and high-risk of the two: the Bethesda IV category.18,19
Risk Model Indeterminate Thyroid Nodules
Current American Thyroid Association guidelines recommend thyroid lobectomy or total thyroidectomy for patients with a diagnosis of follicular neoplasm on cytology. If papillary or FTC is diagnosed on surgical pathology, further surgery in the form of completion thyroidectomy often is required.1 In patients undergoing surgery because of indeterminate thyroid nodules, the use of intraoperative frozen section is still controversial.20,21 Despite sampling error and freeze artifact, intraoperative frozen-section analysis, along with intraoperative touch prep for better nuclear visualization (the cut edge of the surgical specimen is touched onto a slide to allow better characterization of nuclear features otherwise obscured by freeze artifact on a frozen specimen), may provide surgeons with additional information necessary to select total thyroidectomy instead of lobectomy at the initial operation.22,23 In our series of 151 patients with a diagnosis of follicular neoplasm, 51 (33.8 %) patients had cancer. Consistent with existing data, papillary carcinoma was the most common type of cancer.8,10,19,22 We chose to include papillary microcarcinomas (defined as papillary carcinomas smaller than 1 cm) in the noncancer group, because these are lesions that are not typically aspirated preoperatively, and therefore, often are considered an incidental finding on surgical pathology. More importantly, current recommended treatment for papillary microcarcinomas is thyroid lobectomy without the need to undergo subsequent completion thyroidectomy.1 The four variables that we found to be most predictive of thyroid cancer in patients with FNA cytology of follicular neoplasm were presence of nuclear atypia on FNA, nodule size, presence of calcifications on ultrasound, and tobacco use. Calcifications and atypia have been found to be strongly associated with malignant thyroid nodules in most series.4,7,10,13,22 Data regarding nodule size is conflicting. Some authors have shown that larger nodule size is suggestive of malignancy, whereas others have shown that benign nodules tend to grow larger than malignant ones.9,12,13,24 Tobacco is the leading cause of cancer-related deaths in the world.25 However, it is generally associated with a reduced thyroid cancer risk.26 In this study, patients that smoked tobacco had a significant and robust increased risk of thyroid cancer. We can only speculate about this result, but tobacco smoking may well be associated with thyroid cancer in this particular group of patients with Bethesda IV cytology. Finally, in our models, the addition of frozen-section analysis increased the discriminatory power, with an AUC of 0.89 for the model predicting thyroid cancer [1 cm. We acknowledge several limitations of this study. As with any retrospective cohort study, we can demonstrate
associations but not causal relationships between the examined factors and the development of thyroid malignancies. The multivariable model performed well and has shown to be robust; however, we could not perform an external validation in other populations. We did use the bootstrap method to correct for the expected value of the external predictive discrimination. Prior studies have attempted to build a model to predict cancer in this group of patients. A recent study by Lubitz et al. utilized logistic regression analyses to find that nodule size greater than 4 cm, nuclear grooves, and transgressing vessels in FNA are significant predictors of thyroid cancer in patients with follicular neoplasm cytology. However, in this study, most demographic and clinical descriptors, as well as ultrasound findings and frozen-section analysis, were not included.13 Other studies have investigated the potential role of molecular assays of indeterminate FNA specimens in identifying associations with malignancy. A recent study at our institution utilized immunohistochemistry to analyze DNA microarray assays in specimens of patients who underwent thyroidectomy for indeterminate thyroid nodules and either had a postoperative diagnosis of FTC or follicular adenoma. Results showed FTC specimens to be associated with a loss of carbonic anhydrase 4 and an increase in crystallin alpha-B immunoreactivity in follicular carcinoma samples versus adenoma. Future studies involving gene and molecular testing on FNA specimens are needed to continue to investigate such associations.27 In conclusion, we have developed a novel tool to be used for patient counseling and surgical planning when treating a patient with a thyroid nodule and follicular neoplasia cytology. This model does not require any special equipment or additional tests, because it is based on variables readily available and routinely used to make the diagnosis in patients with thyroid nodules. For example, using the equation of our model we can determine the risk (probability) of thyroid cancer in a 20year-old patient, smoker, who has an 18-mm thyroid nodule with calcifications on US and nuclear atypia on FNA cytology. From our equation, first we calculate the prognostic index (L). Thus L = 1.83 - 0.029 9 20 ? 1.52 9 1 ? 0.56 9 Ln(18) ? 1.85 9 1 ? 1.59 9 1 = 7.83. Then, the risk (probability) of having thyroid cancer (p) is calcuL lated: p ¼ e =1 þ eL ¼ 0:99. The risk of thyroid cancer for this particular patient is nearly 1, and thus warrants a total thyroidectomy as the initial operation. Improving the diagnosis of thyroid nodules with indeterminate cytology continues to be a challenge. This model can potentially provide a useful tool for predicting thyroid cancer and planning surgical treatment. DISCLOSURE
C. A. Macias et al.
REFERENCES 1. American Thyroid Association Guidelines Taskforce on Thyroid N, Differentiated Thyroid C, Cooper DS, et al. Revised American Thyroid Association management guidelines for patients with thyroid nodules and differentiated thyroid cancer. Thyroid. 2009;19(11):1167–214. 2. Gharib H, Goellner JR. Fine-needle aspiration biopsy of the thyroid: an appraisal. Ann Intern Med. 1993;118(4):282–9. 3. Davidov T, Trooskin SZ, Shanker BA, et al. Routine secondopinion cytopathology review of thyroid fine needle aspiration biopsies reduces diagnostic thyroidectomy. Surgery. 2010;148(6): 1294–9; discussion 1299–301 4. Chernyavsky VS, Shanker BA, Davidov T, et al. Is one benign fine needle aspiration enough? Ann Surg Oncol. 2012;19(5): 1472–6. 5. Nikiforov YE, Yip L, Nikiforova MN. New strategies in diagnosing cancer in thyroid nodules: impact of molecular markers. Clin Cancer Res. May 1 2013;19(9):2283–2288. 6. Alexander EK, Kennedy GC, Baloch ZW, et al. Preoperative diagnosis of benign thyroid nodules with indeterminate cytology. N Engl J Med. 2012;367(8):705–15. 7. Kelman AS, Rathan A, Leibowitz J, Burstein DE, Haber RS. Thyroid cytology and the risk of malignancy in thyroid nodules: importance of nuclear atypia in indeterminate specimens. Thyroid. 2001;11(3):271–7. 8. Tyler DS, Winchester DJ, Caraway NP, Hickey RC, Evans DB. Indeterminate fine-needle aspiration biopsy of the thyroid: identification of subgroups at high risk for invasive carcinoma. Surgery. 1994;116(6):1054–60. 9. Baloch ZW, Fleisher S, LiVolsi VA, Gupta PK. Diagnosis of ‘‘follicular neoplasm’’: a gray zone in thyroid fine-needle aspiration cytology. Diagn Cytopathol. 2002;26(1):41–4. 10. Goldstein RE, Netterville JL, Burkey B, Johnson JE. Implications of follicular neoplasms, atypia, and lesions suspicious for malignancy diagnosed by fine-needle aspiration of thyroid nodules. Ann Surg. 2002;235(5):656–62; discussion 662–4. 11. Koike E, Noguchi S, Yamashita H, et al. Ultrasonographic characteristics of thyroid nodules: prediction of malignancy. Arch Surg. 2001;136(3):334–7. 12. Raber W, Kaserer K, Niederle B, Vierhapper H. Risk factors for malignancy of thyroid nodules initially identified as follicular neoplasia by fine-needle aspiration: results of a prospective study of one hundred twenty patients. Thyroid. 2000;10(8):709–12. 13. Lubitz CC, Faquin WC, Yang J, et al. Clinical and cytological features predictive of malignancy in thyroid follicular neoplasms. Thyroid. 2010;20(1):25–31.
14. Hosmer DW, Lemeshow LS. Applied logistic regression. Hoboken: Wiley; 2000. 15. Efron B GG. A leisurely look at the bootstrap, the jackknife, and cross-validation. Am Stat. 1983;37(1):36–48. 16. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15(4):361–87. 17. Cibas ES, Ali SZ. The Bethesda system for reporting thyroid cytopathology. Thyroid. 2009;19(11):1159–65. 18. Broome JT, Solorzano CC. The impact of atypia/follicular lesion of undetermined significance on the rate of malignancy in thyroid fine-needle aspiration: evaluation of the Bethesda system for reporting thyroid cytopathology. Surgery. 2011;150(6):1234–41. 19. Krane JF, Vanderlaan PA, Faquin WC, Renshaw AA. The atypia of undetermined significance/follicular lesion of undetermined significance: malignant ratio: a proposed performance measure for reporting in The Bethesda system for thyroid cytopathology. Cancer Cytopathol. 2012;120(2):111–6. 20. Grannan K, Snyder J, McDonough S, Engel A, Farnum J. Operative decision-making for follicular thyroid lesions: a community hospital system experience. Am Surg. 2011;77(4):443–6. 21. Lumachi F, Borsato S, Tregnaghi A, et al. FNA cytology and frozen section examination in patients with follicular lesions of the thyroid gland. Anticancer Res. 2009;29(12):5255–7. 22. Liu FH, Liou MJ, Hsueh C, Chao TC, Lin JD. Thyroid follicular neoplasm: analysis by fine needle aspiration cytology, frozen section, and histopathology. Diagn Cytopathol. 2010;38(11): 801–5. 23. Yoon JH, Kwak JY, Kim EK, et al. How to approach thyroid nodules with indeterminate cytology. Ann Surg Oncol. 2010;17(8): 2147–55. 24. Miller B, Burkey S, Lindberg G, Snyder WH 3rd, Nwariaku FE. Prevalence of malignancy within cytologically indeterminate thyroid nodules. Am J Surg. 2004;188(5):459–62. 25. Oppeltz RF, Jatoi I. Tobacco and the escalating global cancer burden. J Oncol. 2011. doi:10.1155/2011/408104. 26. Mack WJ, Preston-Martin S, Dal Maso L, et al. A pooled analysis of case-control studies of thyroid cancer: cigarette smoking and consumption of alcohol, coffee, and tea. Cancer Causes Control. 2003;14(8):773–85. 27. Davidov T, Nagar M, Kierson M, et al. Carbonic anhydrase 4 and crystallin alpha-B immunoreactivity may distinguish benign from malignant thyroid nodules in patients with indeterminate thyroid cytology. J Surg Res. 2014;190(2):565–74.