3306jum929-1084_Layout 1 5/21/14 8:56 AM Page 1023
ORIGINAL RESEARCH
Development of a Logistic Regression Formula for Evaluation of Subcentimeter Thyroid Nodules Mingbo Zhang, MD, Yan Zhang, MD, Shuai Fu, MD, Faqin Lv, MD, Jie Tang, MD
Objectives—The purpose of this study was to build a logistic regression formula for ameliorating the diagnosis of subcentimeter thyroid nodules. Methods—The sonographic features of 889 subcentimeter nodules were reviewed retrospectively with reference of histologic results. The diagnostic performance of each feature was evaluated. Multivariate binary logistic regression was used to develop the formula for evaluation of subcentimeter nodules, and the cutoff value was decided for recommending biopsy. Results—The logistic regression formula was –0.029age – 2.063US1 – 0.812US2 + 1.781US3 + 1.627height-to-width ratio – 0.333 (for nonhypoechogenicity, US1 = 1; for hypoechogenicity, US1 = 0; for a well-circumscribed margin, US2 = 1 and US3 = 0; for a microlobulated margin, US2 = 0 and US3 = 1; and for an irregular margin, US2 = 0 and US3 = 0). The area under the receiver operating characteristic curve for the formula was 0.860. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of the formula under the cutoff value of 0.284 were 90.9%, 54.0%, 66.4%, 85.6%, and 72.5%, respectively. With the use of the formula, 229 of 830 nodules could avoid surgery. Conclusions—A logistic regression formula with a cutoff value could provide an objective and easy tool with effective diagnostic performance, which could improve diagnosis of subcentimeter thyroid nodules and reduce unnecessary biopsy, decreasing costs and patient discomfort. Key Words—biopsy; nodule; sonography; superficial structures; thyroid
Received August 1, 2013, from the Department of Ultrasound, General Hospital of Chinese PLA, Beijing, China. Revision requested August 20, 2013. Revised manuscript accepted for publication October 7, 2013. Address correspondence to Jie Tang, MD, Department of Ultrasound, General Hospital of Chinese PLA, 28 Fuxing Rd, Haidian District, 100853 Beijing, China. E-mail: [email protected] Abbreviations
NPV, negative predictive value; OR, odds ratio; PPV, positive predictive value doi:10.7863/ultra.33.6.1023
T
hyroid nodules are commonly seen in daily practice. They are detected in 4% to 7% of patients by palpation and 50% at autopsy.1,2 Recently, the use of high-frequency sonography has greatly increased the detection of subcentimeter (maximum diameter, ≤1 cm) thyroid nodules. Although the clinical importance of subcentimeter nodules for malignancy is widely debated,3–7 studies suggest that their malignant prevalence is similar to that of palpable nodules, and their malignant involvement is not less frequent.4,8,9 The American Thyroid Association guideline does not recommend fine-needle aspiration of sonographically suspicious subcentimeter nodules unless there is associated high-risk history.10 However, other organizations and authors suggest that fine-needle aspiration biopsy should also be performed on subcentimeter nodules with suspicious sonographic findings.8,11 Uncertainty from not knowing may adversely affect the quality of life in these patients.11
©2014 by the American Institute of Ultrasound in Medicine | J Ultrasound Med 2014; 33:1023–1030 | 0278-4297 | www.aium.org
3306jum929-1084_Layout 1 5/21/14 8:56 AM Page 1024
Zhang et al—Logistic Regression Formula for Subcentimeter Thyroid Nodules
Thus, detecting suspicious sonographic features of subcentimeter nodules is worthwhile. Many studies have been performed for diagnosis of thyroid carcinoma using sonographic features. However, the features have varying abilities to predict malignancy,8,9,12–21 and most studies included nodules with various sizes, not only subcentimeter nodules. Sharma et al9 suggested the importance of the clinician’s overall impression to the malignancy of subcentimeter nodules, not just a single feature. Thus, our study tried to improve the diagnosis of subcentimeter thyroid nodules with a comprehensive solution, which is better for saving medical resources and reducing patient discomfort. In our study, the sonographic features of a large sample of subcentimeter nodules were retrospectively analyzed with reference to histologic results to develop a logistic regression formula for improving the diagnosis. In addition, a cutoff value was determined for recommending biopsy to reduce unnecessary biopsy and to decrease costs and patient discomfort.
Materials and Methods Patients Our study was approved by the Institutional Review Board of our hospital, and the requirement for informed consent was waived. Subcentimeter thyroid nodules with histologic results from sonographically guided core needle biopsy, thyroidectomy, or both at our institution between April 2009 and April 2012 were reviewed. Our institution is a tertiary health center, and patients from all over the country seek services here. During this period, 2172 thyroid nodules received core needle biopsy, and 889 of them were subcentimeter nodules. No patient had more than 1 core needle biopsy during this period. Among the 889 nodules, 59 were excluded, including 11 inadequate core needle biopsy samples, 25 atypia of undetermined importance (atypical cells that presented without a diagnosis of suspicion for malignancy), 15 follicular neoplasms, and 8 nodules that were suspicious for malignancy (atypia with insufficient evidence for a definite diagnosis of malignancy), which did not receive surgery. Finally, 830 nodules in 788 patients were included in this study, including 197 nodules receiving only surgical resection, 563 nodules receiving only core needle biopsy, and 70 nodules receiving both at our institution. Ages and sexes of the patients were recorded. A flowchart of the study group is presented in Figure 1. Subcentimeter nodules were subjected to core needle biopsy if one of the following suspicious findings was present: hypoechogenicity, calcification, an irregular or
1024
microlobulated margin or no halo, intranodular vascularity, a taller-than-wide shape, sonographic features highly indicating metastatic cervical lymph nodes, and substantial growth (20% increase in 2 minimum dimensions, at least 2 mm) since a previous sonographic examination within 18 months.10 For patients undergoing surgery at our institution, the final diagnosis was made by surgical histologic reports. Otherwise, core needle biopsy histologic results were used instead. At the Department of Pathology of our institution, the core needle biopsy and thyroidectomy specimens were routinely examined by a single boardcertified pathologist and reviewed by a single chief pathologist with more than 10 years of experience in thyroid pathologic examinations. If there were more than 2 nodules, the physician who performed the core needle biopsy or thyroidectomy did the radiologic-pathologic correlation to ensure concordance. Imaging Analysis The examinations before core needle biopsy were used for the analysis. The sonograms in our study were obtained by board-certified sonographers using highfrequency linear probes (L5-17 MHz) and iU22 (Philips Healthcare, Bothell, WA) and LOGIQ 9 (GE Healthcare, Milwaukee, WI) ultrasound systems. Images were stored in our picture archiving and communication system, and the sonographic features of the nodules were reevaluated by 2 sonographers (M.Z. and Y.Z.) with more than 5 years of experience in thyroid examinations, who were blinded to each other during the evaluation process. If there were disagreements about the features, the sonographers reached a consensus after discussion with an expert (F.L.) Figure 1. Flowchart of the study group for developing the multivariate regression formula. CNB indicates core needle biopsy.
J Ultrasound Med 2014; 33:1023–1030
3306jum929-1084_Layout 1 5/21/14 8:56 AM Page 1025
Zhang et al—Logistic Regression Formula for Subcentimeter Thyroid Nodules
who had more than 15 years of experience in thyroid examinations. All of them were blinded to the histopathologic diagnosis. On sonography, the thyroid nodules were described according to their echogenicity, internal component, margin, calcification, halo, size, and vascularization. Echogenicity was classified as hyperechoic, isoechoic, hypoechoic, or anechoic. When the echogenicity of the nodule was similar to that of the thyroid parenchyma, it was classified as isoechoic. The nodule was classified as hypoechoic or hyperechoic if the echogenicity was less than or more than that of the thyroid parenchyma. The internal component of the nodules was classified as solid, mixed solid, or cystic. The margin was classified as well circumscribed, microlobulated, or irregular. A microlobulated margin was defined as the presence of many small lobules on the nodular surface. Calcification, when present, was categorized as microcalcification or macrocalcification. Microcalcification was defined as calcification of 1 mm or smaller in diameter and visualized as tiny punctuate hyperechoic foci, either with or without acoustic shadows. Tiny bright reflectors with a clear-cut comet tail artifact were considered colloid.22 Macrocalcification was defined as hyperechoic foci of 1 mm or larger. When a nodule had both types of calcification (macrocalcification, including rim calcification, intermingled with microcalcification), we regarded it as having microcalcification. A halo was categorized as halo or no halo. Height was defined as the anteroposterior diameter, and width was defined as the larger of the transverse or longitudinal diameter. Width and height were measured, and the ratio of height-to-width ratio was calculated. Vascularization was categorized as no blood flow, internal blood flow, or peripheral blood flow.12 Sonographically Guided Core Needle Biopsy Procedures Sonographically guided core needle biopsy procedures were performed by experienced interventional sonographers using a Magnum biopsy instrument and a 1.5- or 2.2-cm excursion double-action spring-activated 18-gauge needle (Bard Medical Division, Covington, GA). A high-frequency linear probe with a needle guide line and matched guiding frame was used for accurate positioning of the needle track. After induction of local anesthesia with 1% lidocaine, the core needle was inserted in the direction of the guide line. Color Doppler imaging was used for evaluation of vessels along the approach route to avoid hemorrhage. After the tip of the biopsy needle had been advanced into the edge of the nodule, the firing distance was measured to minimize vessel injury for safety. The stylet and cutting cannula of the needle were sequentially fired. Two core biopsy
J Ultrasound Med 2014; 33:1023–1030
passes were made for each nodule. The puncture positions were pressed as soon as the biopsy passes were finished in case of hematoma, which lasted at least 30 minutes. Statistical Analysis Descriptive data were presented as mean ± standard deviation or number (percent). The t test or χ2 test was used to determine the sonographic features and clinical information between benign and malignant nodules. Since hypoechogenicity and a solid internal component are two suspicious malignant features,8,10 the echogenicity and echo structure were dichotomized as hypoechoic/nonhypoechoic and solid/nonsolid during the evaluation process to determine diagnostic performance. The odds ratio (OR) of each sonographic feature was analyzed. If there were more than 2 subgroups of features, the least suspicious (lowest malignant rate) feature was used as the reference, and the ORs of the other suspicious features were analyzed and compared with reference to it. The diagnostic performances of sonographic features were evaluated, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy. Multivariate binary logistic regression with the forward conditional method was used to develop the formula for recommending sonographically guided biopsy. The diagnostic value of the formula was evaluated by a receiver operating characteristic curve. The cutoff value for recommending biopsy and its corresponding diagnostic values were evaluated. Statistical analysis was performed with SPSS version 13.0 software (IBM Corporation, Armonk, NY). P < .05 was considered statistically significant.
Results A total of 830 subcentimeter nodules (425 benign and 405 malignant) in 789 patients (618 female and 171 male) were included in our study. The mean size (longest diameter) of the nodules ± SD was 0.69 ± 0.20 cm (range, 0.20–1.00 cm). The mean age of the patients was 48.1 ± 10.8 years (range, 12–83 years). Most malignant nodules (96.8%) were papillary carcinoma. The pathologic diagnoses of the 830 nodules are listed in Table 1. Patients with malignant nodules were significantly younger than those with benign nodules (P < .001). The height of malignant nodules was significantly larger (P < .001) and their width was significantly smaller (P = .049) than those of benign nodules. Malignant nodules had a greater chance of being solitary than benign nodules (P = .003). No significant difference in malignancy was found between male and female patients. The clinical features of the patients and nodules are listed in Table 2.
1025
3306jum929-1084_Layout 1 5/21/14 8:56 AM Page 1026
Zhang et al—Logistic Regression Formula for Subcentimeter Thyroid Nodules
The frequency distribution of sonographic features of benign and malignant nodules are listed in Table 3. Suspicious sonographic features were a solid echo structure (OR, 41.97), a microlobulated margin (OR, 25.89), hypoechoic echogenicity (OR, 10.36), no halo (OR, 8.38), an irregular margin (OR, 4.26), a taller-than-wide shape (OR, 2.71), microcalcification (OR, 1.92), a solitary nodule
(OR, 1.58), and macrocalcification (OR, 1.28). No significant differences in malignancy were detected among different subtypes of vascularization. Thus, vascularization was not assessed for diagnostic performance. For each sonographic feature, diagnostic performance was calculated, and values are presented in Table 4, including sensitivity, specificity, PPV, NPV, and accuracy.
Table 1. Histologic Findings of the 830 Subcentimeter Nodules
Table 2. Clinical Features of Patients and Nodules
Finding
Feature
Nodules, n (%)
Malignant (n = 405) Papillary carcinoma Follicular variant of papillary carcinoma Medullary carcinoma Benign (n = 425) Nodular hyperplasia Follicular adenomaa Hashimoto thyroiditis de Quervain thyroiditis Other benign lesions
392 (96.8) 10 (2.5) 3 (0.7) 263 (61.9) 25 (5.9) 40 (9.4) 6 (1.4) 91 (21.4)
Age, y Sex, n (%) Male Female Size, cm Height Width
Benign
Pa
Malignant
50.15 ± 11.04
45.94 ± 10.02
.99 >.99
0.57 ± 0.18 0.70 ± 0.21
0.65 ± 0.18 0.67 ± 0.19
1 Echogenicity, n (%) Hypoechoic Nonhypoechoic Echo structure, n (%) Solid Nonsolid Margin, n (%) Well circumscribed Microlobulated Irregular Calcification, n (%) No calcification Microcalcification Macrocalcification Halo, n (%)a No halo Halo Height-to-width ratio, n (%) >1 (taller than wide)
ORIGINAL RESEARCH
Development of a Logistic Regression Formula for Evaluation of Subcentimeter Thyroid Nodules Mingbo Zhang, MD, Yan Zhang, MD, Shuai Fu, MD, Faqin Lv, MD, Jie Tang, MD
Objectives—The purpose of this study was to build a logistic regression formula for ameliorating the diagnosis of subcentimeter thyroid nodules. Methods—The sonographic features of 889 subcentimeter nodules were reviewed retrospectively with reference of histologic results. The diagnostic performance of each feature was evaluated. Multivariate binary logistic regression was used to develop the formula for evaluation of subcentimeter nodules, and the cutoff value was decided for recommending biopsy. Results—The logistic regression formula was –0.029age – 2.063US1 – 0.812US2 + 1.781US3 + 1.627height-to-width ratio – 0.333 (for nonhypoechogenicity, US1 = 1; for hypoechogenicity, US1 = 0; for a well-circumscribed margin, US2 = 1 and US3 = 0; for a microlobulated margin, US2 = 0 and US3 = 1; and for an irregular margin, US2 = 0 and US3 = 0). The area under the receiver operating characteristic curve for the formula was 0.860. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of the formula under the cutoff value of 0.284 were 90.9%, 54.0%, 66.4%, 85.6%, and 72.5%, respectively. With the use of the formula, 229 of 830 nodules could avoid surgery. Conclusions—A logistic regression formula with a cutoff value could provide an objective and easy tool with effective diagnostic performance, which could improve diagnosis of subcentimeter thyroid nodules and reduce unnecessary biopsy, decreasing costs and patient discomfort. Key Words—biopsy; nodule; sonography; superficial structures; thyroid
Received August 1, 2013, from the Department of Ultrasound, General Hospital of Chinese PLA, Beijing, China. Revision requested August 20, 2013. Revised manuscript accepted for publication October 7, 2013. Address correspondence to Jie Tang, MD, Department of Ultrasound, General Hospital of Chinese PLA, 28 Fuxing Rd, Haidian District, 100853 Beijing, China. E-mail: [email protected] Abbreviations
NPV, negative predictive value; OR, odds ratio; PPV, positive predictive value doi:10.7863/ultra.33.6.1023
T
hyroid nodules are commonly seen in daily practice. They are detected in 4% to 7% of patients by palpation and 50% at autopsy.1,2 Recently, the use of high-frequency sonography has greatly increased the detection of subcentimeter (maximum diameter, ≤1 cm) thyroid nodules. Although the clinical importance of subcentimeter nodules for malignancy is widely debated,3–7 studies suggest that their malignant prevalence is similar to that of palpable nodules, and their malignant involvement is not less frequent.4,8,9 The American Thyroid Association guideline does not recommend fine-needle aspiration of sonographically suspicious subcentimeter nodules unless there is associated high-risk history.10 However, other organizations and authors suggest that fine-needle aspiration biopsy should also be performed on subcentimeter nodules with suspicious sonographic findings.8,11 Uncertainty from not knowing may adversely affect the quality of life in these patients.11
©2014 by the American Institute of Ultrasound in Medicine | J Ultrasound Med 2014; 33:1023–1030 | 0278-4297 | www.aium.org
3306jum929-1084_Layout 1 5/21/14 8:56 AM Page 1024
Zhang et al—Logistic Regression Formula for Subcentimeter Thyroid Nodules
Thus, detecting suspicious sonographic features of subcentimeter nodules is worthwhile. Many studies have been performed for diagnosis of thyroid carcinoma using sonographic features. However, the features have varying abilities to predict malignancy,8,9,12–21 and most studies included nodules with various sizes, not only subcentimeter nodules. Sharma et al9 suggested the importance of the clinician’s overall impression to the malignancy of subcentimeter nodules, not just a single feature. Thus, our study tried to improve the diagnosis of subcentimeter thyroid nodules with a comprehensive solution, which is better for saving medical resources and reducing patient discomfort. In our study, the sonographic features of a large sample of subcentimeter nodules were retrospectively analyzed with reference to histologic results to develop a logistic regression formula for improving the diagnosis. In addition, a cutoff value was determined for recommending biopsy to reduce unnecessary biopsy and to decrease costs and patient discomfort.
Materials and Methods Patients Our study was approved by the Institutional Review Board of our hospital, and the requirement for informed consent was waived. Subcentimeter thyroid nodules with histologic results from sonographically guided core needle biopsy, thyroidectomy, or both at our institution between April 2009 and April 2012 were reviewed. Our institution is a tertiary health center, and patients from all over the country seek services here. During this period, 2172 thyroid nodules received core needle biopsy, and 889 of them were subcentimeter nodules. No patient had more than 1 core needle biopsy during this period. Among the 889 nodules, 59 were excluded, including 11 inadequate core needle biopsy samples, 25 atypia of undetermined importance (atypical cells that presented without a diagnosis of suspicion for malignancy), 15 follicular neoplasms, and 8 nodules that were suspicious for malignancy (atypia with insufficient evidence for a definite diagnosis of malignancy), which did not receive surgery. Finally, 830 nodules in 788 patients were included in this study, including 197 nodules receiving only surgical resection, 563 nodules receiving only core needle biopsy, and 70 nodules receiving both at our institution. Ages and sexes of the patients were recorded. A flowchart of the study group is presented in Figure 1. Subcentimeter nodules were subjected to core needle biopsy if one of the following suspicious findings was present: hypoechogenicity, calcification, an irregular or
1024
microlobulated margin or no halo, intranodular vascularity, a taller-than-wide shape, sonographic features highly indicating metastatic cervical lymph nodes, and substantial growth (20% increase in 2 minimum dimensions, at least 2 mm) since a previous sonographic examination within 18 months.10 For patients undergoing surgery at our institution, the final diagnosis was made by surgical histologic reports. Otherwise, core needle biopsy histologic results were used instead. At the Department of Pathology of our institution, the core needle biopsy and thyroidectomy specimens were routinely examined by a single boardcertified pathologist and reviewed by a single chief pathologist with more than 10 years of experience in thyroid pathologic examinations. If there were more than 2 nodules, the physician who performed the core needle biopsy or thyroidectomy did the radiologic-pathologic correlation to ensure concordance. Imaging Analysis The examinations before core needle biopsy were used for the analysis. The sonograms in our study were obtained by board-certified sonographers using highfrequency linear probes (L5-17 MHz) and iU22 (Philips Healthcare, Bothell, WA) and LOGIQ 9 (GE Healthcare, Milwaukee, WI) ultrasound systems. Images were stored in our picture archiving and communication system, and the sonographic features of the nodules were reevaluated by 2 sonographers (M.Z. and Y.Z.) with more than 5 years of experience in thyroid examinations, who were blinded to each other during the evaluation process. If there were disagreements about the features, the sonographers reached a consensus after discussion with an expert (F.L.) Figure 1. Flowchart of the study group for developing the multivariate regression formula. CNB indicates core needle biopsy.
J Ultrasound Med 2014; 33:1023–1030
3306jum929-1084_Layout 1 5/21/14 8:56 AM Page 1025
Zhang et al—Logistic Regression Formula for Subcentimeter Thyroid Nodules
who had more than 15 years of experience in thyroid examinations. All of them were blinded to the histopathologic diagnosis. On sonography, the thyroid nodules were described according to their echogenicity, internal component, margin, calcification, halo, size, and vascularization. Echogenicity was classified as hyperechoic, isoechoic, hypoechoic, or anechoic. When the echogenicity of the nodule was similar to that of the thyroid parenchyma, it was classified as isoechoic. The nodule was classified as hypoechoic or hyperechoic if the echogenicity was less than or more than that of the thyroid parenchyma. The internal component of the nodules was classified as solid, mixed solid, or cystic. The margin was classified as well circumscribed, microlobulated, or irregular. A microlobulated margin was defined as the presence of many small lobules on the nodular surface. Calcification, when present, was categorized as microcalcification or macrocalcification. Microcalcification was defined as calcification of 1 mm or smaller in diameter and visualized as tiny punctuate hyperechoic foci, either with or without acoustic shadows. Tiny bright reflectors with a clear-cut comet tail artifact were considered colloid.22 Macrocalcification was defined as hyperechoic foci of 1 mm or larger. When a nodule had both types of calcification (macrocalcification, including rim calcification, intermingled with microcalcification), we regarded it as having microcalcification. A halo was categorized as halo or no halo. Height was defined as the anteroposterior diameter, and width was defined as the larger of the transverse or longitudinal diameter. Width and height were measured, and the ratio of height-to-width ratio was calculated. Vascularization was categorized as no blood flow, internal blood flow, or peripheral blood flow.12 Sonographically Guided Core Needle Biopsy Procedures Sonographically guided core needle biopsy procedures were performed by experienced interventional sonographers using a Magnum biopsy instrument and a 1.5- or 2.2-cm excursion double-action spring-activated 18-gauge needle (Bard Medical Division, Covington, GA). A high-frequency linear probe with a needle guide line and matched guiding frame was used for accurate positioning of the needle track. After induction of local anesthesia with 1% lidocaine, the core needle was inserted in the direction of the guide line. Color Doppler imaging was used for evaluation of vessels along the approach route to avoid hemorrhage. After the tip of the biopsy needle had been advanced into the edge of the nodule, the firing distance was measured to minimize vessel injury for safety. The stylet and cutting cannula of the needle were sequentially fired. Two core biopsy
J Ultrasound Med 2014; 33:1023–1030
passes were made for each nodule. The puncture positions were pressed as soon as the biopsy passes were finished in case of hematoma, which lasted at least 30 minutes. Statistical Analysis Descriptive data were presented as mean ± standard deviation or number (percent). The t test or χ2 test was used to determine the sonographic features and clinical information between benign and malignant nodules. Since hypoechogenicity and a solid internal component are two suspicious malignant features,8,10 the echogenicity and echo structure were dichotomized as hypoechoic/nonhypoechoic and solid/nonsolid during the evaluation process to determine diagnostic performance. The odds ratio (OR) of each sonographic feature was analyzed. If there were more than 2 subgroups of features, the least suspicious (lowest malignant rate) feature was used as the reference, and the ORs of the other suspicious features were analyzed and compared with reference to it. The diagnostic performances of sonographic features were evaluated, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy. Multivariate binary logistic regression with the forward conditional method was used to develop the formula for recommending sonographically guided biopsy. The diagnostic value of the formula was evaluated by a receiver operating characteristic curve. The cutoff value for recommending biopsy and its corresponding diagnostic values were evaluated. Statistical analysis was performed with SPSS version 13.0 software (IBM Corporation, Armonk, NY). P < .05 was considered statistically significant.
Results A total of 830 subcentimeter nodules (425 benign and 405 malignant) in 789 patients (618 female and 171 male) were included in our study. The mean size (longest diameter) of the nodules ± SD was 0.69 ± 0.20 cm (range, 0.20–1.00 cm). The mean age of the patients was 48.1 ± 10.8 years (range, 12–83 years). Most malignant nodules (96.8%) were papillary carcinoma. The pathologic diagnoses of the 830 nodules are listed in Table 1. Patients with malignant nodules were significantly younger than those with benign nodules (P < .001). The height of malignant nodules was significantly larger (P < .001) and their width was significantly smaller (P = .049) than those of benign nodules. Malignant nodules had a greater chance of being solitary than benign nodules (P = .003). No significant difference in malignancy was found between male and female patients. The clinical features of the patients and nodules are listed in Table 2.
1025
3306jum929-1084_Layout 1 5/21/14 8:56 AM Page 1026
Zhang et al—Logistic Regression Formula for Subcentimeter Thyroid Nodules
The frequency distribution of sonographic features of benign and malignant nodules are listed in Table 3. Suspicious sonographic features were a solid echo structure (OR, 41.97), a microlobulated margin (OR, 25.89), hypoechoic echogenicity (OR, 10.36), no halo (OR, 8.38), an irregular margin (OR, 4.26), a taller-than-wide shape (OR, 2.71), microcalcification (OR, 1.92), a solitary nodule
(OR, 1.58), and macrocalcification (OR, 1.28). No significant differences in malignancy were detected among different subtypes of vascularization. Thus, vascularization was not assessed for diagnostic performance. For each sonographic feature, diagnostic performance was calculated, and values are presented in Table 4, including sensitivity, specificity, PPV, NPV, and accuracy.
Table 1. Histologic Findings of the 830 Subcentimeter Nodules
Table 2. Clinical Features of Patients and Nodules
Finding
Feature
Nodules, n (%)
Malignant (n = 405) Papillary carcinoma Follicular variant of papillary carcinoma Medullary carcinoma Benign (n = 425) Nodular hyperplasia Follicular adenomaa Hashimoto thyroiditis de Quervain thyroiditis Other benign lesions
392 (96.8) 10 (2.5) 3 (0.7) 263 (61.9) 25 (5.9) 40 (9.4) 6 (1.4) 91 (21.4)
Age, y Sex, n (%) Male Female Size, cm Height Width
Benign
Pa
Malignant
50.15 ± 11.04
45.94 ± 10.02
.99 >.99
0.57 ± 0.18 0.70 ± 0.21
0.65 ± 0.18 0.67 ± 0.19
1 Echogenicity, n (%) Hypoechoic Nonhypoechoic Echo structure, n (%) Solid Nonsolid Margin, n (%) Well circumscribed Microlobulated Irregular Calcification, n (%) No calcification Microcalcification Macrocalcification Halo, n (%)a No halo Halo Height-to-width ratio, n (%) >1 (taller than wide)
