CLINICAL REVIEW

David W. Eisele, MD, Section Editor

Meta-analysis of thyroid imaging reporting and data system in the ultrasonographic diagnosis of 10,437 thyroid nodules Xi Wei, MD, PhD,1 Ying Li, MD,2 Sheng Zhang, MD,1 Ming Gao, MD, PhD3* 1

Department of Diagnostic and Therapeutic Ultrasonography, Tianjin Medical University Cancer Institute and Hospital, Key Laboratory of Cancer Prevention and Therapy, Tianjin, China, 2The Third Department of Breast Cancer, Tianjin Medical University Cancer Institute and Hospital, Key Laboratory of Cancer Prevention and Therapy, Tianjin, China, 3 Department of Thyroid and Cervical Tumor, Tianjin Medical University Cancer Institute and Hospital, Key Laboratory of Cancer Prevention and Therapy, Tianjin, China.

Accepted 18 September 2014 Published online 16 June 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/hed.23878

ABSTRACT: Background. The meta-analysis was performed to review the diagnostic accuracy of ultrasound reporting and data system in the diagnosis of thyroid nodules. Methods. We identified the diagnostic accuracy of ultrasound reporting and data system in 5 databases. Meta-analyses were used in selected studies to obtain pooled sensitivity, specificity, and summary receiver operating characteristic (ROC) curves. Fixed or random-effects models were performed to analyze our data. Results. Twelve eligible studies were identified, including 10,437 thyroid nodules. A pooled sensitivity of 0.79 (95% confidence interval [CI] 5 0.77–0.81) and a pooled specificity of 0.71 (95% CI 5 0.70–0.72) of

INTRODUCTION The incidence of thyroid cancer has been rising 200% to 300% in China within the past 30 years.1 Ultrasound diagnosis of thyroid nodules has been considered as a useful tool for quantitative and qualitative evaluation of these lesions. Palpable and hard thyroid nodule by examination is associated with an increased risk of thyroid cancer.2 However, nonpalpable thyroid nodules are difficult to be examined in the clinic, but can be differentiated by ultrasonography and ultrasound-guided fine-needle aspiration biopsy (FNAB), even for nodules measuring more than 3 mm.3,4 The recent technical developments of ultrasonography, such as elastography and contrast-enhanced ultrasound, have led to an increase in the diagnostic accuracy, based on the tissue hardness and microcirculation.5,6 Although several series have performed elastography and contrast-enhanced ultrasound in the diagnosis of thyroid lesions, comparison with gray-scale ultrasonography with suspicious features is necessary in clinical practice.7 Suspicious ultrasonography features, such as marked hypoechogenicity, irregular margins, taller than

*Corresponding author: M. Gao, Department of Thyroid and Cervical Tumor, Tianjin Medical University Cancer Institute and Hospital, Key Laboratory of Cancer Prevention and Therapy, Huanhuxi Road, Hexi District, Tianjin, China 300060. E-mail: [email protected] Xi Wei, Ying Li, and Sheng Zhang contributed equally to this work.

ultrasound reporting system in differentiated diagnosis of thyroid nodules were shown in meta-analyses. Subgroup analyses showed that the most important factor of heterogeneity in studies was the final diagnostic references (histological and cytological standards or only histological results). Conclusion. The thyroid imaging reporting and data system has a good sensitivity and specificity in diagnosis of patients with thyroid nodules. C 2014 Wiley Periodicals, Inc. Head Neck 38: 309–315, 2016 V

KEY WORDS: thyroid reporting and data system, thyroid nodules, diagnostic accuracy

wide shape, and microcalcifications, were basic features to identify thyroid nodules into benign and malignant lesions.8 Horvath et al10 and Park et al9 initially intended to categorize different percentages of thyroid malignancies to establish an ultrasound reporting and data system, which follows the concept of Breast Imaging Reporting and Data System.12,13 Breast Imaging Reporting and Data System were widely used as a standard method to describe mammographic and ultrasonography breast lesions to correlate with breast malignancies.11–13 The thyroid imaging reporting and data system, also called TI-RADS, was established based on main sonographic suspicious features for malignances, such as solid component, hypoechogenicity, marked hypoechogenicity, micro-lobulated or irregular margins, microcalcifications, and taller than wide shape. As the Horvath et al10 report suggested, TI-RADS was classified into 6 categories: TI-RADS 1 5 normal thyroid gland; TI-RADS 2 5 benign conditions; TI-RADS 3 5 probably benign nodules; TI-RADS 4 5 suspicious nodules; TI-RADS 5 5 probably malignant nodules; and TI-RADS 6 5 malignant nodules proven by biopsy. TI-RADS were carried out and utilized by radiologists worldwide on the basis of ultrasonographic imaging features to stratify malignant risks of thyroid nodules.9,10,14–23 In this review, we aimed to perform meta-analyses to evaluate the diagnostic accuracy of TI-RADS in patients with thyroid nodules. HEAD & NECK—DOI 10.1002/HED

FEBRUARY 2016

309

WEI ET AL.

lish or Chinese; (2) methodological quality of included studies were included in this review and quality assessment of diagnostic accuracy studies was used as a quality assessment tool to assess the quality of articles33; (3) TIRADS system was performed in the differentiated diagnosis of thyroid nodules; (4) histological and/or cytological analyses were used as the reference standard; (5) sufficient patients in studies were presented to calculate the true-positive, false-negative, false-positive, and truenegative values for data statistics; and (6) the duplicated articles were chosen to keep the most details or the most recent studies.

Statistical analysis

FIGURE 1. The procedure of study selection in our meta-analysis resulted in a total of 12 studies that were included in this systematic review, which fulfilled all of the inclusion criteria. TI-RADS, Thyroid Imaging Reporting and Data System.

METHODS Data sources We searched for reports in English and Chinese that were about thyroid imaging report and data system used in clinical practice. The PubMed (1966–2013.8), Cochrane Library, Google Scholar, WanFang Data (Chinese 1998– 2013.8), and China National Knowledge Infrastructure electronic databases (1994–2013.8) were used for all publications dated from May 1966 to August 2013. The search terms included “thyroid imaging reporting and data system” and “TI-RADS.” We excluded unpublished data of thyroid ultrasound imaging reports.

Meta-analysis was performed based on construct 2 3 2 contingency tables, which were for sensitivity and specificity calculated in each study. At first, we used Spearman correlation coefficient for diagnostic threshold analysis. If there was the diagnostic threshold in these data, summary receiver operating characteristic (ROC) was fitted using the Mantel–Haenszel or Moses–Shapiro– Littenberg model. Pooled sensitivity and specificity, with 95% confidence intervals (CIs), were obtained without the diagnostic threshold in all studies. The heterogeneity was assessed by the likelihood ratio chi-square test. In the likelihood ratio chi-square test, a p value of < .05 was considered as the apparent heterogeneity. A random effect model and fixed effect model were used for primary meta-analysis when the heterogeneity existed or not.24 Combined sensitivity, specificity, likelihood ratio, and diagnostic odds ratio (DOR) were all calculated after heterogeneity assessment. Forest plot was presented based on these parameters. In addition, summary ROC, a mathematic transformation of sensitivity and specificity (Q* values), has been described by Moses et al.25 The weighted area under the curve (AUC) was obtained to measure the diagnostic performance of TI-RADS. Statistical significance was p < .05. All data in our analyses were carried out using Stata/SE statistical software version 11.1 (StataCorp LP, College Station, TX).

Selection criteria and outcomes

Subgroup analyses

The criteria and outcomes for selection in our metaanalysis were as follows: (1) the language limited in Eng-

We performed subgroup analyses using different criteria as follows: (1) the year of publication; (2) language of

TABLE 1. The basic characteristics of 12 studies. Authors

Publication year Country/region Language No. of patients Average age No. of nodules

Horvath et al [10] Park et al [9] Ma et al [14] Zhou et al [15] Zhang et al [16] Lou et al [17] Xie et al [19] Chen et al [18] Chen et al [21] Lu et al [20] Russ et al [23] Wang and Wang [22]

2009 2009 2011 2011 2012 2012 2013 2013 2013 2013 2013 2013

Chile Korea China China China China China China Taiwan China France China

English English Chinese Chinese Chinese Chinese Chinese Chinese English Chinese English Chinese

Abbreviation: NR, not reported.

310

HEAD & NECK—DOI 10.1002/HED

FEBRUARY 2016

NR 2679 292 149 186 69 180 263 437 520 3543 136

NR 48.8 46 45.5 44.9 48.5 46 41.4 NR 48.7 54 41.3

1097 1707 423 250 210 183 225 524 498 634 4550 136

Study design

Reference standard

Prospective Cytology and histology Retrospective Cytology and histology Retrospective Histology Retrospective Histology Retrospective Histology Retrospective Histology Retrospective Histology Retrospective Histology Retrospective Histology Retrospective Histology Prospective Histology and cytology Retrospective Histology

META–ANALYSIS

OF THYROID IMAGING REPORTING AND DATA SYSTEM

publication; (3) the number of thyroid nodules enrolled; and (4) reference standards.

RESULTS Literature selection Of 575 references, 12 studies9,10,14–23 involving 10,437 thyroid nodules were identified as eligible for inclusion in this meta-analysis. The protocol of literature selection is listed in Figure 1.

Basic characteristics of the studies The basic characteristics of 12 studies are shown in Table 1. As for Figure 2, the funnel plot was performed to evaluate a possible publication bias. The plot indicated there was probably no publication bias, showing a symmetry figure. FIGURE 2. Funnel plot was described to show symmetry that there was probably no publication bias in the meta-analysis. The points in this figure were labeled numbers 1 to 12, which indicated 1 to 12 studies (1 5 Horvath et al 2009 [10]; 2 5 Park et al 2009 [9]; 3 5 Ma et al 2011 [14]; 4 5 Zhou et al 2011 [15]; 5 5 Zhang et al 2012 [16]; 6 5 Lou et al 2012 [17]; 7 5 Xie et al 2013 [19]; 8 5 Chen et al 2013 [18]; 9 5 Chen et al 2013 [21]; 10 5 Lu et al 2013 [20]; 11 5 Russ et al 2013 [23]; 12 5 Wang and Wang 2013 [22]). DOR, diagnostic odds ratio.

Meta-analyses outcomes The pooled sensitivity and specificity of TI-RADS were 0.79 (95% CI 5 0.77–0.81) and 0.71 (95% CI 5 0.70–0.72), respectively (p 5 .000; Figure 3A and 3B). The summary positive and negative likelihood ratio were 6.6 (95% CI 5 4.4–9.9) and 0.20 (95% CI 5 0.14–0.29), respectively (Figure 4A and 4B).

FIGURE 3. Forest plot showed pooled sensitivity (A) (red arrow; 0.79; 95% confidence interval [CI] 5 0.77–0.81) and specificity (B) (red arrow; 0.71; 95% CI 5 0.70–0.72) of Thyroid Imaging Reporting and Data System (TI-RADS) in the differentiated diagnosis of thyroid nodules. The sensitivity (A) and specificity (B) of 12 studies are also demonstrated in this figure. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

HEAD & NECK—DOI 10.1002/HED

FEBRUARY 2016

311

WEI ET AL.

FIGURE 4. Forest plot showed pooled positive likelihood ratio (A) (red arrow; 6.6; 95% confidence interval [CI] 5 4.4–9.9) and negative likelihood ratio (B) (red arrow; 0.20; 95% CI 5 0.14–0.29) of Thyroid Imaging Reporting and Data System (TI-RADS) in the differentiated diagnosis of thyroid nodules. The positive likelihood ratio (A) and negative likelihood ratio (B) of 12 studies were also demonstrated in this figure. LR, likelihood ratio. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

FIGURE 5. Forest plot showed diagnostic odds ratio (DOR; red arrow; 5.2; 95% confidence interval [CI] 5 19.5–63.4) of Thyroid Imaging Reporting and Data System (TI-RADS) in the differentiated diagnosis of thyroid nodules. The DOR of 12 studies are also demonstrated in this figure. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

312

HEAD & NECK—DOI 10.1002/HED

FEBRUARY 2016

META–ANALYSIS

OF THYROID IMAGING REPORTING AND DATA SYSTEM

FIGURE 6. Summary receiver operating characteristic (ROC) curves of Thyroid Imaging Reporting and Data System (TI-RADS) in the evaluation of thyroid nodules. As this figure shown, area under the curve (AUC) was 0.9177 and the Q* index was 0.8507. The points in this figure were labeled number 1 to 12, which indicated 1 to 12 studies (1 5 Horvath et al 2009 [10]; 2 5 Park et al 2009 [9]; 3 5 Ma et al 2011 [14]; 4 5 Zhou et al 2011 [15]; 5 5 Zhang et al 2012 [16]; 6 5 Lou et al 2012 [17]; 7 5 Xie et al 2013 [19]; 8 5 Chen et al 2013 [18]; 9 5 Chen et al 2013 [21]; 10 5 Lu et al 2013 [20]; 11 5 Russ et al 2013 [23]; 12 5 Wang and Wang 2013 [22]). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

The pooled DOR was 35.2 (95% CI 5 19.5–63.4; see Figure 5). The summary ROC was symmetry according to no differences between b and zero (p 5 .07; see Figure 6). The overall AUC was 0.9177 and the Q* index was 0.8507, indicating very good diagnostic accuracy (see Figure 6).

95% CI 5 0.810–0.841 vs 0.756; 95% CI 5 0.729–0.782; 0.662; 95% CI 5 0.650–0.673). Some other subgroups, such as language in English or not, the number of thyroid nodules (500) had no influence on the overall sensitivity and specificity significantly (p > .05).

Subgroup analysis and meta-regression

DISCUSSION

The 4 subgroups in our study are shown in Table 2. The pooled sensitivity and specificity of studies with only surgery are higher than those with FNAB and surgery by the subgroup analysis (0.838; 95% CI 5 0.810–0.863; 0.826;

Ultrasonography has be regarded as a screening tool in the preoperative diagnosis of thyroid nodules.26,27 Multiple sonographic features may be more valuable for accurate assessment of benignity and malignancy than a single one

TABLE 2. Subgroups analyses in the meta-analyses. Subgroups

Total* Year 5 2013 Year < 2013 English Chinese No. of patients (500) No. of patients (

Meta-analysis of thyroid imaging reporting and data system in the ultrasonographic diagnosis of 10,437 thyroid nodules.

The meta-analysis was performed to review the diagnostic accuracy of ultrasound reporting and data system in the diagnosis of thyroid nodules...
1MB Sizes 1 Downloads 4 Views

Recommend Documents