bmj open€¦ · 22 related to dwi in differentiating benign from malignant thyroid nodules....

For peer review only

Diffusion-weighted magnetic resonance imaging in differentiating malignant from benign thyroid nodules: A

meta-analysis

Journal: BMJ Open

Manuscript ID: bmjopen-2015-008413

Article Type: Research

Date Submitted by the Author: 07-Apr-2015

Complete List of Authors: Chen, Lihua; Southwest Hospital, Department of Radiology Xu, Jian; PLA101 Hospital, Department of General Surgery Bao, Jing; Wuxi center for disease control and prevention, Molecular

biology lab Huang, Xuequan; Southwest Hospital, Department of Radiology Xia, Yunbao; PLA101 Hospital, Department of Radiology Wang, Jian; Southwest Hospital, Department of Radiology

<b>Primary Subject Heading</b>:

Radiology and imaging

Secondary Subject Heading: Diagnostics, Evidence based practice

Keywords: Magnetic resonance imaging < RADIOTHERAPY, thyroid nodules, Diffusion-weighted imaging, Meta-analysis

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml

BMJ Open on D

ecember 29, 2020 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-008413 on 5 January 2016. D

ownloaded from

http://bmjopen.bmj.com/


1

Title Page 1

Title: Diffusion-weighted magnetic resonance imaging in differentiating malignant from benign thyroid 2

nodules: A meta-analysis. 3

4

Authors: Lihua Chen1,2

, †MS; Jian Xu3 †MS; Jing Bao

4, MS; Xuequan Huang

1, MD; Yunbao Xia

2, *BS; 5

Jian Wang1, *MD 6

7

Lihua Chen, 1Department of Radiology, Southwest Hospital, Third Military Medical University, 8

Chongqing, 400038, China; 2Department of Radiology, Taihu Hospital, Wuxi, Jiangsu Province, 214044, 9

China, E-mail: [email protected]. 10

Jian Xu, Department of General Surgery, Taihu Hospital, Wuxi, Jiangsu Province, 214044, China, E-mail: 11

[email protected]. 12

Jing Bao, Molecular biology lab, Wuxi center for disease control and prevention, Wuxi, Jiangsu Province, 13

214001, China, E-mail: [email protected]. 14

Xuequan Huang, Department of Radiology, Southwest Hospital, Third Military Medical University, 15

Chongqing, 400038, China, E-mail: [email protected]. 16

Yunbao Xia, Department of Radiology, Taihu Hospital, Wuxi, Jiangsu Province, 214044, China, E-mail: 17


Jian Wang, Department of Radiology, Southwest Hospital, Third Military Medical University, Chongqing, 19


21

† Co-first Author 22

* Co-corresponding Author 23

Page 1 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem

ber 29, 2020 by guest. Protected by copyright.

http://bmjopen.bm

j.com/

BM

J Open: first published as 10.1136/bm

jopen-2015-008413 on 5 January 2016. Dow

nloaded from



2

Correspondence to: 1

Jian Wang 2

Department of Radiology, 3

Southwest Hospital, 4

Third Military Medical University, 5

Chongqing, 400038, China 6

Tel: +86-23-6875-4419 7

Fax: +86-23-6546-3026 8

E-mail: [email protected] 9

Or 10

Yunbao Xia 11

Department of Radiology, 12

Taihu Hospital, 13

Wuxi, 214044, Jiangsu Province, China 14

Tel: +86-510-8514-2222 15

Fax: +86-510-8514-2222 16


18

19

Page 2 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



3

Abstract 1

Objectives: To perform a meta-analysis to evaluate the diagnostic efficacy of diffusion-weighted imaging 2

(DWI) in differentiating malignant from benign thyroid nodules. 3

Design: A meta-analysis. 4

Data sources and study selection: Medical and scientific literature databases were searched for original 5

articles published prior to May, 2014. Studies were selected that (1) included a diagnostic DWI for 6

differentiating malignant versus benign thyroid lesions, (2) included patients who later underwent biopsy, 7

and (3) presented sufficient data to allow construction of contingency tables. 8

Data synthesis: For each study, the true-positive, false-positive, true-negative, and false-negative values 9

were extracted or derived, and 2×2 contingency tables were constructed. Methodological quality was 10

assessed by the Quality Assessment of Diagnostic Studies (QUADAS) instrument. The heterogeneity test, 11

the threshold effect test, the subgroup analyses, and publication bias analyses were performed. 12

Results: From the 79 search results, 15 studies were included in the meta-analysis, representing a total of 13

765 lesions. We detected heterogeneity between studies and found no evidence of publication bias. The 14

methodological quality was moderate. The pooled weighted sensitivity, with a corresponding 95% 15

confidence interval (CI), was 0.90 (95% CI: 0.85, 0. 93); the specificity was 0.95 (95% CI: 0.88, 0.98); 16

the positive likelihood ratio was 16.49 (95% CI: 7.37, 36.86); the negative likelihood ratio was 0.11 (95% 17

CI: 0.08, 0.16); and the diagnostic odds ratio was 150.73 (95% CI: 64.96, 349.75). The area under the 18

receiver operator characteristic curve was 0.95 (95% CI: 0.93, 0.97). 19

Conclusions: Quantitative DWI is a non-invasive, non-radiative, and accurate method of distinguishing 20

malignant from benign thyroid nodules. Nevertheless, large-scale trials are necessary to assess its clinical 21

value and to establish diagnosis standards regarding b values and cut-off values for DWI. 22

23

Page 3 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



4

Keywords: Magnetic resonance imaging, thyroid nodules, Diffusion-weighted imaging, Meta-analysis 1

2

Strengths and limitations of this study 3

Assess SEN and SPE with hierarchical summary receiver operating characteristic curves. 4

The subgroup analysis about the technical aspects of DWI and its additional value in thyroid tumour. 5

A new view that using a high b value may provide better results. 6

Small sample size. 7

Bias with reporting, selecting. 8

9

Abbreviations: 10

AUC the areas under the curves 11

CI confidence interval 12

FN false-negative 13

FP false-positive 14

HSROC the hierarchical summary receiver operating characteristic 15

MR magnetic resonance 16

NLR negative likelihood ratio 17

PLR positive likelihood ratio 18

QUADAS the quality assessment of diagnostic studies 19

SEN sensitivity 20

SPE specificity 21

TN true-negative 22

TP true-positive 23

24

Page 4 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



5

INTRODUCTION 1

Thyroid nodules are the most common pathology involving the thyroid gland. Thyroid nodules 2

consist of discrete lesions within the thyroid gland and are often palpable and usually sonographically 3

different from the surrounding thyroid parenchyma [1]. Less than 5% of palpable thyroid nodules are 4

malignant; however, these nodules must be distinguished from benign thyroid nodules to correctly and 5

efficaciously treat the patient suffering from this pathology [2]. 6

Because the clinical findings do not provide a definitive diagnoses, some useful, non-invasive 7

imaging tests (such as ultrasonography [US] and radionuclide scintigraphy) can be used to determine 8

which nodules should undergo histopathologic evaluation to rule out the possibility of thyroid malignancy. 9

US has been used as a first step in the assessment of these nodules, but no single US criterion has been 10

able to accurately differentiate benign nodules from malignant nodules. Furthermore, the hazards 11

associated with radiation exposure are unavoidable in radio nuclide scintigraphy, and some functioning 12

nodules (hot nodules) found using on scintigraphy are malignant. 13

DWI is a type of functional MRI that is based on the diffusion of water molecules through the tissue 14

of interest (i.e., tumors). DWI can provide crucial information regarding the molecular structure of the 15

underlying pathology and pathophysiologic mechanisms [3]. Specifically, the diffusion of water 16

molecules in malignant tumors is restricted, which results in a lower apparent diffusion coefficient (ADC) 17

value, which facilitates the differentiation of benign tumors from malignant tumors [4]. 18

A large number of studies [1 5-18] have shown that DWI has the potential to differentiate benign 19

from malignant thyroid nodules. However, the sample sizes in these studies were relatively small, and the 20

findings have been inconclusive. The aim of this study was to systematically review all of the studies 21

related to DWI in differentiating benign from malignant thyroid nodules. Moreover, based on the 22

Page 5 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



6

extracted data, an analysis regarding the technical aspects of DWI and its additional value in tumour 1

characterisation is presented. 2

3

METHODS 4

As this meta-analysis was based on previous published studies, ethical approval was not necessary. 5

Literature Search 6

PubMed, Embase, Cochrane Library, and China National Knowledge Infrastructure (CNKI) databases 7

were searched by two independent observers. The terms “Diffusion-Weighted Imaging [MeSH]” or “DWI” 8

were used for the diagnostic test, and the terms “thyroid nodules [MeSH]”, “thyroid lesions,” or “thyroid” 9

were used for the clinical domain. We limited our search to publications that met the following criteria: 10

published in the English and Chinese language; the presence of the search term within the title or abstract 11

of the article; and a publication date no later than May, 2014. Review articles, letters, comments, case 12

reports, and unpublished articles were excluded. The reference lists of all retrieved articles were manually 13

cross-checked. 14

Selection of Articles 15

Two authors initially screened the titles and abstracts of the search results and retrieved the full text 16

of all potentially relevant reports. Next, the authors independently reviewed all relevant reports, according 17

to the predefined inclusion criteria. Disagreements were resolved by a consensus or third author 18

arbitration who assessed all of the involved items. The majority opinion was used to determine whether a 19

particular study met the selection criteria. 20

Studies were eligible if the following criteria were met: (a) the study included a diagnostic DWI for 21

differentiating malignant versus benign thyroid lesions; (b) the control included a histopathologic analysis 22

Page 6 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



7

(surgery/biopsy) and/or follow-up analysis; and (c) the data were sufficient to accurately calculate a 1

true-positive or false-negative value. 2

Studies were excluded if (a) there were fewer than 20 patients; (b) multiple reports were published 3

for the same study population (in this case, the most detailed or recent publication was chosen); and (c) 4

the study included patients who had previously undergone treatment for thyroid lesions. 5

Quality Assessment and Data Extraction 6

The aforementioned three authors extracted data from the selected reports. The methodological 7

quality of the included studies was independently assessed by two observers using the quality assessment 8

of diagnostic studies (QUADAS-2) tool, which was specifically developed for to systematically review 9

diagnostic accuracy studies [19-21]. Meanwhile, relevant data were also extracted from each study, 10

including author, the study nation, and the description of the study population, the study design 11

characteristics, the magnetic field strength, the pulse sequences, and descriptions of interpretations of the 12

diagnostic tests. To resolve disagreement between the reviewers, a third reviewer assessed the disputed 13

material and the majority opinion was used in the analysis. 14

For each study, values for true-positive (TP), false-positive (FP), true-negative (TN), false-negative 15

(FN), sensitivity (SEN), specificity (SPE), positive likelihood ratio (PLR) and negative likelihood ratio 16

(NLR) results for the detection of lesions were extracted and 2×2 contingency tables were constructed. 17

Meta-Analysis 18

Exploring heterogeneity is important in understanding the possible factors that influence accuracy 19

estimates. It is also important in evaluating the appropriateness of statistical pooling of accuracy estimates 20

from various studies. Visual inspection of the forest plots, standard χ2-testing, and the inconsistency index 21

(I-squared, I2) were used to estimate the heterogeneity of the individual studies using Stata software (Stata 22

Corporation, College Station, TX, USA). P < 0.1 or I2 > 50% suggested notable heterogeneity [22]. If 23

Page 7 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



8

notable heterogeneities were detected, the test performance was pooled by using a random-effects 1

coefficient binary regression model; otherwise, a fixed-effects coefficient binary regression model was 2

used [23]. 3

In diagnostic test, one of the primary causes of heterogeneity is the threshold effect, which arises 4

when different cut-offs, or thresholds, are used to define a positive (or negative) test result among 5

different studies. When a threshold effect exists, there is a negative correlation between the sensitivities 6

and specificities [24-26]. The Spearman correlation coefficient between the logit of sensitivity and the 7

logit of (1-specificity) was computed to assess the threshold effect using Meta-Disc version 1.4. A 8

strongly positive correlation suggests a threshold effect, P < 0.05. We constructed hierarchical summary 9

receiver operating characteristic (HSROC) curves to assess SEN and SPE [27]. The areas under the ROC 10

curves (AUC) were used to analyse the diagnostic precision of DWI in differentiating thyroid nodules. 11

Along with variations due to the threshold effect in a review article, there are several other factors 12

that can result in variations in accuracy estimates amongst different test accuracy studies. In this study, 13

meta-regression was used to identify such heterogeneity by comparing the accuracy measurement to study 14

level co-variates (study nation, study design, MRI field strength, reference standard, enrolment, disease 15

spectrum, patient spectrum, or b value). Then, subgroup analyses were performed. The following 16

subgroups were created: (a) the b value; (b) studies with a prospective or retrospective design; (c) 17

magnetic field strengths; (d) reference standard; (f) enrolment. 18

With the Stata software, the presence of publication bias was assessed by producing a Deeks funnel 19

plot and an asymmetry test. Publication bias [28 29] was considered to be present if there was a nonzero 20

slope coefficient (P < 0.05), which suggested that only small studies reporting high accuracy had been 21

published and small studies reporting lower accuracy had likely not been published (P > 0.1), which 22

suggested that there was no evidence of notable publication bias. 23

Page 8 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



9

RESULTS 1

The search initially yielded 83 potential literature citations through database searching; 4 additional 2

results were identified through grey literature searching (Fig 1). After reviewing the titles and abstracts, 3

51 of these studies were excluded because they were duplicate publications or not relevant. After reading 4

the full texts, 17 of the remaining 32 articles were excluded because either (1) the patients had previously 5

undergone treatment, (2) the article lacked sufficient information to complete a 2×2 contingency table, or 6

(3) the study was not published in English or Chinese. Following the final screening, 15 published studies 7

met all of our inclusion and exclusion criteria. The abstracted data of these individual studies is 8

summarized in Tab 1. The quality assessment was moderate in 15 studies according to QUADAS-2 items; 9

the results of the distribution of the study design are shown in Fig 2. 10

Significant heterogeneity was found in the pooled analysis (I2

= 54.5%, P = 0.055). Therefore, SEN, 11

SPE, PLR, NLR were pooled by using a random-effects coefficient binary regression model. The pooled 12

weighted values were determined to be SEN: 0.90 (95% confidence interval (CI): 0.85, 0. 93); SPE: 0.95 13

(95% CI: 0.88, 0.98); PLR: 16.49 (95% CI: 7.37, 36.86); NLR: 0.11 (95% CI: 0.08, 0.16); DOR: 150.73 14

(95% CI: 64.96, 349.75). The AUC was 0.95 (95% CI: 0.93, 0.97). The forest plots and HSROC curves 15

from 15 studies are shown in Fig 3-5. 16

A Spearman rank correlation was performed as a further test for the threshold effect and was 17

determined to be 0.081 (P = 0.775), which indicated that there was an absence of a notable threshold 18

effect in the accuracy estimates among individual studies. 19

The results of meta-regression indicated that study nation, study design, MRI field strength, 20

reference standard, enrolment, disease spectrum, patient spectrum, and b values were not strongly 21

correlated with accuracy. The SEN and SPE estimates for the different subgroups are presented in Table 22

2. 23

Page 9 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



10

The results of Deeks funnel plot asymmetry test (P = 0.786) showed no evidence of notable 1

publication bias (Fig 6). 2

3

DISCUSSION 4

Thyroid nodules are highly prevalent and clinically difficult to manage. Compared with benign 5

thyroid nodules, malignant thyroid nodules have larger nuclei, more dense stroma, and higher cell counts, 6

all of which lead to increased cellularity and a small extracellular space [30]. Many studies[11-14 18 31] 7

and a systematic review [32] demonstrated that the ADCs of malignant thyroid nodules are significantly 8

smaller than those of benign nodules. A meta-analysis [33], which included seven studies regarding the 9

potential of DWI to differentiate between malignant and benign thyroid nodules was published in 2014 10

and suggests that DWI can be used as a diagnostic tool in distinguishing benign from malignant thyroid 11

nodules by measuring the ADC. In our study, eight additional references were included that were not 12

present in the aforementioned meta-analysis. Compared with previous meta-analysis, we analysed the 13

technical aspects of DWI and its additional value in tumour characterisation and presented a new point 14

that using a high b value may provide better results. The result of our meta-analysis showed that the 15

pooled weighted sensitivity and specificity of 15 studies were 0.90 (95% CI: 0.85, 0. 93) and 0.95 (95% 16

CI: 0.88, 0.98), respectively. These results demonstrated that DWI has a high sensitivity and specificity in 17

differentiating malignant from benign thyroid nodules. 18

The DOR test determines the ratio of the odds of positivity in correctly diagnosing a disease relative 19

to the odds of positivity obtained in non-diseased patients (false-positives). The DOR is closely linked to 20

existing indicators and is particularly applicable to meta-analysis studies of diagnostic test performance. 21

Because the DOR is derived from logistic models, it is possible to include additional variables to correct 22

Page 10 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



11

for heterogeneity [34]. In our meta-analysis, we found that the DOR estimate for DWI was 150.73 (95% 1

CI: 64.96, 349.75), indicating that DWI is an accurate modality in detecting malignant thyroid lesions. 2

We observed that the Spearman correlation coefficient was 0.081 (P = 0.775), which indicated that 3

no significant threshold effect exists. Our results also indicated that none of the impact factors in 4

meta-regression analysis contributed to the observed heterogeneity. To determine whether there are other 5

sources of heterogeneity, a subgroup analysis must be performed to detect the factors that impact 6

heterogeneity. 7

The b value is a very important factor that affects image quality and ADC measurements. When low 8

b values are applied, the ADC values tend to be higher due to the contribution of perfusion. Applying high 9

maximum b-values may be preferable when ADC measurements are performed to differentiate malignant 10

versus benign tissues exclusively by their water diffusion characteristics. However, signal-to-noise ratios 11

decrease as the b value increases, thus limiting the maximum b value. There are six studies included in 12

this review that use 3 or more pairs of b values to compute the ADC. Three of these six studies 13

demonstrated that lower b values had a higher sensitivity and accuracy in differentiating benign versus 14

malignant nodules, while the remaining studies had opposite results. In our subgroup analysis, the results 15

demonstrated that the pooled DORs of the subgroup 300, 500, 1000 s/m2 were 47.04 (95% CI: 11.55, 16

190.54), 53.13 (95% CI: 17.95, 219.34), 115.21 (95% CI: 28.42, 298.76), respectively. Higher b value 17

subgroups may be more accurate than lower b value subgroups. However, there were no notable 18

differences in the DOR and AUC among these subgroups. 19

Some inherent limitations exist in our study and should be considered when interpreting our results. 20

First, most studies included in our meta-analysis are from Asian countries, while no studies were from 21

Europe or North. Some studies [35 36] have noted that selective reporting is higher among Chinese 22

researchers than elsewhere, in a number of fields. This issue may represent one source of heterogeneity. 23

Page 11 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



12

Second, the sample size of these studies is relatively small, which is a particular problem in diagnostic 1

studies [37]. This may result in an overestimation of the diagnostic accuracy, particularly in studies 2

including non-representative samples of patients and invalid reference standards [38]. Third, our 3

meta-analysis was based only on published studies, which are prone to report positive or significant 4

results; the studies whose results are not significant or negative are often rejected or are not even 5

submitted. Although it is suggested that the quality of the data reported in articles accepted for publication 6

in peer-reviewed journals is superior to the quality of unpublished data [39], including only published 7

studies may ultimately lead to reporting bias. 8

Conclusions 9

In conclusion, DWI has a high sensitivity and specificity, is a reliable, non-invasive, and 10

no-radiation-exposure imaging modality for the detection of thyroid nodules. Using a high b value may 11

provide better results. Nevertheless, large-scale trials are necessary to assess its clinical value and to 12

establish diagnostic standards regarding b values and DWI cut-off values. 13

14

15

16

Contributions Statement 17

Chen LH participated in quality assessment, data extraction and analysis, performed the statistical 18

analysis, and drafted the manuscript. 19

Xu Jian participated in quality assessment, data extraction and analysis, performed the statistical analysis, 20

and drafted the manuscript. 21

Bao Jing participated in literature search and performed the statistical analysis. 22

Huang Xuequan participated in literature search and selection of articles. 23

Page 12 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



13

Xia Yunbao participated in selection of articles, participated in its design and coordination. 1

Wang J conceived of the study, participated in its design, and edited manuscript. 2

3

4

Competing interest 5

None declared. 6

7

Funding 8

No funding supported the work. 9

10

Data sharing statement 11

No additional data available. 12

13

Page 13 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



14

References 1

1. Bozgeyik Z, Coskun S, Dagli AF, et al. Diffusion-weighted MR imaging of thyroid nodules. Neuroradiology 2009;51(3):193-98 2

2. Erdem G, Erdem T, Karakas HM, et al. Diffusion-weighted images differentiate benign from malignant thyroid nodules. 3

Journal of Magnetic Resonance Imaging 2010;31(1):94-100 4

3. Koh DM, Collins DJ. Diffusion-weighted MRI in the body: applications and challenges in oncology. AJR Am J Roentgenol 5

2007;188(6):1622-35 doi: 10.2214/ajr.06.1403[published Online First: Epub Date]|. 6

4. Henzler T, Schmid-Bindert G, Schoenberg SO, et al. Diffusion and perfusion MRI of the lung and mediastinum. European 7

Journal of Radiology 2010;76(3):329-36 8

5. Elshafey R, Elattar A, Mlees M, et al. Role of quantitative diffusion-weighted MRI and 1H MR spectroscopy in distinguishing 9

between benign and malignant thyroid nodules. The Egyptian Journal of Radiology and Nuclear Medicine 10

2014;45(1):89-96 doi: 10.1016/j.ejrnm.2013.10.001[published Online First: Epub Date]|. 11

6. Wu Y, Yue X, Shen W, et al. Diagnostic value of diffusion-weighted MR imaging in thyroid disease: application in 12

differentiating benign from malignant disease. BMC medical imaging 2013;13(1):23 doi: 13

10.1186/1471-2342-13-23[published Online First: Epub Date]|. 14

7. Shi HF, Feng Q, Qiang JW, et al. Utility of diffusion-weighted imaging in differentiating malignant from benign thyroid 15

nodules with magnetic resonance imaging and pathologic correlation. J Comput Assist Tomogr 2013;37(4):505-10 doi: 16

10.1097/RCT.0b013e31828d28f0[published Online First: Epub Date]|. 17

8. Ilica AT, Artas H, Ayan A, et al. Initial experience of 3 Tesla apparent diffusion coefficient values in differentiating benign and 18

malignant thyroid nodules. Journal of Magnetic Resonance Imaging 2013;37(5):1077-82 19

9. Yue XH, Tao XF, Gao X. Application of diffusion-weighted MR imaging in the diagnosis of thyroid disease. Chinese Journal of 20

Radiology (China) 2012;46(6):500-04 21

10. Nakahira M, Saito N, Murata S, et al. Quantitative diffusion-weighted magnetic resonance imaging as a powerful adjunct 22

to fine needle aspiration cytology for assessment of thyroid nodules. American journal of otolaryngology 23

2012;33(4):408-16 doi: 10.1016/j.amjoto.2011.10.013[published Online First: Epub Date]|. 24

11. Mutlu H, Sivrioglu AK, Sonmez G, et al. Role of apparent diffusion coefficient values and diffusion-weighted magnetic 25

resonance imaging in differentiation between benign and malignant thyroid nodules. Clinical Imaging 2012;36(1):1-7 26

12. El-Hariri MA, Gouhar GK, Said NS, et al. Role of diffusion-weighted imaging with ADC mapping and in vivo 1H-MR 27

spectroscopy in thyroid nodules. Egyptian Journal of Radiology and Nuclear Medicine 2012;43(2):183-92 28

13. Aydin H, Kizilgoz V, Tatar I, et al. The role of proton MR spectroscopy and apparent diffusion coefficient values in the 29

diagnosis of malignant thyroid nodules: preliminary results. Clin Imaging 2012;36(4):323-33 doi: 30

10.1016/j.clinimag.2011.09.009[published Online First: Epub Date]|. 31

14. Yan B, Liu HJ, Wang CB, et al. ADC values in differentiation of benign and malignant thyroid nodules. Chinese Journal of 32

Medical Imaging Technology 2011;27(3):510-14 33

15. Schueller-Weidekamm C, Kaserer K, Schueller G, et al. Can quantitative diffusion-weighted MR imaging differentiate 34

benign and malignant cold thyroid nodules? Initial results in 25 patients. American Journal of Neuroradiology 35

2009;30(2):417-22 36

16. Ren S, Liu CH, Bai RJ. [Value of diffusion weighted imaging in diagnosis of nodular lesions of thyroid: a preliminary study]. 37

Zhonghua yi xue za zhi 2010;90(47):3351-54 38

17. Li RK QJ, Liu W, et al. Application of MR diffusion weighted imaging in the differentiation of malignant from benign thyroid 39

focal lesions. Radiol Practice (China) 2009; 24(7):719-22. 40

18. Razek AA, Sadek AG, Kombar OR, et al. Role of apparent diffusion coefficient values in differentiation between malignant 41

and benign solitary thyroid nodules. AJNR American journal of neuroradiology 2008;29(3):563-8 doi: 42

10.3174/ajnr.A0849[published Online First: Epub Date]|. 43

19. Whiting P, Rutjes AW, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of 44

diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003;3:25 doi: 10.1186/1471-2288-3-25 45

Page 14 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



15

[doi] 1

1471-2288-3-25 [pii][published Online First: Epub Date]|. 2

20. Whiting PF, Weswood ME, Rutjes AW, et al. Evaluation of QUADAS, a tool for the quality assessment of diagnostic accuracy 3

studies. BMC Med Res Methodol 2006;6:9 doi: 10.1186/1471-2288-6-9[published Online First: Epub Date]|. 4

21. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy 5

studies. Ann Intern Med 2011;155(8):529-36 doi: 10.7326/0003-4819-155-8-201110180-00009[published Online 6

First: Epub Date]|. 7

22. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ 2003;327(7414):557-60 doi: 8

10.1136/bmj.327.7414.557[published Online First: Epub Date]|. 9

23. Leeflang MM, Deeks JJ, Gatsonis C, et al. Systematic reviews of diagnostic test accuracy. Annals of internal medicine 10

2008;149(12):889-97 11

24. Zamora J, Abraira V, Muriel A, et al. Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Med Res 12

Methodol 2006;6:31 doi: 10.1186/1471-2288-6-31[published Online First: Epub Date]|. 13

25. Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: 14

data-analytic approaches and some additional considerations. Stat Med 1993;12(14):1293-316 15

26. Reitsma JB, Glas AS, Rutjes AW, et al. Bivariate analysis of sensitivity and specificity produces informative summary 16

measures in diagnostic reviews. Journal of clinical epidemiology 2005;58(10):982-90 doi: 17

10.1016/j.jclinepi.2005.02.022[published Online First: Epub Date]|. 18

27. Menke J. Bivariate random-effects meta-analysis of sensitivity and specificity with SAS PROC GLIMMIX. Methods of 19

information in medicine 2010;49(1):54-62, 62-4 doi: 10.3414/me09-01-0001[published Online First: Epub Date]|. 20

28. StataCorp. Stata Statistical Software: Release 10. . Texas: College Station 2009 21

29. Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic 22

reviews of diagnostic test accuracy was assessed. Journal of Clinical Epidemiology 2005;58(9):882-93 23

30. Schueller-Weidekamm C, Schueller G, Kaserer K, et al. Diagnostic value of sonography, ultrasound-guided fine-needle 24

aspiration cytology, and diffusion-weighted MRI in the characterization of cold thyroid nodules. European Journal of 25

Radiology 2010;73(3):538-44 26

31. Jin Y, Qiang JW, Feng Q, et al. MR diffusion weighted imaging of thyroid nodules: Correlation with pathology. Chinese 27

Journal of Medical Imaging Technology 2012;28(2):256-60 28

32. Vermoolen MA, Kwee TC, Nievelstein RA. Apparent diffusion coefficient measurements in the differentiation between 29

benign and malignant lesions: a systematic review. Insights into imaging 2012;3(4):395-409 doi: 30

10.1007/s13244-012-0175-y[published Online First: Epub Date]|. 31

33. Wu LM, Chen XX, Li YL, et al. On the utility of quantitative diffusion-weighted MR imaging as a tool in differentiation 32

between malignant and benign thyroid nodules. Acad Radiol 2014;21(3):355-63 doi: 33

10.1016/j.acra.2013.10.008[published Online First: Epub Date]|. 34

34. Glas AS, Lijmer JG, Prins MH, et al. The diagnostic odds ratio: a single indicator of test performance. J Clin Epidemiol 35

2003;56(11):1129-35 36

35. Pan Z, Trikalinos TA, Kavvoura FK, et al. Local literature bias in genetic epidemiology: an empirical evaluation of the 37

Chinese literature. PLoS medicine 2005;2(12):e334 doi: 10.1371/journal.pmed.0020334[published Online First: Epub 38

Date]|. 39

36. Andrew Vickers NG, Robert Harland, and Rebecca Rees. Do certain countries produce only positive results? A systematic 40

review of controlled trials. Controlled Clinical Trials 1998;19:159-66 41

37. Rutjes AW, Reitsma JB, Di Nisio M, et al. Evidence of bias and variation in diagnostic accuracy studies. CMAJ : Canadian 42

Medical Association journal = journal de l'Association medicale canadienne 2006;174(4):469-76 doi: 43

10.1503/cmaj.050090[published Online First: Epub Date]|. 44

38. Brazzelli M, Sandercock PA, Chappell FM, et al. Magnetic resonance imaging versus computed tomography for detection 45

of acute vascular lesions in patients presenting with stroke symptoms. Cochrane database of systematic reviews 46

Page 15 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



16

(Online) 2009(4):CD007424 doi: 10.1002/14651858.CD007424.pub2[published Online First: Epub Date]|. 1

39. McAuley L, Ba'Pham, Tugwell P, et al. Does the inclusion of grey literature influence estimates of intervention 2

effectiveness reported in meta-analyses? Lancet 2000;356(9237):1228-31 3

4

5

6

Page 16 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



17

Figure legends 1

Figure 1. Flowchart illustrating the selection of studies. 2

Figure 2. Methodological quality of the 15 included studies. A. Methodological quality graph: each 3

methodological quality item is presented as percentages across all included studies; B. Methodological 4

quality summary. 5

Figure 3. Forest plots of the SEN and SPE with corresponding 95% CIs for DWI in the detection of 6

thyroid nodules. 7

Figure 4. Forest plots of the dOR with corresponding 95% CIs for DWI in the detection of thyroid 8

nodules. 9

Figure 5. Summary receiver operating characteristic (SROC) curves from the bivariate model of DWI in 10

the detection of thyroid nodules. 11

Figure 6. The funnel plot of publication bias. Linear regression of the inverse root of effective sample 12

sizes (ESS) on a log dOR is performed as a test for funnel plot asymmetry. 13

14

15

Supporting Information Legends 16

Checklist S1. PRISMA 2009 checklist. 17

Table S2. Data of included studies. 18

19

Page 17 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



18

Table 1. Characteristics of included studies 1

Notes: ND, not mentioned; pro, prospective; retro, retrospective; Both, Histopathologic and FNAB. 2

3

Study Year Nation SEN SPE Field Enrollment Design Reference Blind b-value

Razek [18] 2008 Egypt 98% 92% 1.5T consecutive pro Histopathologic ND 500

Li [17] 2009 China 79% 86% 1.5T ND retro Histopathologic ND 150

86% 77% 1.5T ND retro Histopathologic ND 300


Schueller [15] 2009 Austria 85% 100% 1.0T ND pro Both unblind 800

Bozgeyik [1] 2009 Turkey 89% 100% 1.5T consecutive pro FNAB blind 100

100% 80% 1.5T consecutive pro FNAB blind 200

90% 100% 1.5T consecutive pro FNAB blind 300

Ren [16] 2010 China 90% 90% 1.5T consecutive retro Histopathologic ND 100

83% 90% 1.5T consecutive retro Histopathologic ND 200



Yan [14] 2011 China 87% 100% 1.5T ND retro Histopathologic blind 500

Aydin [13] 2012 Turkey 92% 75% 1.5T consecutive retro Histopathologic ND 400

El-Hariri [12] 2012 Egypt 94% 95% 1.5T ND pro Both ND 500

Mutlu [11] 2012 Turkey 80% 97% 1.5T ND pro Both ND 1000

Nakahira [10] 2012 Japan 94% 83% 1.5T ND retro Histopathologic blind 1000

Yue [9] 2012 China 86% 79% 1.5T ND retro Histopathologic blind 300

Ilica [8] 2013 Turkey 90% 100% 3T ND pro Both ND 1000

Shi [7] 2013 China 92% 88% 1.5T ND retro Histopathologic ND 500

Wu [6] 2013 China 77% 100% 1.5T ND retro Histopathologic blind 300

68% 64% 1.5T ND retro Histopathologic blind 500

54% 71% 1.5T ND retro Histopathologic blind 800

Elshafey [5] 2013 Egypt 96% 92% 1.5T ND pro Both blind 1000

Page 18 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



19

Table 2. Sensitivity and Specificity Estimates for Each Subgroup 1

Subgroup No. of Studies Mean SEN (%) Mean SPE (%) DOR AUC (%)

B value (s/m2)

300 4 90(80-96) 88(83-93) 47(11-190) 91(88-95)

500 6 92(86-96) 82(77-86) 53(12-239) 93(90-96)

1000 4 88(81-93) 93(84-98) 115(30-446) 96(93-99)

Magnetic field

1.0T 1 85 100 NA NA

1.5T 13 89(85-93) 88(85-91) 64(26-156) 95(91-99)

3.0T 1 90 100 NA NA

Study design

retrospective 8 91(86-94) 83(78-88) 33(13-85) 93(89-97)

prospective 7 87(81-92) 98(94-99) 153(76-446) 97(95-99)

Blinded

yes 9 90(85-93) 88(83-92) 69(32-147) 95(93-97)

unknown or not 6 88(79-94) 90(85-94) 66(12-365) 94(89-98)

Note.—Numbers in parentheses are the 95% CIs. NA = not applicable. 2

3

Page 19 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



185x208mm (300 x 300 DPI)

Page 20 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



82x81mm (300 x 300 DPI)

Page 21 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



121x60mm (300 x 300 DPI)

Page 22 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



80x58mm (300 x 300 DPI)

Page 23 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



80x64mm (300 x 300 DPI)

Page 24 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



80x61mm (300 x 300 DPI)

Page 25 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



PRISMA 2009 ChecklistPRISMA 2009 ChecklistPRISMA 2009 ChecklistPRISMA 2009 Checklist

Section/topic # Checklist item Reported on page #

TITLE

Title 1 Identify the report as a systematic review, meta-analysis, or both. Title

ABSTRACT

Structured summary 2 Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, participants, and interventions; study appraisal and synthesis methods; results; limitations; conclusions and implications of key findings; systematic review registration number.

Abstract

INTRODUCTION

Rationale 3 Describe the rationale for the review in the context of what is already known. Introduction

Objectives 4 Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design (PICOS).

Introduction

METHODS

Protocol and registration 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.

No

Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered,

language, publication status) used as criteria for eligibility, giving rationale.

Selection of articles

Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.

Literature search

Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

Literature search

Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable,

included in the meta-analysis).


Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

data extraction

Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.

data extraction

Risk of bias in individual studies

12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

Quality assessment

Summary measures 13 State the principal summary measures (e.g., risk ratio, difference in means). Meta-analysis

Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency

(e.g., I2) for each meta-analysis.

Meta-analysis

Page 26 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on December 29, 2020 by guest. Protected by copyright. http://bmjopen.bmj.com/ BMJ Open: first published as 10.1136/bmjopen-2015-008413 on 5 January 2016. Downloaded from



PRISMA 2009 ChecklistPRISMA 2009 ChecklistPRISMA 2009 ChecklistPRISMA 2009 Checklist

Page 1 of 2


Risk of bias across studies 15 Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting within studies).

Quality assessment

Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done,

indicating which were pre-specified.

Meta-analysis

RESULTS

Study selection 17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.

Results

Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations.

Results

Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12). Results

Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

Results

Synthesis of results 21 Present results of each meta-analysis done, including confidence intervals and measures of consistency. Results

Risk of bias across studies 22 Present results of any assessment of risk of bias across studies (see Item 15). Results

Additional analysis 23 Give results of additional analyses, if done (e.g., sensitivity or subgroup analyses, meta-regression [see Item 16]). Results

DISCUSSION

Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

Discussion

Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).

Discussion

Conclusions 26 Provide a general interpretation of the results in the context of other evidence, and implications for future research. Discussion

FUNDING

Funding 27 Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the systematic review.

Funding Statement

From: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097

For more information, visit: www.prisma-statement.org.

Page 2 of 2

Page 27 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on December 29, 2020 by guest. Protected by copyright. http://bmjopen.bmj.com/ BMJ Open: first published as 10.1136/bmjopen-2015-008413 on 5 January 2016. Downloaded from



Table S2. Data of included studies

Notes: ND, not mentioned; ADC, apparent diffusion coefficient.

Study (Ref) Year SEN SPE ADCmean

(Benign)

ADCmean

(Malignant)

ADCmean

(Normal) Threshold

Razek [18] 2008 98% 92% 1.80±0.27 0.73±0.19 ND 0.98

Li [17] 2009 79% 86% 2.56±0.56 1.75±0.43 ND 1.99

86% 77% 1.99±0.51 1.40±0.31 ND 1.58

93% 58% 1.68±0.54 1.24±0.30 ND 1.41

Schueller [15] 2009 85% 100% 3.46±0.40 2.73±0.65 ND 2.25

Bozgeyik [1] 2009 89% 100% 3.06±0.71 0.96±0.65 2.98±0.67 1.45

100% 80% 1.80±0.60 0.56±0.43 1.73±0.58 0.65

90% 100% 1.15±0.43 0.30±0.20 1.17±0.46 0.36

Ren [16] 2010 90% 90% 3.00±0.50 2.00±0.50 ND 2.56

83% 90% 2.60±0.40 1.70±0.50 ND 2.17

93% 83% 2.20±0.40 1.30±0.40 ND 1.81

93% 97% 2.30±0.30 1.00±0.40 ND 1.48

Yan [14] 2011 87% 100% 2.20±0.40 1.20±0.30 ND 1.49

Aydin [13] 2012 92% 75% 0.06±0.02 0.12±0.07 ND 0.04

El-Hariri [12] 2012 94% 95% 1.85±0.24 0.89±0.27 ND 1.50

Mutlu [11] 2012 80% 97% 1.60±0.10 0.80±0.20 0.98±0.28 1.00

Nakahira [10] 2012 94% 83% 1.93±0.37 1.20±0.25 1.41±0.14 1.60

Yue [9] 2012 86% 79% 2.39±0.38 1.60±0.56 ND 1.98

Ilica [8] 2013 90% 100% 1.55±0.35 0.81±0.18 1.32±0.21 0.91

Shi [7] 2013 92% 88% 1.89±0.64 1.16±0.32 ND 1.70

Wu [6] 2013 77% 100% 2.37±0.47 1.49±0.60 ND 2.17

68% 64% 1.87±0.25 1.61±0.45 ND 1.74

54% 71% 1.68±0.25 1.41±0.34 ND 1.65

Elshafey [5] 2013 96% 92% 1.78±0.21 0.59±0.24 ND 0.80

Page 28 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



Diffusion-weighted magnetic resonance imaging in differentiating malignant from benign thyroid nodules: A

meta-analysis

Journal: BMJ Open

Manuscript ID bmjopen-2015-008413.R1

Article Type: Research

Date Submitted by the Author: 02-Sep-2015

Complete List of Authors: Chen, Lihua; Southwest Hospital, Department of Radiology Xu, Jian; PLA101 Hospital, Department of General Surgery Bao, Jing; Wuxi center for disease control and prevention, Molecular

biology lab Huang, Xuequan; Southwest Hospital, Department of Radiology Hu, Xiaofei; Southwest Hospital, Department of Radiology Xia, Yunbao; PLA101 Hospital, Department of Radiology Wang, Jian; Southwest Hospital, Department of Radiology

<b>Primary Subject Heading</b>:

Radiology and imaging

Secondary Subject Heading: Diagnostics, Evidence based practice

Keywords: Magnetic resonance imaging < RADIOTHERAPY, thyroid nodules, Diffusion-weighted imaging, Meta-analysis


BMJ Open on D

ecember 29, 2020 by guest. P

rotected by copyright.http://bm

jopen.bmj.com

/B

MJ O

pen: first published as 10.1136/bmjopen-2015-008413 on 5 January 2016. D

ownloaded from



1

Title Page 1

Title: Diffusion-weighted magnetic resonance imaging in differentiating malignant from benign thyroid 2

nodules: A meta-analysis. 3

4

Authors: Lihua Chen1,2

, †MS; Jian Xu3 †MS; Jing Bao

4, MS; Xuequan Huang

1, MD; Xiaofei Hu

1, MS；5

Yunbao Xia2, *BS; Jian Wang

1, *MD 6

7

Lihua Chen, 1Department of Radiology, Southwest Hospital, Third Military Medical University, 8

Chongqing, 400038, China; 2Department of Radiology, Taihu Hospital, Wuxi, Jiangsu Province, 214044, 9

China, E-mail: [email protected]. 10

Jian Xu, Department of General Surgery, Taihu Hospital, Wuxi, Jiangsu Province, 214044, China, E-mail: 11


Jing Bao, Molecular Biology Lab, Wuxi Center for Disease Control and Prevention, Wuxi, Jiangsu 13

Province, 214001, China, E-mail: [email protected]. 14

Xuequan Huang, Department of Radiology, Southwest Hospital, Third Military Medical University, 15


Xiaofei Hu, Department of Radiology, Southwest Hospital, Third Military Medical University, 17


Yunbao Xia, Department of Radiology, Taihu Hospital, Wuxi, Jiangsu Province, 214044, China, E-mail: 19


Jian Wang, Department of Radiology, Southwest Hospital, Third Military Medical University, Chongqing, 21


23

Page 1 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



2

† Co-first authors 1

* Co-corresponding authors 2

Correspondence to: 3

Jian Wang 4

Department of Radiology 5

Southwest Hospital 6

Third Military Medical University 7

Chongqing, 400038, China 8

Tel: +86-23-6875-4419 9

Fax: +86-23-6546-3026 10


Or 12

Yunbao Xia 13

Department of Radiology 14

Taihu Hospital 15

Wuxi, 214044, Jiangsu Province, China 16

Tel: +86-510-8514-2222 17

Fax: +86-510-8514-2222 18


20

21

Page 2 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



3

Abstract 1

OBJECTIVES: To perform a meta-analysis to evaluate the diagnostic efficacy of diffusion-weighted 2

imaging (DWI) in differentiating malignant from benign thyroid nodules. 3

DESIGN: A meta-analysis. 4

DATA SOURCES AND STUDY SELECTION: Medical and scientific literature databases were 5

searched for original articles published up to August 2015. Studies were selected if they (1) included a 6

diagnostic DWI for differentiating malignant from benign thyroid lesions, (2) included patients who later 7

underwent biopsy, and (3) presented sufficient data to enable the construction of contingency tables. 8

DATA SYNTHESIS: For each study, the true-positive, false-positive, true-negative, and false-negative 9

values were extracted or derived, and 2×2 contingency tables were constructed. Methodological quality 10

was assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) instrument. The 11

heterogeneity test, the threshold effect test, subgroup analyses, and publication bias analyses were 12

performed. 13

RESULTS: From the 113 identified search results, 15 studies, representing a total of 765 lesions, were 14

included in the meta-analysis. We detected heterogeneity between studies but found no evidence of 15

publication bias. The methodological quality was moderate. The pooled weighted sensitivity was 0.90 (95% 16

confidence interval (CI): 0.85, 0.93); the specificity was 0.95 (95% CI: 0.88, 0.98); the positive likelihood 17

ratio was 16.49 (95% CI: 7.37, 36.86); the negative likelihood ratio was 0.11 (95% CI: 0.08, 0.16); and 18

the diagnostic odds ratio was 150.73 (95% CI: 64.96, 349.75). The area under the receiver operator 19

characteristic curve was 0.95 (95% CI: 0.93, 0.97). 20

CONCLUSIONS: Quantitative DWI may be a non-invasive, non-radiative, and accurate method of 21

distinguishing malignant from benign thyroid nodules. Nevertheless, large-scale trials are necessary to 22

Page 3 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



4

assess its clinical value and to establish standards regarding b values and cut-off values for DWI-based 1

diagnosis. 2

3

Keywords: Magnetic resonance imaging, Thyroid nodules, Diffusion-weighted imaging, Meta-analysis 4

5

Strengths and limitations of this study 6

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement was used to 7

improve the reporting of our research. 8

Hierarchical summary receiver operating characteristic (HSROC) curves was constructed to assess SEN 9

and SPE. 10

We presented a new point that using a high b value may provide higher diagnostic accuracy. 11

Studies included in our meta-analysis lacked a description of ADC reproducibility and the sample size of 12

included studies was relatively small. 13

14

Abbreviations: 15

AUC area under the curve 16

CI confidence interval 17

FN false-negative 18

FP false-positive 19

HSROC hierarchical summary receiver operating characteristic 20

MR magnetic resonance 21

NLR negative likelihood ratio 22

PLR positive likelihood ratio 23

Page 4 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



5

QUADAS Quality Assessment of Diagnostic Accuracy Studies 1

SEN sensitivity 2

SPE specificity 3

TN true-negative 4

TP true-positive 5

6

7

8

Page 5 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



6

Introduction 1

Thyroid nodules, the most common pathology involving the thyroid gland, consist of discrete lesions 2

within the thyroid gland that are often palpable and are typically sonographically distinct from the 3

surrounding thyroid parenchyma [1]. Less than 5% of palpable thyroid nodules are malignant; however, 4

these nodules must be distinguished from benign thyroid nodules to correctly and efficaciously treat 5

patients suffering from this pathology [2]. 6

Because clinical findings do not provide a definitive diagnosis, several useful, non-invasive imaging 7

tests (such as ultrasonography [US] and radionuclide scintigraphy) can be used to determine which 8

nodules should be histopathologically evaluated to rule out the possibility of thyroid malignancy. US has 9

been used as a first step in the assessment of these nodules, but no single US criterion has been 10

demonstrated to accurately differentiate benign nodules from malignant nodules. Furthermore, the 11

hazards associated with radiation exposure during radionuclide scintigraphy are unavoidable, and some 12

functioning nodules (hot nodules) found on scintigraphy are malignant. 13

Diffusion-weighted imaging (DWI) is a type of functional MRI that is based on the diffusion of 14

water molecules through the tissue of interest (i.e., tumor tissue). DWI can provide crucial information 15

regarding the molecular profile of the underlying pathology and pathophysiologic mechanisms [3]. 16

Specifically, the diffusion of water molecules in malignant tumors is restricted, which results in a 17

decreased apparent diffusion coefficient (ADC); this difference in the ADC facilitates the differentiation 18

of benign tumors from malignant tumors [4]. 19

Many studies [1 5-18] have shown that DWI has the potential to differentiate benign from malignant 20

thyroid nodules. However, the sample sizes of these studies were relatively small, and the findings have 21

been inconclusive. The aim of this study was to systematically review all of the studies related to the 22

ability of DWI to differentiate benign from malignant thyroid nodules. Moreover, based on the extracted 23

Page 6 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



7

data, an analysis of the technical aspects of DWI and its additional value for tumor characterization is 1

presented. 2

3

Methods 4

As this meta-analysis was based on previously published studies, ethical approval was not necessary. 5

Literature Search 6

The PubMed, Embase, Cochrane Library, and China National Knowledge Infrastructure (CNKI) 7

databases were searched by two independent observers. The terms “Diffusion-Weighted Imaging [MeSH]” 8

or “DWI” were used for the diagnostic test, and the terms “thyroid nodules [MeSH]”, “thyroid lesions,” 9

or “thyroid” were used for the clinical domain (Search Strategy S1). We limited our search to publications 10

that met the following criteria: published in the English or Chinese language; the presence of the search 11

term within the title or abstract of the article; and a publication date no later than May 2014. Review 12

articles, letters, comments, case reports, and unpublished articles were excluded. The reference lists of all 13

retrieved articles were manually cross-checked. 14

Selection of Articles 15

Two authors initially screened the titles and abstracts of the search results and retrieved the full texts 16

of all potentially relevant reports. Next, the authors independently reviewed all relevant reports according 17

to the predefined inclusion criteria. Disagreements were resolved by consensus or arbitration by a third 18

author who assessed all of the involved items. The majority opinion was used to determine whether a 19

particular study met the selection criteria. 20

Studies were considered as eligible if the following criteria were met: (a) the study included a 21

diagnostic DWI for differentiating malignant from benign thyroid lesions; (b) the controls underwent 22

Page 7 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



8

histopathologic analysis (surgery/biopsy) and/or follow-up analysis; and (c) the data were sufficient to 1

accurately determine the true-positive or false-negative results. 2

Studies were excluded if (a) there were fewer than 20 patients; (b) multiple reports were published 3

for the same study population (in this case, the most detailed or recent publication was chosen); or (c) the 4

study included patients who had previously undergone treatment for thyroid lesions. 5

Quality Assessment and Data Extraction 6

The aforementioned three authors extracted data from the selected reports. The methodological 7

quality of the included studies was independently assessed by two observers using the Quality 8

Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool, which was specifically developed to 9

systematically review the diagnostic accuracy of studies [19-21]. Additionally, relevant data, including 10

author, the study nation, the description of the study population, the study design characteristics, the 11

magnetic field strength, the pulse sequences, and descriptions of the interpretations of the diagnostic tests, 12

were extracted from each study. To resolve disagreements between the reviewers, a third reviewer 13

assessed the disputed material, and the majority opinion was used in the analysis. 14

For each study, the estimated true-positive (TP), false-positive (FP), true-negative (TN), and 15

false-negative (FN) values, sensitivity (SEN), specificity (SPE), positive likelihood ratio (PLR) and 16

negative likelihood ratio (NLR) for the detection of lesions were extracted, and 2×2 contingency tables 17

were constructed. 18

Meta-Analysis 19

Exploring study heterogeneity is important in understanding the possible factors that influence 20

accuracy estimates and in evaluating the appropriateness of statistical pooling of accuracy estimates from 21

various studies. Visual inspection of the forest plots, standard χ2-tests, and the inconsistency index 22

(I-squared, I2) were used to estimate the heterogeneity of the individual studies using Stata software (Stata 23

Page 8 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



9

Corporation, College Station, TX, USA). P < 0.1 or I2 > 50% suggested notable heterogeneity [22]. If 1

notable heterogeneities were detected, the relevant data was pooled using a random-effects coefficient 2

binary regression model; otherwise, a fixed-effects coefficient binary regression model was used [23]. 3

In the diagnostic test, one primary cause of heterogeneity is the threshold effect, which arises when 4

different cut-off values, or thresholds, are used to define a positive (or negative) test result between 5

different studies. When a threshold effect exists, there is a negative correlation between the SEN and the 6

SPE [24-26]. The Spearman correlation coefficient between the logit of SEN and the logit of (1-SPE) was 7

computed to assess the threshold effect using Meta-Disc version 1.4; a strongly positive correlation (P < 8

0.05) suggests a threshold effect. We constructed hierarchical summary receiver operating characteristic 9

(HSROC) curves to assess SEN and SPE [27]. The areas under the ROC curves (AUCs) were used to 10

analyze the diagnostic precision of DWI in differentiating thyroid nodules. 11

In addition to the threshold effect on systematic review results, several other factors can result in 12

variations in accuracy estimates between different test accuracy studies. In this study, meta-regression 13

was used to identify such heterogeneity by comparing the accuracy measurement to study-level 14

co-variates (study nation, study design, MRI field strength, reference standard, enrollment, disease 15

spectrum, patient spectrum, or b value). Then, subgroup analyses were performed. Stratification was 16

performed according to the following parameters: (a) the b value; (b) studies with a prospective or 17

retrospective design; (c) magnetic field strength; (d) reference standard; and (f) enrolment. 18

Using Stata software, the presence of publication bias was assessed by producing a Deeks funnel 19

plot and performing an asymmetry test. Publication bias [28 29] was considered to be present if there was 20

a nonzero slope coefficient (P < 0.05), which suggested that only small studies reporting high accuracy 21

had been published; alternatively, P > 0.1 suggested that there was no evidence of notable publication 22

bias. 23

Page 9 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



10

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement[30] was used 1

to improve the reporting of our research (Figure 1 and Checklist S2). 2

3

Results 4

The database search initially yielded 113 potential literature citations; 4 additional results were 5

identified by searching the gray literature (Fig 1). After reviewing the titles and abstracts, 48 of these 6

studies were excluded because they were duplicate publications or were not relevant to this analysis. After 7

reading the full texts, 19 of the remaining 34 articles were excluded because (1) the patients had 8

previously undergone treatment, (2) the article lacked sufficient information to complete a 2×2 9

contingency table, or (3) the study was not published in English or Chinese. Following the final screening 10

process, 15 published studies were considered to have met all of our inclusion and exclusion criteria. The 11

data abstracted from these individual studies are summarized in Table 1. The quality was moderate in 15 12

studies according to the QUADAS-2 items; the results for the distribution of the study design are shown 13

in Fig 2. 14

Significant heterogeneity was found based on the pooled analysis (I2

= 54.5%, P = 0.055). Therefore, 15

SEN, SPE, PLR, and NLR were pooled using a random-effects coefficient binary regression model. The 16

pooled weighted values were as follows: SEN: 0.90 (95% confidence interval (CI): 0.85, 0.93); SPE: 0.95 17

(95% CI: 0.88, 0.98); PLR: 16.49 (95% CI: 7.37, 36.86); NLR: 0.11 (95% CI: 0.08, 0.16); and diagnostic 18

odds ratio (DOR): 150.73 (95% CI: 64.96, 349.75). The AUC was 0.95 (95% CI: 0.93, 0.97). The forest 19

plots and HSROC curves for the 15 studies are shown in Fig 3-5. 20

A Spearman rank correlation was performed as a further assessment of the threshold effect; the 21

Spearman correlation coefficient was determined to be 0.081 (P = 0.775). This result indicated that no 22

notable threshold effect was detected in the accuracy estimates among individual studies. 23

Page 10 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



11

The results of meta-regression indicated that study nation, study design, MRI field strength, 1

reference standard, enrolment, disease spectrum, patient spectrum, and b values were not strongly 2

correlated with accuracy. The estimated SEN and SPE for each subgroup are presented in Table 2. 3

The results of the Deeks funnel plot asymmetry test (P = 0.786) showed no evidence of notable 4

publication bias (Fig 6). 5

6

Discussion 7

Thyroid nodules are highly prevalent and clinically difficult to manage. Compared with benign 8

thyroid nodules, malignant thyroid nodules have larger nuclei, denser stroma, and higher cell counts, all 9

of which lead to increased cellularity and reduced extracellular space [31]. Many studies [11-14 18 32 33] 10

and a systematic review [34] demonstrated that the ADCs of malignant thyroid nodules are significantly 11

smaller than those of benign nodules. A meta-analysis [35] that included seven studies regarding the 12

potential of DWI to differentiate between malignant and benign thyroid nodules was published in 2014 13

and suggested that DWI can be used as a diagnostic tool to distinguish benign from malignant thyroid 14

nodules by measuring the ADC. In our study, eight additional references were included that were not 15

present in the aforementioned meta-analysis. Compared with the previous meta-analysis, our 16

meta-analysis analyzed the technical aspects of DWI and its additional value for tumor characterization. 17

Our results revealed for the first time that using a high b value may provide better results. The results of 18

our meta-analysis showed that the pooled weighted SEN and SPE of the 15 included studies were 0.90 19

(95% CI: 0.85, 0. 93) and 0.95 (95% CI: 0.88, 0.98), respectively. These results demonstrated that DWI 20

has a high SEN and SPE for differentiating malignant from benign thyroid nodules. 21

The DOR represents the ratio of the odds of correctly diagnosing the diseased patients (true-positives) 22

relative to the odds of obtaining a positive result among the non-diseased patients (false-positives). The 23

Page 11 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



12

DOR is closely linked to existing indicators and is particularly applicable to meta-analyses of diagnostic 1

test performance. Because the DOR is derived from logistic models, it is possible to include additional 2

variables to correct for heterogeneity [36]. In our meta-analysis, we found that the estimated DOR for 3

DWI was 150.73 (95% CI: 64.96, 349.75). This result indicated that DWI is an accurate modality for 4

detecting malignant thyroid lesions. 5

We observed that the Spearman correlation coefficient was 0.081 (P = 0.775); this result indicated 6

that no significant threshold effect was detected. Additionally, our results indicated that none of the 7

factors potentially impacting the meta-regression analysis results contributed to the observed 8

heterogeneity. To determine whether there were other sources of heterogeneity, a subgroup analysis must 9

be performed to detect the factors that impact heterogeneity. 10

The b value is a very important factor that affects image quality and ADC measurements. When low 11

b values are applied, the ADCs tend to be higher due to the contribution of perfusion. Applying high 12

maximum b values may be preferable when ADC measurements are performed to differentiate malignant 13

from benign tissues exclusively based on their water diffusion characteristics. However, the 14

signal-to-noise ratio decreases as the b value increases, thus limiting the maximum b value. Six of the 15

studies included in this review used 3 or more pairs of b values to compute the ADC. Three of these six 16

studies demonstrated that lower b values had a higher SEN and accuracy for differentiating benign from 17

malignant nodules, whereas the remaining studies reported the opposite results. In our subgroup analysis, 18

the results demonstrated that the pooled DORs of the 300, 500, and 1000 s/m2 subgroups were 47.04 (95% 19

CI: 11.55, 190.54), 53.13 (95% CI: 17.95, 219.34), and 115.21 (95% CI: 28.42, 298.76), respectively. The 20

diagnostic accuracy may be greater in the higher b value subgroups than in the lower b value subgroups. 21

However, there were no notable differences in the DOR or the AUC between these subgroups. 22

Page 12 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



13

Our study contains several inherent limitations that should be considered when interpreting our 1

results. First, most of the studies included in our meta-analysis lacked a description of ADC 2

reproducibility and were performed in Asian countries, and no studies were from Europe or North 3

America . Some studies [37 38] have noted that selective reporting is higher among Chinese studies than 4

elsewhere across several fields. This issue may represent one source of heterogeneity. Second, the sample 5

size of these studies was relatively small, which is a particular problem in diagnostic studies [39]. This 6

limitation may result in an overestimation of the diagnostic accuracy, particularly in studies including 7

non-representative samples of patients and invalid reference standards [40]. Third, our meta-analysis was 8

based only on published studies, which are prone to report positive or significant results; the studies 9

whose results are not significant or negative are often rejected or not even submitted. Although it is 10

suggested that the quality of the data reported in articles accepted for publication in peer-reviewed 11

journals is superior to the quality of unpublished data [41], including only published studies may 12

ultimately lead to reporting bias. 13

Conclusions 14

In conclusion, DWI has a high SEN and SPE, may be a reliable, non-invasive, and non-radiative 15

imaging modality for the detection of thyroid nodules. Using a high b value may provide higher 16

diagnostic accuracy. Nevertheless, large-scale trials are necessary to assess its clinical value and to 17

establish standards regarding b values and cut-off values for DWI-based diagnosis. 18

19

20

Contributorship statement 21

Conceived and designed the experiments: Lihua Chen, Jian Xu, Yunbao Xia, Jian Wang. 22

Performed the experiments: Lihua Chen, Jian Xu, Jian Wang. 23

Page 13 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



14

Analyzed the data: Jing Bao, Xuequan Huang, Xiaofei Hu. 1

Contributed reagents/materials/analysis tools: Jing Bao, Xuequan Huang, Xiaofei Hu, Yunbao Xia. 2

Wrote the paper: Lihua Chen, Jian Xu, Jian Wang. 3

4

Acknowledgements 5

The authors thank AJE for help in English-language editing. 6

7

Competing Interests 8

None. 9

10

Funding 11

No funding supported this work. 12

13

14

Page 14 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



15

References 1

1. Bozgeyik Z, Coskun S, Dagli AF, et al. Diffusion-weighted MR imaging of thyroid nodules. Neuroradiology 2009;51(3):193-98 2

2. Erdem G, Erdem T, Karakas HM, et al. Diffusion-weighted images differentiate benign from malignant thyroid nodules. 3

Journal of Magnetic Resonance Imaging 2010;31(1):94-100 4

3. Koh DM, Collins DJ. Diffusion-weighted MRI in the body: applications and challenges in oncology. AJR Am J Roentgenol 5

2007;188(6):1622-35 doi: 10.2214/ajr.06.1403[published Online First: Epub Date]|. 6

4. Henzler T, Schmid-Bindert G, Schoenberg SO, et al. Diffusion and perfusion MRI of the lung and mediastinum. European 7

Journal of Radiology 2010;76(3):329-36 8

5. Elshafey R, Elattar A, Mlees M, et al. Role of quantitative diffusion-weighted MRI and 1H MR spectroscopy in distinguishing 9

between benign and malignant thyroid nodules. The Egyptian Journal of Radiology and Nuclear Medicine 10

2014;45(1):89-96 doi: 10.1016/j.ejrnm.2013.10.001[published Online First: Epub Date]|. 11

6. Wu Y, Yue X, Shen W, et al. Diagnostic value of diffusion-weighted MR imaging in thyroid disease: application in 12

differentiating benign from malignant disease. BMC medical imaging 2013;13(1):23 doi: 13

10.1186/1471-2342-13-23[published Online First: Epub Date]|. 14

7. Shi HF, Feng Q, Qiang JW, et al. Utility of diffusion-weighted imaging in differentiating malignant from benign thyroid 15

nodules with magnetic resonance imaging and pathologic correlation. J Comput Assist Tomogr 2013;37(4):505-10 doi: 16

10.1097/RCT.0b013e31828d28f0[published Online First: Epub Date]|. 17

8. Ilica AT, Artas H, Ayan A, et al. Initial experience of 3 Tesla apparent diffusion coefficient values in differentiating benign and 18

malignant thyroid nodules. Journal of Magnetic Resonance Imaging 2013;37(5):1077-82 19

9. Yue XH, Tao XF, Gao X. Application of diffusion-weighted MR imaging in the diagnosis of thyroid disease. Chinese Journal of 20

Radiology (China) 2012;46(6):500-04 21

10. Nakahira M, Saito N, Murata S, et al. Quantitative diffusion-weighted magnetic resonance imaging as a powerful adjunct 22

to fine needle aspiration cytology for assessment of thyroid nodules. American journal of otolaryngology 23

2012;33(4):408-16 doi: 10.1016/j.amjoto.2011.10.013[published Online First: Epub Date]|. 24

11. Mutlu H, Sivrioglu AK, Sonmez G, et al. Role of apparent diffusion coefficient values and diffusion-weighted magnetic 25

resonance imaging in differentiation between benign and malignant thyroid nodules. Clinical Imaging 2012;36(1):1-7 26

12. El-Hariri MA, Gouhar GK, Said NS, et al. Role of diffusion-weighted imaging with ADC mapping and in vivo 1H-MR 27

spectroscopy in thyroid nodules. Egyptian Journal of Radiology and Nuclear Medicine 2012;43(2):183-92 28

13. Aydin H, Kizilgoz V, Tatar I, et al. The role of proton MR spectroscopy and apparent diffusion coefficient values in the 29

diagnosis of malignant thyroid nodules: preliminary results. Clin Imaging 2012;36(4):323-33 doi: 30

10.1016/j.clinimag.2011.09.009[published Online First: Epub Date]|. 31

14. Yan B, Liu HJ, Wang CB, et al. ADC values in differentiation of benign and malignant thyroid nodules. Chinese Journal of 32

Medical Imaging Technology 2011;27(3):510-14 33

15. Schueller-Weidekamm C, Kaserer K, Schueller G, et al. Can quantitative diffusion-weighted MR imaging differentiate 34

benign and malignant cold thyroid nodules? Initial results in 25 patients. American Journal of Neuroradiology 35

2009;30(2):417-22 36

16. Ren S, Liu CH, Bai RJ. [Value of diffusion weighted imaging in diagnosis of nodular lesions of thyroid: a preliminary study]. 37

Zhonghua yi xue za zhi 2010;90(47):3351-54 38

17. Li RK QJ, Liu W, et al. Application of MR diffusion weighted imaging in the differentiation of malignant from benign thyroid 39

focal lesions. Radiol Practice (China) 2009; 24(7):719-22. 40

18. Razek AA, Sadek AG, Kombar OR, et al. Role of apparent diffusion coefficient values in differentiation between malignant 41

and benign solitary thyroid nodules. AJNR American journal of neuroradiology 2008;29(3):563-8 doi: 42

10.3174/ajnr.A0849[published Online First: Epub Date]|. 43

19. Whiting P, Rutjes AW, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of 44

diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003;3:25 doi: 10.1186/1471-2288-3-25 45

Page 15 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



16

[doi] 1

1471-2288-3-25 [pii][published Online First: Epub Date]|. 2

20. Whiting PF, Weswood ME, Rutjes AW, et al. Evaluation of QUADAS, a tool for the quality assessment of diagnostic accuracy 3

studies. BMC Med Res Methodol 2006;6:9 doi: 10.1186/1471-2288-6-9[published Online First: Epub Date]|. 4

21. Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy 5

studies. Ann Intern Med 2011;155(8):529-36 doi: 10.7326/0003-4819-155-8-201110180-00009[published Online 6

First: Epub Date]|. 7

22. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ 2003;327(7414):557-60 doi: 8

10.1136/bmj.327.7414.557[published Online First: Epub Date]|. 9

23. Leeflang MM, Deeks JJ, Gatsonis C, et al. Systematic reviews of diagnostic test accuracy. Annals of internal medicine 10

2008;149(12):889-97 11

24. Zamora J, Abraira V, Muriel A, et al. Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Med Res 12

Methodol 2006;6:31 doi: 10.1186/1471-2288-6-31[published Online First: Epub Date]|. 13

25. Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: 14

data-analytic approaches and some additional considerations. Stat Med 1993;12(14):1293-316 15

26. Reitsma JB, Glas AS, Rutjes AW, et al. Bivariate analysis of sensitivity and specificity produces informative summary 16

measures in diagnostic reviews. Journal of clinical epidemiology 2005;58(10):982-90 doi: 17

10.1016/j.jclinepi.2005.02.022[published Online First: Epub Date]|. 18

27. Menke J. Bivariate random-effects meta-analysis of sensitivity and specificity with SAS PROC GLIMMIX. Methods of 19

information in medicine 2010;49(1):54-62, 62-4 doi: 10.3414/me09-01-0001[published Online First: Epub Date]|. 20

28. StataCorp. Stata Statistical Software: Release 10. . Texas: College Station 2009 21

29. Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic 22

reviews of diagnostic test accuracy was assessed. Journal of Clinical Epidemiology 2005;58(9):882-93 23

30. Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA 24

statement. PLoS medicine 2009;6(7):e1000097 doi: 10.1371/journal.pmed.1000097[published Online First: Epub 25

Date]|. 26

31. Schueller-Weidekamm C, Schueller G, Kaserer K, et al. Diagnostic value of sonography, ultrasound-guided fine-needle 27

aspiration cytology, and diffusion-weighted MRI in the characterization of cold thyroid nodules. European Journal of 28

Radiology 2010;73(3):538-44 29

32. Jin Y, Qiang JW, Feng Q, et al. MR diffusion weighted imaging of thyroid nodules: Correlation with pathology. Chinese 30

Journal of Medical Imaging Technology 2012;28(2):256-60 31

33. Brown AM, Nagala S, McLean MA, et al. Multi-institutional validation of a novel textural analysis tool for preoperative 32

stratification of suspected thyroid tumors on diffusion-weighted MRI. Magnetic resonance in medicine : official 33

journal of the Society of Magnetic Resonance in Medicine / Society of Magnetic Resonance in Medicine 2015 doi: 34

10.1002/mrm.25743[published Online First: Epub Date]|. 35

34. Vermoolen MA, Kwee TC, Nievelstein RA. Apparent diffusion coefficient measurements in the differentiation between 36

benign and malignant lesions: a systematic review. Insights into imaging 2012;3(4):395-409 doi: 37

10.1007/s13244-012-0175-y[published Online First: Epub Date]|. 38

35. Wu LM, Chen XX, Li YL, et al. On the utility of quantitative diffusion-weighted MR imaging as a tool in differentiation 39

between malignant and benign thyroid nodules. Acad Radiol 2014;21(3):355-63 doi: 40

10.1016/j.acra.2013.10.008[published Online First: Epub Date]|. 41

36. Glas AS, Lijmer JG, Prins MH, et al. The diagnostic odds ratio: a single indicator of test performance. J Clin Epidemiol 42

2003;56(11):1129-35 43

37. Pan Z, Trikalinos TA, Kavvoura FK, et al. Local literature bias in genetic epidemiology: an empirical evaluation of the 44

Chinese literature. PLoS medicine 2005;2(12):e334 doi: 10.1371/journal.pmed.0020334[published Online First: Epub 45

Date]|. 46

Page 16 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



17

38. Andrew Vickers NG, Robert Harland, and Rebecca Rees. Do certain countries produce only positive results? A systematic 1

review of controlled trials. Controlled Clinical Trials 1998;19:159-66 2

39. Rutjes AW, Reitsma JB, Di Nisio M, et al. Evidence of bias and variation in diagnostic accuracy studies. CMAJ : Canadian 3

Medical Association journal = journal de l'Association medicale canadienne 2006;174(4):469-76 doi: 4

10.1503/cmaj.050090[published Online First: Epub Date]|. 5

40. Brazzelli M, Sandercock PA, Chappell FM, et al. Magnetic resonance imaging versus computed tomography for detection 6

of acute vascular lesions in patients presenting with stroke symptoms. Cochrane database of systematic reviews 7

(Online) 2009(4):CD007424 doi: 10.1002/14651858.CD007424.pub2[published Online First: Epub Date]|. 8

41. McAuley L, Ba'Pham, Tugwell P, et al. Does the inclusion of grey literature influence estimates of intervention 9

effectiveness reported in meta-analyses? Lancet 2000;356(9237):1228-31 10

11

12

13

Page 17 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



18

Figure legends 1

Figure 1. Flowchart illustrating the selection of studies. 2

Figure 2. Methodological quality of the 15 included studies. A. Methodological quality graph: each 3

methodological quality item is presented as the percentages across all included studies. B. Methodological 4

quality summary. 5

Figure 3. Forest plots of the SEN and the SPE with corresponding 95% CIs for DWI in the detection of 6

thyroid nodules. 7

Figure 4. Forest plots of the diagnostic odds ratio (DOR) with corresponding 95% CIs for DWI in the 8

detection of thyroid nodules. 9

Figure 5. Hierarchical summary receiver operating characteristic (HSROC) curves from the bivariate 10

model of DWI in the detection of thyroid nodules. 11

Figure 6. The funnel plot of publication bias. Linear regression of the inverse root of the effective sample 12

size (ESS) against the log DOR was performed to assess funnel plot asymmetry. 13

14

15

16

Page 18 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



19

Table 1. Characteristics of the Included Studies 1

Notes: ND, not mentioned; FNAB; fine needle aspiration biopsy; pro, prospective; retro, retrospective; Both, Histopathologic and FNAB. 2

3

Study Year Nation SEN SPE Field Enrollment Design Reference Blinding b value

Razek [18] 2008 Egypt 98% 92% 1.5T Consecutive pro Histopathologic ND 500

Li [17] 2009 China 79% 86% 1.5T ND retro Histopathologic ND 150



Schueller [15] 2009 Austria 85% 100% 1.0T ND pro Both Unblinded 800

Bozgeyik [1] 2009 Turkey 89% 100% 1.5T Consecutive pro FNAB Blinded 100

100% 80% 1.5T Consecutive pro FNAB Blinded 200

90% 100% 1.5T Consecutive pro FNAB Blinded 300

Ren [16] 2010 China 90% 90% 1.5T Consecutive retro Histopathologic ND 100

83% 90% 1.5T Consecutive retro Histopathologic ND 200



Yan [14] 2011 China 87% 100% 1.5T ND retro Histopathologic Blinded 500

Aydin [13] 2012 Turkey 92% 75% 1.5T Consecutive retro Histopathologic ND 400

El-Hariri [12] 2012 Egypt 94% 95% 1.5T ND pro Both ND 500

Mutlu [11] 2012 Turkey 80% 97% 1.5T ND pro Both ND 1000

Nakahira [10] 2012 Japan 94% 83% 1.5T ND retro Histopathologic Blinded 1000

Yue [9] 2012 China 86% 79% 1.5T ND retro Histopathologic Blinded 300

Ilica [8] 2013 Turkey 90% 100% 3T ND pro Both ND 1000

Shi [7] 2013 China 92% 88% 1.5T ND retro Histopathologic ND 500

Wu [6] 2013 China 77% 100% 1.5T ND retro Histopathologic Blinded 300

68% 64% 1.5T ND retro Histopathologic Blinded 500

54% 71% 1.5T ND retro Histopathologic Blinded 800

Elshafey [5] 2013 Egypt 96% 92% 1.5T ND pro Both Blinded 1000

Page 19 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



20

Table 2. Sensitivity and Specificity Estimates for Each Subgroup 1

Subgroup No. of studies Mean SEN (%) Mean SPE (%) DOR AUC (%)

b value (s/m2)

300 4 90 (80-96) 88 (83-93) 47 (11-190) 91 (88-95)

500 6 92 (86-96) 82 (77-86) 53 (12-239) 93 (90-96)

1000 4 88 (81-93) 93 (84-98) 115 (30-446) 96 (93-99)

Magnetic field

strength

1.0T 1 85 100 NA NA

1.5T 13 89 (85-93) 88 (85-91) 64 (26-156) 95 (91-99)

3.0T 1 90 100 NA NA

Study design

retrospective 8 91 (86-94) 83 (78-88) 33 (13-85) 93 (89-97)

prospective 7 87 (81-92) 98 (94-99) 153 (76-446) 97 (95-99)

Blinding

Yes 9 90 (85-93) 88 (83-92) 69 (32-147) 95 (93-97)

Unknown or no 6 88 (79-94) 90 (85-94) 66 (12-365) 94 (89-98)

Notes: The numbers in parentheses are the 95% CIs. NA = not applicable. 2

3

Page 20 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



Figure 1. Flowchart illustrating the selection of studies. 112x125mm (300 x 300 DPI)

Page 21 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



90x89mm (300 x 300 DPI)

Page 22 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



90x44mm (300 x 300 DPI)

Page 23 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



90x65mm (300 x 300 DPI)

Page 24 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



90x73mm (600 x 600 DPI)

Page 25 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from



90x68mm (300 x 300 DPI)

Page 26 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from




TITLE

Title 1 Identify the report as a systematic review, meta-analysis, or both. Title

ABSTRACT

Structured summary 2 Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, participants, and interventions; study appraisal and synthesis methods; results; limitations; conclusions and implications of key findings; systematic review registration number.

Abstract

INTRODUCTION

Rationale 3 Describe the rationale for the review in the context of what is already known. Introduction

Objectives 4 Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design (PICOS).

Introduction

METHODS

Protocol and registration 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.

No

Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered,

language, publication status) used as criteria for eligibility, giving rationale.


Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.

Literature search

Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

Literature search

Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).


Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

data extraction

Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.

data extraction

Risk of bias in individual studies

12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

Quality assessment

Summary measures 13 State the principal summary measures (e.g., risk ratio, difference in means). Meta-analysis

Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I

2) for each meta-analysis.

Meta-analysis

Page 1 of 2

Page 27 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from




Risk of bias across studies 15 Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting within studies).

Quality assessment

Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified.

Meta-analysis

RESULTS

Study selection 17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.

Results

Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations.

Results

Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12). Results

Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

Results

Synthesis of results 21 Present results of each meta-analysis done, including confidence intervals and measures of consistency. Results

Risk of bias across studies 22 Present results of any assessment of risk of bias across studies (see Item 15). Results

Additional analysis 23 Give results of additional analyses, if done (e.g., sensitivity or subgroup analyses, meta-regression [see Item 16]). Results

DISCUSSION

Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

Discussion

Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).

Discussion

Conclusions 26 Provide a general interpretation of the results in the context of other evidence, and implications for future research. Discussion

FUNDING

Funding 27 Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the systematic review.

Funding Statement

From: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097

For more information, visit: www.prisma-statement.org.

Page 2 of 2

Page 28 of 28


BMJ Open

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

on Decem


http://bmjopen.bm

j.com/

BM



nloaded from


bmj open€¦ · 22 related to dwi in differentiating benign from malignant thyroid nodules....

Documents