ORIGINAL ARTICLE
Differential Diagnosis of Benign and Malignant Breast Tumors Using Apparent Diffusion Coefficient Value Measured Through Diffusion-Weighted Magnetic Resonance Imaging Rong-Feng Qu, MD,* Dong-Rui Guo, MD,† Zhe-Xing Chang, MD,‡ Jie Meng, PhD,§ Yan Sun, PhD,† Shu-Hong Hao, MD,† Guang Shi, MD,† and Jian Sun, PhD*
Objective: Apparent diffusion coefficient (ADC) value measurement of nodes in diffusion-weighted imaging was widely used in differentiating different types of human tumors. The aim of this meta-analysis was to evaluate the clinical value of ADC measurement through diffusion-weighted magnetic resonance imaging (DW-MRI) in the differential diagnosis of benign and malignant breast tumors. Methods: Relevant studies were identified through computer-based search of databases, which were supplemented through manual search strategies. Case-control studies were selected in adherence with our strict inclusion and exclusion criteria. Statistical analysis was conducted using Stata 12.0 statistical software (StataCorp, College Station, Tex). Results: Our database searches initially retrieved 602 studies (320 studies in Chinese and 282 studies in English), and 31 studies (18 studies in English and 13 studies in Chinese) were eventually selected for metaanalysis. These 31 case-control studies included a total of 926 normal breast tissues and 2323 breast tumors (911 benign tumors and 1412 malignant tumors). Our meta-analysis showed that ADC values measured through DW-MRI were higher in benign breast tumors compared with malignant breast tumors, and this difference was statistically significant. In addition, the ADC values in the normal breast tissues were markedly higher than the benign breast tumors, which were also at a statistically significant level. Consistent with these observations, the ADC values in the normal breast tissues were significantly higher when compared with the values found in the malignant breast tumors. Conclusions: Our data strongly support the conclusion that the ADC value measured through DW-MRI is an important radiographic index for differential diagnosis of benign and malignant breast tumors and is critical to our assessment of the internal structure of tumors. Key Words: diffusion-weighted magnetic resonance imaging, apparent diffusion coefficient, breast cancer, benign tumor, malignant tumor, meta-analysis (J Comput Assist Tomogr 2015;39: 513–522)
B
reast cancer (BC) is one of the most frequently diagnosed malignancies affecting women all over the world.1 In United States, approximately 232,340 women were diagnosed with invasive BC in 2013 and 39,620 deaths were linked to BC in the same year.2 Breast cancer cases around the world have been on the rise, increasing by more than 1 million cases annually, with higher
From the *Department of Cardiology, First Bethune Hospital of Jilin University; †Department of Hematology and Oncology, the Second Hospital of Jilin University, Changchun; ‡Department of Oncology, Affiliated Hospital of Beihua University, Jilin; and §Department of Ultrasonic Diagnosis, the Second Hospital of Jilin University, Changchun, China. Received for publication July 24, 2014; accepted January 7, 2015. Reprints: Jian Sun, PhD, Department of Cardiology, First Bethune Hospital of Jilin University, Xinmin St, No. 71, Changchun 130021, People's Republic of China (e‐mail: [email protected]). The authors declare no conflict of interest. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
incidence rates found in developed countries. However, the incidence rate in China, particularly in Beijing, has dramatically increased by 23% to 31% within the last 10 years.3 Depending on the context, BC is variously classified according to grade, stage, histopathology, receptor status, genetic or epigenetic profiles, and other more recent DNA/RNA/protein-based approaches, underscoring the significant difficulties in predicting BC disease progression within the current classification systems.4 Various risk factors, such as diet, hormonal levels, family history, lack of physical exercise, alcohol consumption, menophania, late menopause, menopausal period, and genetic predisposition, are linked to the pathogenesis of BC.5,6 At present, the all-in approach of treatment of patients with BC includes a combination of approaches such as endocrine, chemotherapeutic, or biological therapies to produce a series of remissions and a more drastic approach of single/double mastectomy along with other treatment modalities for more aggressive BC.7 Recently, studies have shown that the apparent diffusion coefficient (ADC), which was derived from the diffusion-weighted magnetic resonance imaging (DWMRI), has the potential to accurately diagnose BC and that DWMRI–based measurements can provide information on the tumor internal structure and predict BC behavior.8,9 Diffusion-weighted magnetic resonance imaging, a recent advancement of the magnetic resonance imaging (MRI) technique, measures the mobility of water (water diffusion) in live tissues within the body as a diagnostic tool.8 The principle of DW-MRI used in clinical applications is that live tissue structures can be probed at levels beyond the usual image resolution by measuring the behavior of water molecules when they are diffusing randomly in the tissue by displacement, thus providing a realistic picture of the tissue's internal structure.10 In DW-MRI, a diffusionsensitizing gradient generates images, and by changing the “b value” on the scanner, the gradient strength can be altered.11 Performing DW-MRI using 2 or more b values allows for the calculation of a quantitative parameter known as the ADC.12 The ADC, which is expressed in square millimeter per second in unit measurement, combines the effects of capillary perfusion and water diffusion in the extracellular extravascular space. Apparent diffusion coefficient can be used to compare diffusion in lesions; thus, it is important in evaluating patients with cancer including BC, liver cancer, and prostate cancer.13 Therefore, DW-MRI has the potential to differentiate malignant breast tumors from benign breast tumors by measuring ADC values, which is an overall reflection of the tumor parameters that includes blood flow, tissue cellularity, and membrane permeability.14 In general, malignant breast tumors demonstrate decreased ADC with respect to normal breast tissue and benign tumors because of the higher cell density.15,16 A few previous studies have shown that ADC value measured through DW-MRI is indeed an important radiographic index in differential diagnosis of benign and malignant breast tumors.17,18 However, other studies differed in their opinion of DW-MRI as an accurate diagnostic tool for tumors.19,20 We, therefore, performed
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jcat.org
513
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
Qu et al
a meta-analysis to investigate the clinical value of ADC measurement through DW-MRI in the differential diagnosis of benign and malignant breast tumors.
(11) baseline equivalence of groups (MINORS11); and (12) adequate statistical analyses (MINORS12).
Statistical Analysis MATERIALS AND METHODS Data Sources and Key Words A comprehensive list of computer-based bibliographic databases (PubMed, EMBASE, EBSCO, Ovid, SpringerLink, Wiley, Web of Science, Cochrane Library, Cumulative Index to Nursing and Allied Health, Wanfang database, VIP database, China Biomedicine, and China National Knowledge Infrastructure; last update on October 1, 2014) was searched using combinations of free words and key words. The search terms were as follows: (“diffusion magnetic resonance imaging” or “diffusion MRI” or “diffusion weighted MRI” or “diffusion weighted imaging” or “DWI” or “WB-DWI” or “DMRI” or “diffusion”) and (“breast neoplasms” or “breast cancer” or “breast carcinoma” or “tumors, breast” or “mammary neoplasms, human” or “carcinoma, human mammary” or “mammary cancer” or “malignant neoplasm of breast” or “malignant tumor of breast” or “cancer of the breast”). Restrictions such as language, time, and data set were not used in this meta-analysis. Manual searches were also used for the screening and selection of other eligible studies.
Statistical analysis was performed using the Stata statistical software, version 12.0 (StataCorp, College Station, Tex). To calculate the effect size for each study, the summary standard mean difference (SMD) and 95% CI were applied to evaluate the correlations between ADC value measured through DW-MRI as well as benign and malignant breast tumors with the utilization of Z test.22 Cochran Q statistic and I2 tests were also adopted to quantify heterogeneity among studies.23,24 Random-effects model was applied for the existence of significant heterogeneity (P < 0.05 or I2 > 50%), whereas fixed-effects model was applied when no statistical heterogeneity was observed (P > 0.05 or I2 < 50%).25 When there was significant heterogeneity, subgroup analyses were performed to find potential explanatory variables. Forest plots were drawn to reflect the comparisons of odds risk and 95% CI among the study groups. Furthermore, 1-way sensitivity analysis was performed to evaluate the influence of individual studies on the overall effect estimates. Contour-enhanced funnel plots and Egger linear regression test were inspected for indication of publication bias, so as to clear out the reliability of original analysis results.26 All tests were 2 sided, and P < 0.05 indicated a statistical significance.
RESULTS
Study Selection After reading the Abstract, full articles were retrieved and assessed for their suitability using the following inclusion criteria: (1) clinical case-control studies concerning the ADC measurement through DW-MRI in the differential diagnosis of benign and malignant breast tumors; (2) study subjects should be patients with BC and healthy controls; (3) included studies provided complete data of age, country, language, ethnicity, sex, pathological type, MRI machine type, b value, case number, and lesions; and (4) only the most recent and complete study was extracted if studies were published by the same author after a careful reexamination. Exclusion criteria were as follows: (1) studies unrelated to the research subjects; (2) studies with incomplete data; (3) nonChinese or non-English literature; (4) duplicate publications; and (5) journals of non–science and technology core or non–Peking University core.
Data Extraction and Quality Assessment We used a standard reporting form to extract data from each study, and the following descriptive information were collected: the first author, publication time, country, ethnicity, language, age, sex, study design, pathological type, MRI machine type, b value, case number, and lesions. Discrepancies on the inclusion of studies were resolved by discussion or by consulting a third investigator. Quality assessment of the included studies was carried out by 2 reviewers independently based on methodological index for nonrandomized studies (MINORS).21 The MINORS were standardized as follows: (1) a clearly stated aim (MINORS01); (2) inclusion of consecutive patients (MINORS02); (3) prospective collection of data (MINORS03); (4) end points appropriate to the aim of the study (MINORS04); (5) unbiased assessment of the study end point (MINORS05); (6) appropriate follow-up period (MINORS06); (7) loss to follow up of less than 5% (MINORS07); (8) prospective calculation for the study size of 95% confidence interval (95% CI; MINORS08); (9) an adequate control group or a criterion standard diagnostic test and therapeutic intervention (MINORS09); (10) contemporary groups (MINORS10);
514
www.jcat.org
Baseline Characteristics of Included Study Figure 1 shows the flowchart of published studies and the main exclusion criteria. Our present meta-analysis initially retrieved 602 studies through electronic database searching accompanied by manual search. From these 602 studies, 150 studies were eliminated for being duplicates; 12, for being letters, reviews, or meta-analysis; 15, for being nonhuman studies; and 336, for being unrelated to the research topic. After reviewing the remaining 88 studies, additional 54 studies were excluded for not being a case-control study (n = 21), irrelevant to DWMRI (n = 17), or irrelevant to BC (n = 16). Finally, 31 studies were enrolled in the meta-analysis, following an exclusion of 3 studies for having insufficient information (n = 3). The finally selected 31 case-control studies,27–57 which were published between 2002 and 2014, evaluated the ADC values of DW-MRI in the differential diagnosis of benign and malignant breast tumors in Asian populations (23 studies) as well as in white populations (8 studies) and included a total of 926 normal breast tissues and 2323 breast tumors (911 benign tumors and 1412 malignant tumors). The countries where the studies were performed are China (n = 19), America (n = 3), Japan (n = 3), Korea (n = 1), Venezuela (n = 1), Turkey (n = 1), Brazil (n = 1), Austria (n = 1), and Belgium (n = 1). The MRI machine types were Siemens, General Electric (GE), and Philips. The b values were 400 s/mm2, 500 s/ mm2, 600 s/mm2, 800 s/mm2, 1000 s/mm2, and others, respectively. The baseline characteristics and the MINORS quality evaluation of the extracted studies are presented in Table 1 and Figure 2.
Pooled Outcomes of Meta-Analysis The results of heterogeneity test suggested that there was evidence of heterogeneity in the studies of ADC values in different tissues (benign versus malignant: P < 0.001, I2 = 95.3%; normal versus benign: P < 0.001, I2 = 84.5%; normal versus malignant: P < 0.001, I2 = 97.2%); thus, random-effects model was applied in this meta-analysis. Pooled data from studies demonstrated that the ADC values measured through DW-MRI in the benign breast © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
ADC Values of DW-MRI in BC
FIGURE 1. Flowchart shows the study selection procedure. Thirty-one studies were included in this meta-analysis. Figure 1 can be viewed online in color at www.jcat.org.
tumors were significantly higher than those of the malignant breast tumors, showing a statistically significant difference (SMD, 2.10; 95% CI, 1.71–2.49; P < 0.001; Fig. 3). The ADC values measured through DW-MRI in the normal breast tissues were markedly higher than those of the benign breast tumors, indicating significant differences between the 2 groups (SMD, 1.09; 95% CI, 0.82–1.36; P < 0.001; Fig. 4). The ADC values measured through DW-MRI in the normal breast tissues were significantly higher than those of the malignant breast tumors, and there was a statistically significant difference in the ADC values between these 2 groups as well (SMD, 3.51; 95% CI, 2.81–4.22; P < 0.001; Fig. 5).
Pooled Outcomes of Subgroup Analyses Subgroup analysis based on ethnicity found that, in Asians, the ADC values in the benign breast tumors were remarkably higher than those of the malignant breast tumors, the ADC values in the normal breast tissues were significantly higher than those of the benign breast tumors, and the ADC values in the normal breast tissues were higher than those of the malignant breast tumors, which were all at statistically significant levels (all P < 0.05). On the other hand, in whites, the ADC values in the benign breast tumors were markedly higher than that of the malignant breast tumors and the ADC values the in normal breast tissues were significantly higher than that of the benign breast tumors (both P < 0.05), but we could not detect statistically significant differences between the ADC values in the normal breast tissues and the malignant breast tumors (P = 0.064). Subgroup analysis based on MRI machine type showed that, in GE and Siemens, the ADC values in the benign breast tumors
were remarkably higher than those of the malignant breast tumors, the ADC values in the normal breast tissues were markedly higher than those of the benign breast tumors, and the ADC values in the normal breast tissues were significantly higher than those of the malignant breast tumors (all P < 0.05). The MRI from Philips showed that the ADC values in the benign breast tumors were significantly higher than those of the malignant breast tumors and that the ADC values in the normal breast tissues were markedly higher than those of the malignant breast tumors (both P < 0.05), but no significant differences were observed in the ADC values in the normal breast tissues and in the benign breast tumors using Philips MRI (P = 0.058). Subgroup analysis based on b value revealed that, when the b value was 400 s/mm2, 500 s/mm2, 600 s/mm2, 800 s/mm2, 1000 s/ mm2, and others, respectively, the ADC values in the benign breast tumors were significantly higher than those of the malignant breast tumors and the ADC values in the normal breast tissues were remarkably higher than those of the benign breast tumors (all P < 0.05). When the b value was 400 s/mm2, 600 s/mm2, 800 s/mm2, 1000 s/mm2, and others, respectively, the ADC values in the normal breast tissues were higher than those of the malignant breast tumors with statistical significance (all P < 0.05). However, when the b value was 500 s/mm2, there was no statistical significance of the ADC values in the normal breast tissues and in the malignant breast tumors (P = 0.083; Table 2).
Sensitivity Analysis and Publication Bias The sensitivity analysis results revealed that the included studies had no significant effect on the pooled effect value SMD. Contour-enhanced funnel plot for the comparisons between
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jcat.org
515
516
30
He et al Yoo et al56 Nogueira et al49 Cai et al43 Baba et al40 Bokacheva et al42 Wang et al37 Zhang et al57 Orguc et al50 Chen et al44* Chen et al44† Chen et al44‡ Bai et al27§ Bai et al27* Bai et al27† Bai et al27‡ Partridge et al51 Inoue et al47 Xie et al38|| Xie et al38† Wang et al36 Han et al29|| Han et al29‡ Jin et al48* Jin et al48‡ Ei Khouli et al45 Yuan et al39 Wang and Duan35|| Wang and Duan35† Wang and Duan35¶ Luo et al34* Luo et al34‡ Zhang et al55|| Zhang et al55‡ Pereira et al52|| Pereira et al52‡ Bogner et al41§
Author
2014 2014 2014 2014 2014 2013 2012 2012 2012 2012 2012 2012 2011 2011 2011 2011 2011 2011 2010 2010 2010 2010 2010 2010 2010 2010 2009 2009 2009 2009 2009 2009 2009 2009 2009 2009 2009
Year
China Korea Portugal China Japan United States China China Turkey China China China China China China China United States Japan China China China China China China China United States China China China China China China China China Brazil Brazil Austria
Country Asians Asians Whites Asians Asians Whites Asians Asians Whites Asians Asians Asians Asians Asians Asians Asians Whites Asians Asians Asians Asians Asians Asians Asians Asians Whites Asians Asians Asians Asians Asians Asians Asians Asians Whites Whites Whites
Ethnicity
TABLE 1. Baseline Characteristics of the Included Studies
Malignant
44 (21–72) 44 (18–69) 52.1 ± 12.4 48.4 ± 11.1 39.1 ± 8.1 49 ± 18 49 (28–70) 44.1 (26–81) 42 ± 13.5 43 ± 16.5 44.9 (19–77) 49.6 (23–75) 49.6 (23–75) 49.6 (23–75) 40.6 (19–68) 49.7 (33–77) 40.6 (19–68) 49.7 (33–77) 40.6 (19–68) 49.7 (33–77) 40.6 (19–68) 49.7 (33–77) 49 (22–85) 53.8 (26–85) 44 (30–62) 44 (30–62) 50 (30–86) 18–75 18–75 18–80 18–80 50 ± 1.7 53.5 ± 11 45 (18-69) 39 (21–67) 43 (32–67) 39 (21–67) 43 (32–67) 39 (21–67) 43 (32–67) 42 (12–72) 42 (12–72) 46.3 (31–77) 46.3 (31–77) 46.1 (22–80) 46.1 (22–80) 52 ± 13
Benign
Age, y
GE Siemens Siemens GE Philips GE GE Philips GE GE GE GE GE GE GE GE GE Siemens GE GE GE Philips Philips GE GE Philips Philips Siemens Siemens Siemens GE GE Philips Philips GE GE Siemens
MRI Machine Type 800 Others Others 800 1000 600 1000 800 600 600 800 1000 400 600 800 1000 600 1000 500 800 800 500 1000 600 1000 600 800 500 800 Others 600 1000 500 1000 500 1000 400
b Value, s/mm2 48 89 53 234 79 35 40 60 108 55 55 55 100 100 100 100 77 105 53 53 50 53 53 56 56 93 36 52 52 52 43 43 54 54 45 45 51
Case Number
57 169 59 234 83 40 40 60 124 57 57 57 100 100 100 100 100 106 66 66 50 57 57 60 60 101 36 60 60 60 47 47 57 57 5 52 41
Lesions
Qu et al J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
www.jcat.org
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
ADC Values of DW-MRI in BC
Data under column “Age” are expressed as the mean ± standard deviation or as the median (interquartile range). *600 s/mm2. †800 s/mm2. ‡1000 s/mm2. §400 s/mm2. ||500 s/mm2. ¶Others.
40 (24–65) 47 42 (18–71) 42 (18–71) 42 (18–71) 42 (18–71) 46.9 (30–72) 46.9 (30–72) 50 (30–70) 53.1 (25–74) 53 (14–88) 38.3 ± 7.1 50.3 ± 11.6
Bogner et al41‡ Jin et al31* Jin et al31‡ Luo et al33 Lou et al32§ Lou et al32* Lou et al32† Lou et al32‡ Feng et al28† Feng et al28‡ Rubesova et al53 Woodhams et al54 Guo et al46
2009 2008 2008 2007 2007 2007 2007 2007 2007 2007 2006 2005 2002
Austria China China China China China China China China China Belgium Japan China
Whites Asians Asians Asians Asians Asians Asians Asians Asians Asians Whites Asians Asians
52 ± 13 — —
Siemens GE GE Siemens Siemens Siemens Siemens Siemens GE GE Siemens GE GE
1000 600 1000 800 400 600 800 1000 800 1000 Others Others Others
51 56 56 52 50 50 50 50 65 65 78 190 52
41 60 60 63 58 58 58 58 68 68 87 191 55
the normal breast tissues and the benign breast tumors suggested that most of the studies were in the range of P > 0.05, and Egger linear regression analysis further confirmed that there was no existence of publication bias (P > 0.05). However, contourenhanced funnel plot for the comparisons between the benign breast tumors and the malignant breast tumors as well as the normal breast tissues and the malignant breast tumors showed that most of the included studies were interspersed in the range of P < 0.01, revealing that there was publication bias, and the Egger linear regression analysis further confirmed publication bias (P > 0.05; Fig. 6).
FIGURE 2. Quality assessment of the included studies by MINORS score. Figure 2 can be viewed online in color at www.jcat.org.
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jcat.org
517
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
Qu et al
DISCUSSION To investigate the clinical value of ADC measurements through DW-MRI as a diagnostic tool to differentiate between benign and malignant breast tumors, a systematic meta-analysis was performed. Our meta-analysis results suggest that the ADC values measured through DW-MRI provide an accurate picture of the internal structure of the tumors and that it is an important radiographic index in the differential diagnosis of benign and malignant breast tumors. Diffusion-weighted MRI, which is a noncontrast MRI technique, provides critical information on cell density, tissue architecture, and integrity of membrane by measuring the diffusion of water molecule, with lower loss of signal representing the low water diffusion and higher loss of signal representing high water diffusion.58,59 Apparent diffusion coefficient is a frequently used parameter to quantify diffusibility of tissues by correlating the active
movement or attenuation of the motion of water molecules to tissue complexity; thus, ADC values are high in tissues without obstacles for water motion and are low in tissues with obstacles.60,61 In tissues, apparent diffusion of water molecules is altered by their interactions with cell membranes and macromolecules, and in tumor tissues with high cell density, the disorganized extracellular space and the higher density of hydrophobic cellular membranes restrict the apparent diffusion of water. Apparent diffusion coefficient is quantified by mean diffusivity measurement in 3 orthogonal directions, which is influenced by tissue cellularity, fluid viscosity, membrane permeability, and blood flow.62 A few studies have reported that the varied treatment responses reflecting diverse clinical outcomes are directly linked to the tumor morphology and could be predicted from ADC values, that is, high pretreatment ADC indicated high viability and shows a favorable clinical response and low ADC values indicated necrosis.12 With specific relevance to
FIGURE 3. Forest plots for the comparisons of ADC values between the benign breast tumors and the malignant breast tumors. Figure 3 can be viewed online in color at www.jcat.org.
518
www.jcat.org
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
ADC Values of DW-MRI in BC
FIGURE 4. Forest plots for the comparisons of ADC values between the normal breast tissues and the benign breast tumors. Figure 4 can be viewed online in color at www.jcat.org.
FIGURE 5. Forest plots for the comparisons of ADC values between the normal breast tissues and the malignant breast tumors. Figure 5 can be viewed online in color at www.jcat.org. © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jcat.org
519
520
Ethnicity Asians Whites Machine type GE Siemens Philips b Value 400 500 600 800 1000 Others
1.68–2.43 0.94–3.58
1.03–2.04 2.27–4.00 1.79–2.80
0.61–1.30 1.28–3.68 0.06–1.66 1.52–3.30 1.47–2.80 2.40–5.61
1.53 3.13 2.29
0.96 2.48 0.86 2.41 2.14 4.01
95% CI
2.05 2.26
SMD
Benign vs Malignant
www.jcat.org
Differential Diagnosis of Benign and Malignant Breast Tumors Using Apparent Diffusion Coefficient Value Measured Through Diffusion-Weighted Magnetic Resonance Imaging Rong-Feng Qu, MD,* Dong-Rui Guo, MD,† Zhe-Xing Chang, MD,‡ Jie Meng, PhD,§ Yan Sun, PhD,† Shu-Hong Hao, MD,† Guang Shi, MD,† and Jian Sun, PhD*
Objective: Apparent diffusion coefficient (ADC) value measurement of nodes in diffusion-weighted imaging was widely used in differentiating different types of human tumors. The aim of this meta-analysis was to evaluate the clinical value of ADC measurement through diffusion-weighted magnetic resonance imaging (DW-MRI) in the differential diagnosis of benign and malignant breast tumors. Methods: Relevant studies were identified through computer-based search of databases, which were supplemented through manual search strategies. Case-control studies were selected in adherence with our strict inclusion and exclusion criteria. Statistical analysis was conducted using Stata 12.0 statistical software (StataCorp, College Station, Tex). Results: Our database searches initially retrieved 602 studies (320 studies in Chinese and 282 studies in English), and 31 studies (18 studies in English and 13 studies in Chinese) were eventually selected for metaanalysis. These 31 case-control studies included a total of 926 normal breast tissues and 2323 breast tumors (911 benign tumors and 1412 malignant tumors). Our meta-analysis showed that ADC values measured through DW-MRI were higher in benign breast tumors compared with malignant breast tumors, and this difference was statistically significant. In addition, the ADC values in the normal breast tissues were markedly higher than the benign breast tumors, which were also at a statistically significant level. Consistent with these observations, the ADC values in the normal breast tissues were significantly higher when compared with the values found in the malignant breast tumors. Conclusions: Our data strongly support the conclusion that the ADC value measured through DW-MRI is an important radiographic index for differential diagnosis of benign and malignant breast tumors and is critical to our assessment of the internal structure of tumors. Key Words: diffusion-weighted magnetic resonance imaging, apparent diffusion coefficient, breast cancer, benign tumor, malignant tumor, meta-analysis (J Comput Assist Tomogr 2015;39: 513–522)
B
reast cancer (BC) is one of the most frequently diagnosed malignancies affecting women all over the world.1 In United States, approximately 232,340 women were diagnosed with invasive BC in 2013 and 39,620 deaths were linked to BC in the same year.2 Breast cancer cases around the world have been on the rise, increasing by more than 1 million cases annually, with higher
From the *Department of Cardiology, First Bethune Hospital of Jilin University; †Department of Hematology and Oncology, the Second Hospital of Jilin University, Changchun; ‡Department of Oncology, Affiliated Hospital of Beihua University, Jilin; and §Department of Ultrasonic Diagnosis, the Second Hospital of Jilin University, Changchun, China. Received for publication July 24, 2014; accepted January 7, 2015. Reprints: Jian Sun, PhD, Department of Cardiology, First Bethune Hospital of Jilin University, Xinmin St, No. 71, Changchun 130021, People's Republic of China (e‐mail: [email protected]). The authors declare no conflict of interest. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
incidence rates found in developed countries. However, the incidence rate in China, particularly in Beijing, has dramatically increased by 23% to 31% within the last 10 years.3 Depending on the context, BC is variously classified according to grade, stage, histopathology, receptor status, genetic or epigenetic profiles, and other more recent DNA/RNA/protein-based approaches, underscoring the significant difficulties in predicting BC disease progression within the current classification systems.4 Various risk factors, such as diet, hormonal levels, family history, lack of physical exercise, alcohol consumption, menophania, late menopause, menopausal period, and genetic predisposition, are linked to the pathogenesis of BC.5,6 At present, the all-in approach of treatment of patients with BC includes a combination of approaches such as endocrine, chemotherapeutic, or biological therapies to produce a series of remissions and a more drastic approach of single/double mastectomy along with other treatment modalities for more aggressive BC.7 Recently, studies have shown that the apparent diffusion coefficient (ADC), which was derived from the diffusion-weighted magnetic resonance imaging (DWMRI), has the potential to accurately diagnose BC and that DWMRI–based measurements can provide information on the tumor internal structure and predict BC behavior.8,9 Diffusion-weighted magnetic resonance imaging, a recent advancement of the magnetic resonance imaging (MRI) technique, measures the mobility of water (water diffusion) in live tissues within the body as a diagnostic tool.8 The principle of DW-MRI used in clinical applications is that live tissue structures can be probed at levels beyond the usual image resolution by measuring the behavior of water molecules when they are diffusing randomly in the tissue by displacement, thus providing a realistic picture of the tissue's internal structure.10 In DW-MRI, a diffusionsensitizing gradient generates images, and by changing the “b value” on the scanner, the gradient strength can be altered.11 Performing DW-MRI using 2 or more b values allows for the calculation of a quantitative parameter known as the ADC.12 The ADC, which is expressed in square millimeter per second in unit measurement, combines the effects of capillary perfusion and water diffusion in the extracellular extravascular space. Apparent diffusion coefficient can be used to compare diffusion in lesions; thus, it is important in evaluating patients with cancer including BC, liver cancer, and prostate cancer.13 Therefore, DW-MRI has the potential to differentiate malignant breast tumors from benign breast tumors by measuring ADC values, which is an overall reflection of the tumor parameters that includes blood flow, tissue cellularity, and membrane permeability.14 In general, malignant breast tumors demonstrate decreased ADC with respect to normal breast tissue and benign tumors because of the higher cell density.15,16 A few previous studies have shown that ADC value measured through DW-MRI is indeed an important radiographic index in differential diagnosis of benign and malignant breast tumors.17,18 However, other studies differed in their opinion of DW-MRI as an accurate diagnostic tool for tumors.19,20 We, therefore, performed
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jcat.org
513
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
Qu et al
a meta-analysis to investigate the clinical value of ADC measurement through DW-MRI in the differential diagnosis of benign and malignant breast tumors.
(11) baseline equivalence of groups (MINORS11); and (12) adequate statistical analyses (MINORS12).
Statistical Analysis MATERIALS AND METHODS Data Sources and Key Words A comprehensive list of computer-based bibliographic databases (PubMed, EMBASE, EBSCO, Ovid, SpringerLink, Wiley, Web of Science, Cochrane Library, Cumulative Index to Nursing and Allied Health, Wanfang database, VIP database, China Biomedicine, and China National Knowledge Infrastructure; last update on October 1, 2014) was searched using combinations of free words and key words. The search terms were as follows: (“diffusion magnetic resonance imaging” or “diffusion MRI” or “diffusion weighted MRI” or “diffusion weighted imaging” or “DWI” or “WB-DWI” or “DMRI” or “diffusion”) and (“breast neoplasms” or “breast cancer” or “breast carcinoma” or “tumors, breast” or “mammary neoplasms, human” or “carcinoma, human mammary” or “mammary cancer” or “malignant neoplasm of breast” or “malignant tumor of breast” or “cancer of the breast”). Restrictions such as language, time, and data set were not used in this meta-analysis. Manual searches were also used for the screening and selection of other eligible studies.
Statistical analysis was performed using the Stata statistical software, version 12.0 (StataCorp, College Station, Tex). To calculate the effect size for each study, the summary standard mean difference (SMD) and 95% CI were applied to evaluate the correlations between ADC value measured through DW-MRI as well as benign and malignant breast tumors with the utilization of Z test.22 Cochran Q statistic and I2 tests were also adopted to quantify heterogeneity among studies.23,24 Random-effects model was applied for the existence of significant heterogeneity (P < 0.05 or I2 > 50%), whereas fixed-effects model was applied when no statistical heterogeneity was observed (P > 0.05 or I2 < 50%).25 When there was significant heterogeneity, subgroup analyses were performed to find potential explanatory variables. Forest plots were drawn to reflect the comparisons of odds risk and 95% CI among the study groups. Furthermore, 1-way sensitivity analysis was performed to evaluate the influence of individual studies on the overall effect estimates. Contour-enhanced funnel plots and Egger linear regression test were inspected for indication of publication bias, so as to clear out the reliability of original analysis results.26 All tests were 2 sided, and P < 0.05 indicated a statistical significance.
RESULTS
Study Selection After reading the Abstract, full articles were retrieved and assessed for their suitability using the following inclusion criteria: (1) clinical case-control studies concerning the ADC measurement through DW-MRI in the differential diagnosis of benign and malignant breast tumors; (2) study subjects should be patients with BC and healthy controls; (3) included studies provided complete data of age, country, language, ethnicity, sex, pathological type, MRI machine type, b value, case number, and lesions; and (4) only the most recent and complete study was extracted if studies were published by the same author after a careful reexamination. Exclusion criteria were as follows: (1) studies unrelated to the research subjects; (2) studies with incomplete data; (3) nonChinese or non-English literature; (4) duplicate publications; and (5) journals of non–science and technology core or non–Peking University core.
Data Extraction and Quality Assessment We used a standard reporting form to extract data from each study, and the following descriptive information were collected: the first author, publication time, country, ethnicity, language, age, sex, study design, pathological type, MRI machine type, b value, case number, and lesions. Discrepancies on the inclusion of studies were resolved by discussion or by consulting a third investigator. Quality assessment of the included studies was carried out by 2 reviewers independently based on methodological index for nonrandomized studies (MINORS).21 The MINORS were standardized as follows: (1) a clearly stated aim (MINORS01); (2) inclusion of consecutive patients (MINORS02); (3) prospective collection of data (MINORS03); (4) end points appropriate to the aim of the study (MINORS04); (5) unbiased assessment of the study end point (MINORS05); (6) appropriate follow-up period (MINORS06); (7) loss to follow up of less than 5% (MINORS07); (8) prospective calculation for the study size of 95% confidence interval (95% CI; MINORS08); (9) an adequate control group or a criterion standard diagnostic test and therapeutic intervention (MINORS09); (10) contemporary groups (MINORS10);
514
www.jcat.org
Baseline Characteristics of Included Study Figure 1 shows the flowchart of published studies and the main exclusion criteria. Our present meta-analysis initially retrieved 602 studies through electronic database searching accompanied by manual search. From these 602 studies, 150 studies were eliminated for being duplicates; 12, for being letters, reviews, or meta-analysis; 15, for being nonhuman studies; and 336, for being unrelated to the research topic. After reviewing the remaining 88 studies, additional 54 studies were excluded for not being a case-control study (n = 21), irrelevant to DWMRI (n = 17), or irrelevant to BC (n = 16). Finally, 31 studies were enrolled in the meta-analysis, following an exclusion of 3 studies for having insufficient information (n = 3). The finally selected 31 case-control studies,27–57 which were published between 2002 and 2014, evaluated the ADC values of DW-MRI in the differential diagnosis of benign and malignant breast tumors in Asian populations (23 studies) as well as in white populations (8 studies) and included a total of 926 normal breast tissues and 2323 breast tumors (911 benign tumors and 1412 malignant tumors). The countries where the studies were performed are China (n = 19), America (n = 3), Japan (n = 3), Korea (n = 1), Venezuela (n = 1), Turkey (n = 1), Brazil (n = 1), Austria (n = 1), and Belgium (n = 1). The MRI machine types were Siemens, General Electric (GE), and Philips. The b values were 400 s/mm2, 500 s/ mm2, 600 s/mm2, 800 s/mm2, 1000 s/mm2, and others, respectively. The baseline characteristics and the MINORS quality evaluation of the extracted studies are presented in Table 1 and Figure 2.
Pooled Outcomes of Meta-Analysis The results of heterogeneity test suggested that there was evidence of heterogeneity in the studies of ADC values in different tissues (benign versus malignant: P < 0.001, I2 = 95.3%; normal versus benign: P < 0.001, I2 = 84.5%; normal versus malignant: P < 0.001, I2 = 97.2%); thus, random-effects model was applied in this meta-analysis. Pooled data from studies demonstrated that the ADC values measured through DW-MRI in the benign breast © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
ADC Values of DW-MRI in BC
FIGURE 1. Flowchart shows the study selection procedure. Thirty-one studies were included in this meta-analysis. Figure 1 can be viewed online in color at www.jcat.org.
tumors were significantly higher than those of the malignant breast tumors, showing a statistically significant difference (SMD, 2.10; 95% CI, 1.71–2.49; P < 0.001; Fig. 3). The ADC values measured through DW-MRI in the normal breast tissues were markedly higher than those of the benign breast tumors, indicating significant differences between the 2 groups (SMD, 1.09; 95% CI, 0.82–1.36; P < 0.001; Fig. 4). The ADC values measured through DW-MRI in the normal breast tissues were significantly higher than those of the malignant breast tumors, and there was a statistically significant difference in the ADC values between these 2 groups as well (SMD, 3.51; 95% CI, 2.81–4.22; P < 0.001; Fig. 5).
Pooled Outcomes of Subgroup Analyses Subgroup analysis based on ethnicity found that, in Asians, the ADC values in the benign breast tumors were remarkably higher than those of the malignant breast tumors, the ADC values in the normal breast tissues were significantly higher than those of the benign breast tumors, and the ADC values in the normal breast tissues were higher than those of the malignant breast tumors, which were all at statistically significant levels (all P < 0.05). On the other hand, in whites, the ADC values in the benign breast tumors were markedly higher than that of the malignant breast tumors and the ADC values the in normal breast tissues were significantly higher than that of the benign breast tumors (both P < 0.05), but we could not detect statistically significant differences between the ADC values in the normal breast tissues and the malignant breast tumors (P = 0.064). Subgroup analysis based on MRI machine type showed that, in GE and Siemens, the ADC values in the benign breast tumors
were remarkably higher than those of the malignant breast tumors, the ADC values in the normal breast tissues were markedly higher than those of the benign breast tumors, and the ADC values in the normal breast tissues were significantly higher than those of the malignant breast tumors (all P < 0.05). The MRI from Philips showed that the ADC values in the benign breast tumors were significantly higher than those of the malignant breast tumors and that the ADC values in the normal breast tissues were markedly higher than those of the malignant breast tumors (both P < 0.05), but no significant differences were observed in the ADC values in the normal breast tissues and in the benign breast tumors using Philips MRI (P = 0.058). Subgroup analysis based on b value revealed that, when the b value was 400 s/mm2, 500 s/mm2, 600 s/mm2, 800 s/mm2, 1000 s/ mm2, and others, respectively, the ADC values in the benign breast tumors were significantly higher than those of the malignant breast tumors and the ADC values in the normal breast tissues were remarkably higher than those of the benign breast tumors (all P < 0.05). When the b value was 400 s/mm2, 600 s/mm2, 800 s/mm2, 1000 s/mm2, and others, respectively, the ADC values in the normal breast tissues were higher than those of the malignant breast tumors with statistical significance (all P < 0.05). However, when the b value was 500 s/mm2, there was no statistical significance of the ADC values in the normal breast tissues and in the malignant breast tumors (P = 0.083; Table 2).
Sensitivity Analysis and Publication Bias The sensitivity analysis results revealed that the included studies had no significant effect on the pooled effect value SMD. Contour-enhanced funnel plot for the comparisons between
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jcat.org
515
516
30
He et al Yoo et al56 Nogueira et al49 Cai et al43 Baba et al40 Bokacheva et al42 Wang et al37 Zhang et al57 Orguc et al50 Chen et al44* Chen et al44† Chen et al44‡ Bai et al27§ Bai et al27* Bai et al27† Bai et al27‡ Partridge et al51 Inoue et al47 Xie et al38|| Xie et al38† Wang et al36 Han et al29|| Han et al29‡ Jin et al48* Jin et al48‡ Ei Khouli et al45 Yuan et al39 Wang and Duan35|| Wang and Duan35† Wang and Duan35¶ Luo et al34* Luo et al34‡ Zhang et al55|| Zhang et al55‡ Pereira et al52|| Pereira et al52‡ Bogner et al41§
Author
2014 2014 2014 2014 2014 2013 2012 2012 2012 2012 2012 2012 2011 2011 2011 2011 2011 2011 2010 2010 2010 2010 2010 2010 2010 2010 2009 2009 2009 2009 2009 2009 2009 2009 2009 2009 2009
Year
China Korea Portugal China Japan United States China China Turkey China China China China China China China United States Japan China China China China China China China United States China China China China China China China China Brazil Brazil Austria
Country Asians Asians Whites Asians Asians Whites Asians Asians Whites Asians Asians Asians Asians Asians Asians Asians Whites Asians Asians Asians Asians Asians Asians Asians Asians Whites Asians Asians Asians Asians Asians Asians Asians Asians Whites Whites Whites
Ethnicity
TABLE 1. Baseline Characteristics of the Included Studies
Malignant
44 (21–72) 44 (18–69) 52.1 ± 12.4 48.4 ± 11.1 39.1 ± 8.1 49 ± 18 49 (28–70) 44.1 (26–81) 42 ± 13.5 43 ± 16.5 44.9 (19–77) 49.6 (23–75) 49.6 (23–75) 49.6 (23–75) 40.6 (19–68) 49.7 (33–77) 40.6 (19–68) 49.7 (33–77) 40.6 (19–68) 49.7 (33–77) 40.6 (19–68) 49.7 (33–77) 49 (22–85) 53.8 (26–85) 44 (30–62) 44 (30–62) 50 (30–86) 18–75 18–75 18–80 18–80 50 ± 1.7 53.5 ± 11 45 (18-69) 39 (21–67) 43 (32–67) 39 (21–67) 43 (32–67) 39 (21–67) 43 (32–67) 42 (12–72) 42 (12–72) 46.3 (31–77) 46.3 (31–77) 46.1 (22–80) 46.1 (22–80) 52 ± 13
Benign
Age, y
GE Siemens Siemens GE Philips GE GE Philips GE GE GE GE GE GE GE GE GE Siemens GE GE GE Philips Philips GE GE Philips Philips Siemens Siemens Siemens GE GE Philips Philips GE GE Siemens
MRI Machine Type 800 Others Others 800 1000 600 1000 800 600 600 800 1000 400 600 800 1000 600 1000 500 800 800 500 1000 600 1000 600 800 500 800 Others 600 1000 500 1000 500 1000 400
b Value, s/mm2 48 89 53 234 79 35 40 60 108 55 55 55 100 100 100 100 77 105 53 53 50 53 53 56 56 93 36 52 52 52 43 43 54 54 45 45 51
Case Number
57 169 59 234 83 40 40 60 124 57 57 57 100 100 100 100 100 106 66 66 50 57 57 60 60 101 36 60 60 60 47 47 57 57 5 52 41
Lesions
Qu et al J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
www.jcat.org
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
ADC Values of DW-MRI in BC
Data under column “Age” are expressed as the mean ± standard deviation or as the median (interquartile range). *600 s/mm2. †800 s/mm2. ‡1000 s/mm2. §400 s/mm2. ||500 s/mm2. ¶Others.
40 (24–65) 47 42 (18–71) 42 (18–71) 42 (18–71) 42 (18–71) 46.9 (30–72) 46.9 (30–72) 50 (30–70) 53.1 (25–74) 53 (14–88) 38.3 ± 7.1 50.3 ± 11.6
Bogner et al41‡ Jin et al31* Jin et al31‡ Luo et al33 Lou et al32§ Lou et al32* Lou et al32† Lou et al32‡ Feng et al28† Feng et al28‡ Rubesova et al53 Woodhams et al54 Guo et al46
2009 2008 2008 2007 2007 2007 2007 2007 2007 2007 2006 2005 2002
Austria China China China China China China China China China Belgium Japan China
Whites Asians Asians Asians Asians Asians Asians Asians Asians Asians Whites Asians Asians
52 ± 13 — —
Siemens GE GE Siemens Siemens Siemens Siemens Siemens GE GE Siemens GE GE
1000 600 1000 800 400 600 800 1000 800 1000 Others Others Others
51 56 56 52 50 50 50 50 65 65 78 190 52
41 60 60 63 58 58 58 58 68 68 87 191 55
the normal breast tissues and the benign breast tumors suggested that most of the studies were in the range of P > 0.05, and Egger linear regression analysis further confirmed that there was no existence of publication bias (P > 0.05). However, contourenhanced funnel plot for the comparisons between the benign breast tumors and the malignant breast tumors as well as the normal breast tissues and the malignant breast tumors showed that most of the included studies were interspersed in the range of P < 0.01, revealing that there was publication bias, and the Egger linear regression analysis further confirmed publication bias (P > 0.05; Fig. 6).
FIGURE 2. Quality assessment of the included studies by MINORS score. Figure 2 can be viewed online in color at www.jcat.org.
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jcat.org
517
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
Qu et al
DISCUSSION To investigate the clinical value of ADC measurements through DW-MRI as a diagnostic tool to differentiate between benign and malignant breast tumors, a systematic meta-analysis was performed. Our meta-analysis results suggest that the ADC values measured through DW-MRI provide an accurate picture of the internal structure of the tumors and that it is an important radiographic index in the differential diagnosis of benign and malignant breast tumors. Diffusion-weighted MRI, which is a noncontrast MRI technique, provides critical information on cell density, tissue architecture, and integrity of membrane by measuring the diffusion of water molecule, with lower loss of signal representing the low water diffusion and higher loss of signal representing high water diffusion.58,59 Apparent diffusion coefficient is a frequently used parameter to quantify diffusibility of tissues by correlating the active
movement or attenuation of the motion of water molecules to tissue complexity; thus, ADC values are high in tissues without obstacles for water motion and are low in tissues with obstacles.60,61 In tissues, apparent diffusion of water molecules is altered by their interactions with cell membranes and macromolecules, and in tumor tissues with high cell density, the disorganized extracellular space and the higher density of hydrophobic cellular membranes restrict the apparent diffusion of water. Apparent diffusion coefficient is quantified by mean diffusivity measurement in 3 orthogonal directions, which is influenced by tissue cellularity, fluid viscosity, membrane permeability, and blood flow.62 A few studies have reported that the varied treatment responses reflecting diverse clinical outcomes are directly linked to the tumor morphology and could be predicted from ADC values, that is, high pretreatment ADC indicated high viability and shows a favorable clinical response and low ADC values indicated necrosis.12 With specific relevance to
FIGURE 3. Forest plots for the comparisons of ADC values between the benign breast tumors and the malignant breast tumors. Figure 3 can be viewed online in color at www.jcat.org.
518
www.jcat.org
© 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Comput Assist Tomogr • Volume 39, Number 4, July/August 2015
ADC Values of DW-MRI in BC
FIGURE 4. Forest plots for the comparisons of ADC values between the normal breast tissues and the benign breast tumors. Figure 4 can be viewed online in color at www.jcat.org.
FIGURE 5. Forest plots for the comparisons of ADC values between the normal breast tissues and the malignant breast tumors. Figure 5 can be viewed online in color at www.jcat.org. © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jcat.org
519
520
Ethnicity Asians Whites Machine type GE Siemens Philips b Value 400 500 600 800 1000 Others
1.68–2.43 0.94–3.58
1.03–2.04 2.27–4.00 1.79–2.80
0.61–1.30 1.28–3.68 0.06–1.66 1.52–3.30 1.47–2.80 2.40–5.61
1.53 3.13 2.29
0.96 2.48 0.86 2.41 2.14 4.01
95% CI
2.05 2.26
SMD
Benign vs Malignant
www.jcat.org
