- Email: [email protected]
Real-time elastography for the differentiation of benign and malignant superficial lymph nodes: A meta-analysis
European Journal of Radiology 81 (2012) 2576–2584
Contents lists available at SciVerse ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Real-time elastography for the differentiation of benign and malignant superficial lymph nodes: A meta-analysis Li Ying ∗ , Yao Hou, Hua-Min Zheng, Xiao Lin, Zuo-Liu Xie, Yuan-Ping Hu Department of Ultrasonography, the First Affiliated Hospital of Wenzhou Medical College, No. 2 Fuxue Lane, Wenzhou 325000, China
a r t i c l e
i n f o
Article history: Received 13 September 2011 Received in revised form 31 October 2011 Accepted 31 October 2011 Keywords: Diagnostic accuracy Sensitivity Specificity Lymph nodes Elastography
a b s t r a c t Background: Real-time elastography (RTE), as a non-invasive method, is used for the classification of benign and malignant lymph nodes (LNs) and developed as an alternative to biopsy. Elasticity score (ES) and strain ratio (SR) are used for the interpretation of RTE. We studied the performance of RTE for diagnosis of malignant LNs using meta-analysis. Methods: PubMed, the Cochrane Library, ISI Web of Knowledge, China National Knowledge Infrastructure were searched. The studies published in English or Chinese relating to the diagnostic value of RTE for superficial LNs were collected. Hierarchical summary receiver operating characteristic (HSROC) curve was used to examine the RTE accuracy. Clinical utility of RTE for LNs was evaluated by Fagan plot analysis. Results: A total of 9 studies which included 835 LNs were analyzed. The summary sensitivity and specificity for the diagnosis of malignant LNs were 0.74 (95% confidence interval (CI), 0.66–0.81) and 0.90 (95% CI, 0.82–0.94) for ES, and 0.88 (95% CI, 0.79–0.93) and 0.81 (95% CI, 0.49–0.95) for SR, respectively. Compared to ES, SR obviously improved the diagnostic sensitivity value. The HSROCs were 0.88 for ES and 0.91 for SR, respectively. After RTE results over the cut-off value for malignant LNs (“positive” result), the corresponding post-test probability for the presence (if pre-test probability was 50%) was 88% for ES and 82% for SR, respectively; while, in “negative” measurement, the post-test probability was 22% and 13%, respectively. Conclusion: RTE has a high accuracy in the classification of superficial LNs and can potentially help to select suspicious LNs for biopsy. © 2011 Elsevier Ireland Ltd. All rights reserved.
1. Introduction The status of lymph nodes (LNs) is so significant that it is used for cancer staging which decides the treatment to be employed, and for determining the prognosis [1]. Various imaging techniques, including computed tomography (CT), magnetic resonance imaging (MRI), and ultrasonography are used to evaluate the possible characteristics of LNs [2]. Because of its simplest and least expensive, ultrasonography is the most extensively used for classification of superficial LNs. However, no single ultrasonography criterion for malignant LNs had satisfactory sensitivity and specificity [3].
Abbreviations: CI, confidence interval; CT, computed tomography; DOR, diagnostic odds ratio; ES, elasticity score; HSROC, hierarchical summary receiver operating characteristic; LNs, lymph nodes; MRI, magnetic resonance imaging; NPV, negative predictive value; PPV, positive predictive value; QUADAS, quality assessment of diagnostic accuracy studies; RTE, real-time elastography; SR, strain ratio. ∗ Corresponding author. Tel.: +86 577 88078232; fax: +86 577 88078262. E-mail addresses: [email protected] (L. Ying), [email protected] (Y. Hou), zhenghm [email protected] (H.-M. Zheng), Xiaolin [email protected] (X. Lin), [email protected] (Z.-L. Xie), [email protected] (Y.-P. Hu). 0720-048X/$ – see front matter © 2011 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2011.10.026
A simple, reliable, noninvasive imaging method for identification of malignant LNs is needed. Real-time elastography (RTE) is a new ultrasound technique that evaluates the elasticity of any given tissue. It is a complimentary imaging technique to the conventional ultrasonography and could potentially reduce unnecessary biopsies [4]. Because RTE combines the penetration depth and resolution of ultrasound with high sensitivity to stiffness contrast, it can detect the nodules including superficial LNs, which elude palpation by virtue of their small size or their deep location within the body [5]. Elasticity score (ES) and strain ratio (SR) are used for the interpretation of RTE [6]. Four-RTE pattern is the most frequently used ES in detecting malignant LNs. In this scheme, RTE pattern 1 is a nodule that displays predominantly purple, green or yellow with less than 10% displaying red (soft), RTE pattern 2 is a nodule that displays predominantly in yellow or green and with red areas comprising between 10% and 50% (moderately soft), RTE pattern 3 is a nodule that displays predominantly in red and with yellow or green areas comprising between 10% and 50% (moderately stiff), and RTE pattern 4 is a nodule that displays predominantly red and with less than 10% appearing yellow or green (stiff). RTE pattern 1 and 2 indicate benign LNs and RTE pattern 3 and 4 indicate malignant LNs [7]. The SR is the ratio
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
between the LNs and the background tissue, of which a cut-off value would be set for identification of malignant LNs. In the present study, we used meta-analysis to assess the performance of RTE in the classification of superficial LNs. 2. Materials and methods 2.1. Search strategy and selection criteria We searched PubMed, the Cochrane Library, ISI Web of Knowledge and China National Knowledge Infrastructure for studies published prior to August 30th, 2011, by using the following search terms: elastography, sonoelastography, real-time tissue elastography, lymph node, superficial, metastasis, diagnosis, and diagnostic test. A manual search was also done by using references of eligible articles. Language was limited to English or Chinese. Two investigators independently assessed reports for eligibility. To be included, studies had to meet the following inclusion criteria: (1) the study evaluated the performance of the RTE for the differentiation of benign and malignant LNs, with qualitative (ES) or quantitative (SR) measurements. (2) Using appropriate cytology acquired by fine needle aspiration biopsy or histology by surgery or imaging findings (contrast CT, or contrast MIR, or positron emission tomography) as reference standard for the diagnosis of malignancy. (3) Reported on data necessary to calculate the true positive, false positive, true negative and false negative diagnostic results of RTE for the differentiation of benign and malignant LNs. If such data were unavailable, the corresponding author was contacted via to provide them; if he/she failed to reply, the study was excluded. 2.2. Data extraction and quality assessment All data were extracted by two investigators independently, with disagreements resolved in consultation with a third investigator. For each study, the following information was abstracted according to a fixed protocol: author, study publication year, country, number of patients, number of LNs available for analysis, rate of the malignant LNs, patient age, gender, measurement, ES score with cut-off value, and SR cut-off value. The true positive, false positive, true negative and false negative diagnostic results of ES or SR measurements were extracted allowing the calculation of sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for each reported test threshold. The quality of the studies included in the meta-analysis was assessed using quality assessment of diagnostic accuracy studies (QUADAS) questionnaire, which was designed to assess the internal and external validity of diagnostic accuracy studies included in systematic reviews [8]. The QUADAS tool has 14 items that assess study designrelated issues, and the validity of the results of the study. Each item may be scored “yes” if reported; “no” if not reported; or “unclear” if there is no adequate information in the article to make an accurate judgment.
between primary studies, the random effect method was used for pooled analyses. To explore sources of heterogeneity in the studies, a metaregression technique was used according to the following predefined characteristics: study design (retrospective versus prospective), blinded interpretation of RTE and reference standard (yes versus no), number of patients, median age, male proportion, location of study (Asian or not), number of LNs, prevalence of malignant LNs, LNs position (Cervical or not), language (English versus Chinese), and QUADAS score. A p value of <0.05 was considered to be representative of statistically significance. Deeks’ funnel plot asymmetry test was used to investigate publication bias, in which formal testing for publication bias may be conducted by a regression of diagnostic log odds ratio against 1/sqrt (effective sample size), weighting by effective sample size, with p < 0.10 for the slope coefficient indicating significant asymmetry [9]. Sensitivity analysis was also performed by excluding studies in which patients were not all referred to histopathologic findings. We evaluated pre-test probabilities of 25%, 50%, and 75% versus corresponding post-test probabilities following a “positive” or “negative” RTE result based on the summary sensitivity and specificity using a Fagan plot, which showed the relationship between the prior probability specified, the likelihood ratio, and posterior test probability [10]. “Positive” RTE results were defined as all results above the optimal cut-off value for malignant LNs, given in each individual study, while “negative” test results were all results below the same cut-off value. MIDAS and METANDI modules in Stata 11.0 (College Station, TX) were used for statistical analysis. 3. Result 3.1. Characteristics of studies in our analysis Of the 78 references identified, 22 potentially relevant studies were identified for evaluation. Ultimately, 13 studies were excluded for undesirable article types (n = 11), not written English or Chinese (n = 1), and insufficient data (n = 1). Thus, 9 studies were included in our final dataset for the meta-analysis. The flowchart of study selection was shown in Fig. 1.
2.3. Statistical analysis and data synthesis Summary sensitivities and specificities, and diagnostic odds ratio (DOR) (with corresponding 95% confidence interval (CI)) were used to examine the ES and SR accuracy for the differentiation of benign and malignant LNs. The DOR expresses how much greater the odds of having the disease are for the people with a positive test result than for the people with a negative test result. Hierarchical summary receiver operating characteristic (HSROC) curve was also plotted to graphically present the results. The between-study heterogeneity was evaluated by computing Higgins’s I2 and 2 tests for heterogeneity using the generic inverse variance method of metaanalysis of DOR. An I2 value of more than 50% or a 2 p value of 0.10 was considered substantial heterogeneity. If heterogeneity existed
2577
Fig. 1. Flowchart of study selection. ES: elasticity score; SR: strain ratio.
2578
Table 1 Main characteristics of all studies included in the meta-analysis. LN position
Primary tumors of LN
Reference
Age (year)
Male (%)
RTE measurement
ES scale (cut-off)
SR cut-off
QUADAS score
141
42.6
Cervical
Papillary thyroid cancer, follicular thyroid adenoma, squamous cell cancer of the hypopharynx
Histopathologic
58
48.8
ES/SR
4 (2/3)
1.5
14
37
85
62.4
Cervical
Squamous cell carcinoma, thyroid carcinoma, breast cancer, lung cancer
Histopathologic/ imaging
55
67.6
ES
5 (2/3)
–
11
Italy
53
53
52.8
Cervical
Metastatic, non-Hodgkin lymphomas
Histopathologic
NA
NA
ES
4 (2/3)
–
6
[14]
China
82
155
56.1
Cervical/ supraclavicular region/ armpit/groin
Thyroid cancer, breast cancer, Hashimoto’s disease combined with adenoma, lung cancer, nasopharyngeal carcinoma, lymphoma, primitive neuroectodermal tumors
Histopathologic/ imaging
38.2
NA
ES/SR
4 (2/3)
2.395
8
Bhatia et al., 2010
[7]
China
74
74
50.0
Cervical
Lung carcinoma, breast adenocarcinoma, nasopharyngeal carcinoma, thyroid papillary carcinoma, squamous carcinoma, uterine mixed mullerian tumor, ovarian adenocarcinoma
Histopathologic
50
41.9
ES
4 (2/3)
–
14
Shi et al., 2010
[13]
China
37
85
62.4
Cervical
Lung cancer, thyroid carcinoma, breast cancer, gastric cancer
Histopathologic
55
67.6
ES
5 (2/3)
–
13
Tan, 2010
[15]
China
107
128
54.7
Cervical
Non-Hodgkin’s lymphoma, leukemia, Hodgkin’s lymphoma, squamous cell carcinoma, melanoma, adenocarcinoma
Histopathologic
53.4
52.3
ES/SR
4 (2/3)
1.5
13
Choi et al., 2011
[16]
Korea
62
64
48.4
Axillary
Breast cancer
Histopathologic
53
0
ES/SR
4 (2/3)
2.3
13
Taylor et al., 2011
[17]
UK
50
50
42.0
Axillary
Breast cancer
Histopathologic
57
0
ES/SR
4 (2/3)
0.65
13
Ref.
Country
No. of patients
Lyshchik et al., 2007
[5]
Japan
43
Alam et al., 2008
[11]
Japan
Rubaltelli et al., 2009
[12]
Zhang et al., 2009
No. of LN
ES: elasticity score; LNs: lymph nodes; NA: not available; QUADAS: quality assessment of diagnostic accuracy studies. RTE: real-time elastography; SR: strain ratio.
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
Malignant LN rate (%)
Author, year
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Unclear Yes Unclear Unclear Yes Yes Yes Yes Yes No Yes Yes
Yes Yes Yes Yes
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Yes
Yes Yes Yes Unclear Unclear Yes Yes Yes No No Unclear Yes Yes
No
Yes Unclear No Unclear Unclear No Yes Yes Yes Yes Unclear No Unclear
Yes
Yes Yes Yes Yes Yes Yes Yes Yes No No Yes Yes Yes
No
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Intermediate results Clinical review bias Diagnostic review bias Test review bias Reference execution details Test execution details Incorporation bias Selection criteria Spectrum composition
2579
Lyshchik et al., 2007 Alam et al., 2008 Rubaltelli et al., 2009 Zhang et al., 2009 Bhatia et al., 2010 Shi et al., 2010 Tan, 2010 Choi et al., 2011 Taylor et al., 2011
The Fagan plot demonstrated that ES measurement was very informative lowering the negative post-probability of malignant LNs to as low as 9% when “negative” measurement from 25% preprobability; however, only 71% probability of correctly diagnosed malignant LNs following a “positive” measurement (Fig. 5A). When
Author, year
3.4. Fagan plot analysis
Table 2 Quality assessment of included studies.
3.3. Publication bias According to Deeks’ funnel plot asymmetry test, there were no publication bias among ES studies (p = 0.74, Supplementary Fig. 1A) and RS studies (p = 0.35, Supplementary Fig. 1B).
Differential verification bias
Q2 Q1
Q3
Q4
Q5
Q6
Q7
Q8
Q9
Q10
The summary sensitivity and specificity of the ES measurement for the differentiation of benign and malignant LNs were 0.74 (95% CI, 0.66–0.81) and 0.90 (95% CI, 0.82–0.94), respectively (Fig. 2A). The summary DOR was 24.84 (95% CI, 11.59–53.24, Fig. 3A), and the HSROC was 0.88 (95% CI, 0.85–0.91) (Fig. 4A). Based on these values, and assuming a 62.7% malignant LNs (as observed in the included studies), the estimated PPV and NPV were 0.76 (95% CI, 0.72–0.80) and 0.86 (95% CI, 0.81–0.91), respectively. Sensitivity analysis was performed after excluding two studies in which patients were not all referred to histopathologic findings [11,14]. The results were not influenced excessively after omitting the two studies. There was statistically significant heterogeneity in DOR (2 p < 0.001, I2 = 98.38%). However, according to the meta-regression analysis, ES measurement accuracy was not affected by the covariates. The summary sensitivity of the SR measurement was 0.88 (95% CI, 0.79–0.93) (Fig. 2B), which was significantly higher than it was for ES measurement (p < 0.001). The summary specificity of the SR measurement was 0.81 (95% CI, 0.49–0.95) (Fig. 2B), which was lower than that of ES measurement (p < 0.001). The summary DOR was 31.63 (95% CI, 9.95–100.56) (Fig. 3B), and the HSROC was 0.91 (95% CI, 0.88–0.93) for the SR measurement (Fig. 4B). The accuracy of the SR measurement was similar to that of ES measurement (p = 0.573). Based on these values, and assuming a 49.6% malignant LNs (as observed in the included studies), the estimated PPV and NPV were 0.86 (95% CI, 0.74–0.97) and 0.81 (95% CI, 0.70–0.92), respectively. Sensitivity analysis was performed after excluding one study in which patients who were not all referred to histopathologic findings [14]. The results were not influenced excessively after omitting the study. There was statistically significant heterogeneity in DOR (2 p < 0.001, I2 = 99.99%). However, according to the meta-regression analysis, SR measurement accuracy was not affected by the covariates.
Partial verification bias
Q11
3.2. Accuracy of RTE for the differentiation of benign and malignant LNs
Appropriate Disease reference progression standard bias
Q12
Q13
Q14
Table 1 summarized the main characteristics of the included studies. A total of 835 LNs (395 benign, 440 malignant) in 545 patients (median age, 51.5 years) were evaluated. Some variation of the classifications of RTE for LNs existed in the studies. Four studies performed ES measurement [7,11–13], and 5 studies performed both ES and SR measurement [5,14–17]. ES measurement, 7 described the measurement on a numerical scale of 1–4, 2 described a scale of 1–5. Histopathologic findings or imaging findings were used as the reference standard for the final classification of LNs in two studies [11,14]. Seven studies included patients who were only referred to histopathologic findings. According to the QUADAS scale, included studies had a very good methodological quality except two studies [12,14] (Table 2).
Withdrawals
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
2580
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
Fig. 2. Forest sensitivity and specificity. (A) ES for LNs and (B) SR for LNs. CI: confidence interval; ES: elasticity score; LNs: lymph nodes; SR: strain ratio.
there is a high pre-test index of suspicion (pre-test probability = 50% or 75%), then the probability of a correct diagnosis following a “positive” measurement reached or exceeded 90% for malignant LNs, but the diagnosis would be wrong in 22–46% of patients with a “negative” measurement (Fig. 5B and C).
SR measurement was very informative with 82% probability of correctly diagnosing malignant LNs following a “positive” measurement when pre-test probability was 50% and lowering the probability of disease to as low as 13% when “negative” measurement (Fig. 6B). However, when pre-test probability was 25%,
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
2581
Fig. 3. Forest diagnostic odds ratio. (A) ES for LNs and (B) SR for LNs. CI: confidence interval; ES: elasticity score; LNs: lymph nodes; SR: strain ratio.
Fig. 4. HSROC curve of the RTE for classification of LNs. (A) ES for LNs and (B) SR for LNs. The size of the dots for 1-specificity and sensitivity of the single studies in the ROC space was derived from the respective sample size. ES: elasticity score; HSROC: hierarchical summary receiver operating characteristic; LNs: lymph nodes; SR: strain ratio.
2582
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
Fig. 5. Fagan plot analysis to evaluate the clinical utility of ES for classification of LNs. (A) Pre-test probability = 25%; (B) Pre-test probability = 50% and (C) Pre-test probability = 75%. The Fagan plot consisted of a vertical axis on the left with the pre-test probability, an axis in the middle representing the likelihood ratio and a vertical axis on the right representing the post-test probability. ES: elasticity score; LNs: lymph nodes; NLR: negative likelihood ratio; PLR: positive likelihood ratio.
Fig. 6. Fagan plot analysis to evaluate the clinical utility of SR for classification of LNs. (A) Pre-test probability = 25%; (B) Pre-test probability = 50% and (C) Pre-test probability = 75%. The Fagan plot consisted of a vertical axis on the left with the pre-test probability, an axis in the middle representing the likelihood ratio and a vertical axis on the right representing the post-test probability. LNs: lymph nodes; NLR: negative likelihood ratio; PLR: positive likelihood ratio; SR: strain ratio.
SR measurement had only 61% probability of correctly diagnosing malignant LNs following a “positive” measurement (Fig. 6A). In addition, the diagnosis would be wrong in 31% of patients with a “negative” measurement when pre-test probability was 75% (Fig. 6C).
4. Discussion In this meta-analysis, we evaluated the performance of RTE, a non-invasive technique, for the classification of superficial LNs. The results indicated that RTE had a high accuracy for identification of
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
malignant LNs. The HSROC for identification of malignant LNs by RTE was 0.88 (95% CI, 0.85–0.91) for ES measurement; SR measurement also had a high HSROC (0.91 (95% CI, 0.88–0.93)). In this respect, RTE is promising and worthy to translate into clinical practice that it is a reliable and noninvasive procedure and is able to reveal LNs with small size or deep location. Therefore, RTE could be integrated in the classification of superficial LNs. RTE might also be useful to select suspicious LNs for biopsy. In addition, in patients with cancer, RTE could be used for cancer staging by identifying LNs most likely to be malignant [5]. The ease of availability and application, non-invasive nature of RTE are set to make it increasingly popular among radiologist. Although, ES measurement had a good diagnostic accuracy, it depended on the examiner resulting in significant inter-observer variability. Therefore, a quantitative method of analysis for elasticity images of LNs is needed. The SR measurement is a quantitative approach that could relatively objectively evaluate the LNs stiffness. Although the accuracy of SR measurement was similar to ES measurement, it obviously improved the diagnostic sensitivity value. A strength of our study was that the Fagan plot analysis had been used for exploring clinical utilities of the RTE. Our results show that SR measurement could be used to classify the superficial LNs (when pre-test probability = 50%), with 82% probability of correctly diagnosing LNs following a “positive” measurement. A “negative” measurement was also informative, as malignant LNs were present in only 13% of patients. However, results were less promising for ES measurement. Although the probability of correctly identification of malignant LNs could reach 88% following a “positive” measurement, the diagnosis would be wrong in 22% of patients with a “negative” measurement. The above results, especially the following a “positive” SR measurement, have been considered encouraging in individual studies and in some reviews. However, the major drawback was that the SR cut-off value was different across studies and has not been validated [15,16]. Pooling such “optimal” results from these studies might artificially increase the summary sensitivity and specificity. Because of a scarcity of publications, we could not address these important issues. The efficacy of SR measurement in the classification of superficial LNs should be further evaluated in more studies in the future. Because significant heterogeneity was present in this analysis, our study explored factors that may be the source of heterogeneity by meta-regression analysis. Although 11 specific covariates of patient and study were examined, none was found that could affect ES measurement accuracy or SR measurement accuracy. We could not explain this finding. Specifically, LNs position, which might influence the stiffness value, was not significant in the metaregression analysis. Despite its potential, there were some technique limitations of RTE. The RTE technique is labor intensive and time consuming, requiring additional time for off-line computer processing. In addition, some artifacts would be caused by no standardized compression load applied with freehand RTE and motion of surrounding tissues and vessels during compression scanning [5,18]. Therefore, investigators in future studies need to analyze inter and intra-observer variability. Some limitations of this study should be taken into consideration: first, in our meta-analysis, the results were pooled for LNs with different positions. The different depth of the LNs could influence the stiffness value. The perception of the pressure and frequency of some deep LNs such as LNs in the supraclavicular or auxiliary areas will be affected because of signal attenuating [15]. We would encourage investigators to be rigorous in their patient selection in future studies. In addition, significant heterogeneity was also
2583
present in the evaluation of ES measurement and SR measurement accuracy in these studies, and therefore interpretation should be cautious. In conclusion, our meta-analysis shows that RTE could be used as a good identification tool for malignant LNs, with an exceeded 80% disease probability following a “positive” measurement. A “negative” measurement is also accurate and informative for SR measurement, with disease being present in only 13% of patients, but not for ES measurement. It is necessary to further evaluate the potential role of RTE in a large, prospective, international, multicenter study. Conflicts of interest No conflicts of interest exist for all authors. Acknowledgement The authors thank Ke-Qing Shi, MD, Department of Infection and Liver Diseases, Liver Research Center, the First Affiliated Hospital of Wenzhou Medical College, Wenzhou 325000, China, for his valuable statistical assistance. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.ejrad.2011.10.026. References [1] Lecuru F, Mathevet P, Querleu D, et al. Bilateral negative sentinel nodes accurately predict absence of lymph node metastasis in early cervical cancer: results of the SENTICOL study. J Clin Oncol 2011;29(13):1686–91. [2] Choi HJ, Ju W, Myung SK, et al. Diagnostic performance of computer tomography, magnetic resonance imaging, and positron emission tomography or positron emission tomography/computer tomography for detection of metastatic lymph nodes in patients with cervical cancer: meta-analysis. Cancer Sci 2010;101(6):1471–9. [3] Gritzmann N, Hollerweger A, Macheiner P, et al. Sonography of soft tissue masses of the neck. J Clin Ultrasound 2002;30(6):356–73. [4] Tan SM, Teh HS, Mancer JF, et al. Improving B mode ultrasound evaluation of breast lesions with real-time ultrasound elastography: a clinical approach. Breast 2008;17(3):252–7. [5] Lyshchik A, Higashi T, Asato R, et al. Cervical lymph node metastases: diagnosis at sonoelastography: initial experience. Radiology 2007;243(1): 258–67. [6] Landoni V, Francione V, Marzi S, et al. Quantitative analysis of elastography images in the detection of breast cancer. Eur J Radiol 2011, doi:10.1016/j.ejrad.2011.04.012. [7] Bhatia KS, Cho CC, Yuen YH, et al. Real-time qualitative ultrasound elastography of cervical lymph nodes in routine clinical practice: interobserver agreement and correlation with malignancy. Ultrasound Med Biol 2010;36(12):1990–7. [8] Whiting P, Rutjes AW, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003;3:25. [9] Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol 2005;58(9):882–93. [10] Hellmich M, Lehmacher W. A ruler for interpreting diagnostic test results. Methods Inf Med 2005;44(1):124–6. [11] Alam F, Naito K, Horiguchi J, et al. Accuracy of sonographic elastography in the differential diagnosis of enlarged cervical lymph nodes: comparison with conventional B-mode sonography. AJR Am J Roentgenol 2008;191(2): 604–10. [12] Rubaltelli L, Stramare R, Tregnaghi A, et al. The role of sonoelastography in the differential diagnosis of neck nodules. J Ultrasound 2009;12(3):93–100. [13] Shi GH, Wang XM, Ou GC, et al. Comparative study of ultrasonic elastography with conventional ultrasonography in cervical lymph nodes [in Chinese]. Chin J Ultrasound Med 2010;26(8):730–3. [14] Zhang YR, Lv Q, Yin YH, et al. The value of ultrasound elastography in differential diagnosis of superficial lymph nodes. Front Med China 2009;3(3):368–74. [15] Tan R, Xiao Y, He Q. Ultrasound elastography: its potential role in assessment of cervical lymphadenopathy. Acad Radiol 2010;17(7):849–55. [16] Choi JJ, Kang BJ, Kim SH, et al. Role of sonographic elastography in the differential diagnosis of axillary lymph nodes in breast cancer. J Ultrasound Med 2011;30(4):429–36.
2584
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
[17] Taylor K, O‘Keeffe S, Britton PD, et al. Ultrasound elastography as an adjuvant to conventional ultrasound in the preoperative assessment of axillary lymph nodes in suspected breast cancer: a pilot study. Clin Radiol 2011;66(11):1064–71.
[18] Havre RF, Elde E, Gilja OH, et al. Freehand real-time elastography: impact of scanning parameters on image quality and in vitro intra- and interobserver validations. Ultrasound Med Biol 2008;34(10):1638–50.
Contents lists available at SciVerse ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Real-time elastography for the differentiation of benign and malignant superficial lymph nodes: A meta-analysis Li Ying ∗ , Yao Hou, Hua-Min Zheng, Xiao Lin, Zuo-Liu Xie, Yuan-Ping Hu Department of Ultrasonography, the First Affiliated Hospital of Wenzhou Medical College, No. 2 Fuxue Lane, Wenzhou 325000, China
a r t i c l e
i n f o
Article history: Received 13 September 2011 Received in revised form 31 October 2011 Accepted 31 October 2011 Keywords: Diagnostic accuracy Sensitivity Specificity Lymph nodes Elastography
a b s t r a c t Background: Real-time elastography (RTE), as a non-invasive method, is used for the classification of benign and malignant lymph nodes (LNs) and developed as an alternative to biopsy. Elasticity score (ES) and strain ratio (SR) are used for the interpretation of RTE. We studied the performance of RTE for diagnosis of malignant LNs using meta-analysis. Methods: PubMed, the Cochrane Library, ISI Web of Knowledge, China National Knowledge Infrastructure were searched. The studies published in English or Chinese relating to the diagnostic value of RTE for superficial LNs were collected. Hierarchical summary receiver operating characteristic (HSROC) curve was used to examine the RTE accuracy. Clinical utility of RTE for LNs was evaluated by Fagan plot analysis. Results: A total of 9 studies which included 835 LNs were analyzed. The summary sensitivity and specificity for the diagnosis of malignant LNs were 0.74 (95% confidence interval (CI), 0.66–0.81) and 0.90 (95% CI, 0.82–0.94) for ES, and 0.88 (95% CI, 0.79–0.93) and 0.81 (95% CI, 0.49–0.95) for SR, respectively. Compared to ES, SR obviously improved the diagnostic sensitivity value. The HSROCs were 0.88 for ES and 0.91 for SR, respectively. After RTE results over the cut-off value for malignant LNs (“positive” result), the corresponding post-test probability for the presence (if pre-test probability was 50%) was 88% for ES and 82% for SR, respectively; while, in “negative” measurement, the post-test probability was 22% and 13%, respectively. Conclusion: RTE has a high accuracy in the classification of superficial LNs and can potentially help to select suspicious LNs for biopsy. © 2011 Elsevier Ireland Ltd. All rights reserved.
1. Introduction The status of lymph nodes (LNs) is so significant that it is used for cancer staging which decides the treatment to be employed, and for determining the prognosis [1]. Various imaging techniques, including computed tomography (CT), magnetic resonance imaging (MRI), and ultrasonography are used to evaluate the possible characteristics of LNs [2]. Because of its simplest and least expensive, ultrasonography is the most extensively used for classification of superficial LNs. However, no single ultrasonography criterion for malignant LNs had satisfactory sensitivity and specificity [3].
Abbreviations: CI, confidence interval; CT, computed tomography; DOR, diagnostic odds ratio; ES, elasticity score; HSROC, hierarchical summary receiver operating characteristic; LNs, lymph nodes; MRI, magnetic resonance imaging; NPV, negative predictive value; PPV, positive predictive value; QUADAS, quality assessment of diagnostic accuracy studies; RTE, real-time elastography; SR, strain ratio. ∗ Corresponding author. Tel.: +86 577 88078232; fax: +86 577 88078262. E-mail addresses: [email protected] (L. Ying), [email protected] (Y. Hou), zhenghm [email protected] (H.-M. Zheng), Xiaolin [email protected] (X. Lin), [email protected] (Z.-L. Xie), [email protected] (Y.-P. Hu). 0720-048X/$ – see front matter © 2011 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2011.10.026
A simple, reliable, noninvasive imaging method for identification of malignant LNs is needed. Real-time elastography (RTE) is a new ultrasound technique that evaluates the elasticity of any given tissue. It is a complimentary imaging technique to the conventional ultrasonography and could potentially reduce unnecessary biopsies [4]. Because RTE combines the penetration depth and resolution of ultrasound with high sensitivity to stiffness contrast, it can detect the nodules including superficial LNs, which elude palpation by virtue of their small size or their deep location within the body [5]. Elasticity score (ES) and strain ratio (SR) are used for the interpretation of RTE [6]. Four-RTE pattern is the most frequently used ES in detecting malignant LNs. In this scheme, RTE pattern 1 is a nodule that displays predominantly purple, green or yellow with less than 10% displaying red (soft), RTE pattern 2 is a nodule that displays predominantly in yellow or green and with red areas comprising between 10% and 50% (moderately soft), RTE pattern 3 is a nodule that displays predominantly in red and with yellow or green areas comprising between 10% and 50% (moderately stiff), and RTE pattern 4 is a nodule that displays predominantly red and with less than 10% appearing yellow or green (stiff). RTE pattern 1 and 2 indicate benign LNs and RTE pattern 3 and 4 indicate malignant LNs [7]. The SR is the ratio
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
between the LNs and the background tissue, of which a cut-off value would be set for identification of malignant LNs. In the present study, we used meta-analysis to assess the performance of RTE in the classification of superficial LNs. 2. Materials and methods 2.1. Search strategy and selection criteria We searched PubMed, the Cochrane Library, ISI Web of Knowledge and China National Knowledge Infrastructure for studies published prior to August 30th, 2011, by using the following search terms: elastography, sonoelastography, real-time tissue elastography, lymph node, superficial, metastasis, diagnosis, and diagnostic test. A manual search was also done by using references of eligible articles. Language was limited to English or Chinese. Two investigators independently assessed reports for eligibility. To be included, studies had to meet the following inclusion criteria: (1) the study evaluated the performance of the RTE for the differentiation of benign and malignant LNs, with qualitative (ES) or quantitative (SR) measurements. (2) Using appropriate cytology acquired by fine needle aspiration biopsy or histology by surgery or imaging findings (contrast CT, or contrast MIR, or positron emission tomography) as reference standard for the diagnosis of malignancy. (3) Reported on data necessary to calculate the true positive, false positive, true negative and false negative diagnostic results of RTE for the differentiation of benign and malignant LNs. If such data were unavailable, the corresponding author was contacted via to provide them; if he/she failed to reply, the study was excluded. 2.2. Data extraction and quality assessment All data were extracted by two investigators independently, with disagreements resolved in consultation with a third investigator. For each study, the following information was abstracted according to a fixed protocol: author, study publication year, country, number of patients, number of LNs available for analysis, rate of the malignant LNs, patient age, gender, measurement, ES score with cut-off value, and SR cut-off value. The true positive, false positive, true negative and false negative diagnostic results of ES or SR measurements were extracted allowing the calculation of sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for each reported test threshold. The quality of the studies included in the meta-analysis was assessed using quality assessment of diagnostic accuracy studies (QUADAS) questionnaire, which was designed to assess the internal and external validity of diagnostic accuracy studies included in systematic reviews [8]. The QUADAS tool has 14 items that assess study designrelated issues, and the validity of the results of the study. Each item may be scored “yes” if reported; “no” if not reported; or “unclear” if there is no adequate information in the article to make an accurate judgment.
between primary studies, the random effect method was used for pooled analyses. To explore sources of heterogeneity in the studies, a metaregression technique was used according to the following predefined characteristics: study design (retrospective versus prospective), blinded interpretation of RTE and reference standard (yes versus no), number of patients, median age, male proportion, location of study (Asian or not), number of LNs, prevalence of malignant LNs, LNs position (Cervical or not), language (English versus Chinese), and QUADAS score. A p value of <0.05 was considered to be representative of statistically significance. Deeks’ funnel plot asymmetry test was used to investigate publication bias, in which formal testing for publication bias may be conducted by a regression of diagnostic log odds ratio against 1/sqrt (effective sample size), weighting by effective sample size, with p < 0.10 for the slope coefficient indicating significant asymmetry [9]. Sensitivity analysis was also performed by excluding studies in which patients were not all referred to histopathologic findings. We evaluated pre-test probabilities of 25%, 50%, and 75% versus corresponding post-test probabilities following a “positive” or “negative” RTE result based on the summary sensitivity and specificity using a Fagan plot, which showed the relationship between the prior probability specified, the likelihood ratio, and posterior test probability [10]. “Positive” RTE results were defined as all results above the optimal cut-off value for malignant LNs, given in each individual study, while “negative” test results were all results below the same cut-off value. MIDAS and METANDI modules in Stata 11.0 (College Station, TX) were used for statistical analysis. 3. Result 3.1. Characteristics of studies in our analysis Of the 78 references identified, 22 potentially relevant studies were identified for evaluation. Ultimately, 13 studies were excluded for undesirable article types (n = 11), not written English or Chinese (n = 1), and insufficient data (n = 1). Thus, 9 studies were included in our final dataset for the meta-analysis. The flowchart of study selection was shown in Fig. 1.
2.3. Statistical analysis and data synthesis Summary sensitivities and specificities, and diagnostic odds ratio (DOR) (with corresponding 95% confidence interval (CI)) were used to examine the ES and SR accuracy for the differentiation of benign and malignant LNs. The DOR expresses how much greater the odds of having the disease are for the people with a positive test result than for the people with a negative test result. Hierarchical summary receiver operating characteristic (HSROC) curve was also plotted to graphically present the results. The between-study heterogeneity was evaluated by computing Higgins’s I2 and 2 tests for heterogeneity using the generic inverse variance method of metaanalysis of DOR. An I2 value of more than 50% or a 2 p value of 0.10 was considered substantial heterogeneity. If heterogeneity existed
2577
Fig. 1. Flowchart of study selection. ES: elasticity score; SR: strain ratio.
2578
Table 1 Main characteristics of all studies included in the meta-analysis. LN position
Primary tumors of LN
Reference
Age (year)
Male (%)
RTE measurement
ES scale (cut-off)
SR cut-off
QUADAS score
141
42.6
Cervical
Papillary thyroid cancer, follicular thyroid adenoma, squamous cell cancer of the hypopharynx
Histopathologic
58
48.8
ES/SR
4 (2/3)
1.5
14
37
85
62.4
Cervical
Squamous cell carcinoma, thyroid carcinoma, breast cancer, lung cancer
Histopathologic/ imaging
55
67.6
ES
5 (2/3)
–
11
Italy
53
53
52.8
Cervical
Metastatic, non-Hodgkin lymphomas
Histopathologic
NA
NA
ES
4 (2/3)
–
6
[14]
China
82
155
56.1
Cervical/ supraclavicular region/ armpit/groin
Thyroid cancer, breast cancer, Hashimoto’s disease combined with adenoma, lung cancer, nasopharyngeal carcinoma, lymphoma, primitive neuroectodermal tumors
Histopathologic/ imaging
38.2
NA
ES/SR
4 (2/3)
2.395
8
Bhatia et al., 2010
[7]
China
74
74
50.0
Cervical
Lung carcinoma, breast adenocarcinoma, nasopharyngeal carcinoma, thyroid papillary carcinoma, squamous carcinoma, uterine mixed mullerian tumor, ovarian adenocarcinoma
Histopathologic
50
41.9
ES
4 (2/3)
–
14
Shi et al., 2010
[13]
China
37
85
62.4
Cervical
Lung cancer, thyroid carcinoma, breast cancer, gastric cancer
Histopathologic
55
67.6
ES
5 (2/3)
–
13
Tan, 2010
[15]
China
107
128
54.7
Cervical
Non-Hodgkin’s lymphoma, leukemia, Hodgkin’s lymphoma, squamous cell carcinoma, melanoma, adenocarcinoma
Histopathologic
53.4
52.3
ES/SR
4 (2/3)
1.5
13
Choi et al., 2011
[16]
Korea
62
64
48.4
Axillary
Breast cancer
Histopathologic
53
0
ES/SR
4 (2/3)
2.3
13
Taylor et al., 2011
[17]
UK
50
50
42.0
Axillary
Breast cancer
Histopathologic
57
0
ES/SR
4 (2/3)
0.65
13
Ref.
Country
No. of patients
Lyshchik et al., 2007
[5]
Japan
43
Alam et al., 2008
[11]
Japan
Rubaltelli et al., 2009
[12]
Zhang et al., 2009
No. of LN
ES: elasticity score; LNs: lymph nodes; NA: not available; QUADAS: quality assessment of diagnostic accuracy studies. RTE: real-time elastography; SR: strain ratio.
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
Malignant LN rate (%)
Author, year
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Unclear Yes Unclear Unclear Yes Yes Yes Yes Yes No Yes Yes
Yes Yes Yes Yes
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Yes
Yes Yes Yes Unclear Unclear Yes Yes Yes No No Unclear Yes Yes
No
Yes Unclear No Unclear Unclear No Yes Yes Yes Yes Unclear No Unclear
Yes
Yes Yes Yes Yes Yes Yes Yes Yes No No Yes Yes Yes
No
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Intermediate results Clinical review bias Diagnostic review bias Test review bias Reference execution details Test execution details Incorporation bias Selection criteria Spectrum composition
2579
Lyshchik et al., 2007 Alam et al., 2008 Rubaltelli et al., 2009 Zhang et al., 2009 Bhatia et al., 2010 Shi et al., 2010 Tan, 2010 Choi et al., 2011 Taylor et al., 2011
The Fagan plot demonstrated that ES measurement was very informative lowering the negative post-probability of malignant LNs to as low as 9% when “negative” measurement from 25% preprobability; however, only 71% probability of correctly diagnosed malignant LNs following a “positive” measurement (Fig. 5A). When
Author, year
3.4. Fagan plot analysis
Table 2 Quality assessment of included studies.
3.3. Publication bias According to Deeks’ funnel plot asymmetry test, there were no publication bias among ES studies (p = 0.74, Supplementary Fig. 1A) and RS studies (p = 0.35, Supplementary Fig. 1B).
Differential verification bias
Q2 Q1
Q3
Q4
Q5
Q6
Q7
Q8
Q9
Q10
The summary sensitivity and specificity of the ES measurement for the differentiation of benign and malignant LNs were 0.74 (95% CI, 0.66–0.81) and 0.90 (95% CI, 0.82–0.94), respectively (Fig. 2A). The summary DOR was 24.84 (95% CI, 11.59–53.24, Fig. 3A), and the HSROC was 0.88 (95% CI, 0.85–0.91) (Fig. 4A). Based on these values, and assuming a 62.7% malignant LNs (as observed in the included studies), the estimated PPV and NPV were 0.76 (95% CI, 0.72–0.80) and 0.86 (95% CI, 0.81–0.91), respectively. Sensitivity analysis was performed after excluding two studies in which patients were not all referred to histopathologic findings [11,14]. The results were not influenced excessively after omitting the two studies. There was statistically significant heterogeneity in DOR (2 p < 0.001, I2 = 98.38%). However, according to the meta-regression analysis, ES measurement accuracy was not affected by the covariates. The summary sensitivity of the SR measurement was 0.88 (95% CI, 0.79–0.93) (Fig. 2B), which was significantly higher than it was for ES measurement (p < 0.001). The summary specificity of the SR measurement was 0.81 (95% CI, 0.49–0.95) (Fig. 2B), which was lower than that of ES measurement (p < 0.001). The summary DOR was 31.63 (95% CI, 9.95–100.56) (Fig. 3B), and the HSROC was 0.91 (95% CI, 0.88–0.93) for the SR measurement (Fig. 4B). The accuracy of the SR measurement was similar to that of ES measurement (p = 0.573). Based on these values, and assuming a 49.6% malignant LNs (as observed in the included studies), the estimated PPV and NPV were 0.86 (95% CI, 0.74–0.97) and 0.81 (95% CI, 0.70–0.92), respectively. Sensitivity analysis was performed after excluding one study in which patients who were not all referred to histopathologic findings [14]. The results were not influenced excessively after omitting the study. There was statistically significant heterogeneity in DOR (2 p < 0.001, I2 = 99.99%). However, according to the meta-regression analysis, SR measurement accuracy was not affected by the covariates.
Partial verification bias
Q11
3.2. Accuracy of RTE for the differentiation of benign and malignant LNs
Appropriate Disease reference progression standard bias
Q12
Q13
Q14
Table 1 summarized the main characteristics of the included studies. A total of 835 LNs (395 benign, 440 malignant) in 545 patients (median age, 51.5 years) were evaluated. Some variation of the classifications of RTE for LNs existed in the studies. Four studies performed ES measurement [7,11–13], and 5 studies performed both ES and SR measurement [5,14–17]. ES measurement, 7 described the measurement on a numerical scale of 1–4, 2 described a scale of 1–5. Histopathologic findings or imaging findings were used as the reference standard for the final classification of LNs in two studies [11,14]. Seven studies included patients who were only referred to histopathologic findings. According to the QUADAS scale, included studies had a very good methodological quality except two studies [12,14] (Table 2).
Withdrawals
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
2580
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
Fig. 2. Forest sensitivity and specificity. (A) ES for LNs and (B) SR for LNs. CI: confidence interval; ES: elasticity score; LNs: lymph nodes; SR: strain ratio.
there is a high pre-test index of suspicion (pre-test probability = 50% or 75%), then the probability of a correct diagnosis following a “positive” measurement reached or exceeded 90% for malignant LNs, but the diagnosis would be wrong in 22–46% of patients with a “negative” measurement (Fig. 5B and C).
SR measurement was very informative with 82% probability of correctly diagnosing malignant LNs following a “positive” measurement when pre-test probability was 50% and lowering the probability of disease to as low as 13% when “negative” measurement (Fig. 6B). However, when pre-test probability was 25%,
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
2581
Fig. 3. Forest diagnostic odds ratio. (A) ES for LNs and (B) SR for LNs. CI: confidence interval; ES: elasticity score; LNs: lymph nodes; SR: strain ratio.
Fig. 4. HSROC curve of the RTE for classification of LNs. (A) ES for LNs and (B) SR for LNs. The size of the dots for 1-specificity and sensitivity of the single studies in the ROC space was derived from the respective sample size. ES: elasticity score; HSROC: hierarchical summary receiver operating characteristic; LNs: lymph nodes; SR: strain ratio.
2582
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
Fig. 5. Fagan plot analysis to evaluate the clinical utility of ES for classification of LNs. (A) Pre-test probability = 25%; (B) Pre-test probability = 50% and (C) Pre-test probability = 75%. The Fagan plot consisted of a vertical axis on the left with the pre-test probability, an axis in the middle representing the likelihood ratio and a vertical axis on the right representing the post-test probability. ES: elasticity score; LNs: lymph nodes; NLR: negative likelihood ratio; PLR: positive likelihood ratio.
Fig. 6. Fagan plot analysis to evaluate the clinical utility of SR for classification of LNs. (A) Pre-test probability = 25%; (B) Pre-test probability = 50% and (C) Pre-test probability = 75%. The Fagan plot consisted of a vertical axis on the left with the pre-test probability, an axis in the middle representing the likelihood ratio and a vertical axis on the right representing the post-test probability. LNs: lymph nodes; NLR: negative likelihood ratio; PLR: positive likelihood ratio; SR: strain ratio.
SR measurement had only 61% probability of correctly diagnosing malignant LNs following a “positive” measurement (Fig. 6A). In addition, the diagnosis would be wrong in 31% of patients with a “negative” measurement when pre-test probability was 75% (Fig. 6C).
4. Discussion In this meta-analysis, we evaluated the performance of RTE, a non-invasive technique, for the classification of superficial LNs. The results indicated that RTE had a high accuracy for identification of
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
malignant LNs. The HSROC for identification of malignant LNs by RTE was 0.88 (95% CI, 0.85–0.91) for ES measurement; SR measurement also had a high HSROC (0.91 (95% CI, 0.88–0.93)). In this respect, RTE is promising and worthy to translate into clinical practice that it is a reliable and noninvasive procedure and is able to reveal LNs with small size or deep location. Therefore, RTE could be integrated in the classification of superficial LNs. RTE might also be useful to select suspicious LNs for biopsy. In addition, in patients with cancer, RTE could be used for cancer staging by identifying LNs most likely to be malignant [5]. The ease of availability and application, non-invasive nature of RTE are set to make it increasingly popular among radiologist. Although, ES measurement had a good diagnostic accuracy, it depended on the examiner resulting in significant inter-observer variability. Therefore, a quantitative method of analysis for elasticity images of LNs is needed. The SR measurement is a quantitative approach that could relatively objectively evaluate the LNs stiffness. Although the accuracy of SR measurement was similar to ES measurement, it obviously improved the diagnostic sensitivity value. A strength of our study was that the Fagan plot analysis had been used for exploring clinical utilities of the RTE. Our results show that SR measurement could be used to classify the superficial LNs (when pre-test probability = 50%), with 82% probability of correctly diagnosing LNs following a “positive” measurement. A “negative” measurement was also informative, as malignant LNs were present in only 13% of patients. However, results were less promising for ES measurement. Although the probability of correctly identification of malignant LNs could reach 88% following a “positive” measurement, the diagnosis would be wrong in 22% of patients with a “negative” measurement. The above results, especially the following a “positive” SR measurement, have been considered encouraging in individual studies and in some reviews. However, the major drawback was that the SR cut-off value was different across studies and has not been validated [15,16]. Pooling such “optimal” results from these studies might artificially increase the summary sensitivity and specificity. Because of a scarcity of publications, we could not address these important issues. The efficacy of SR measurement in the classification of superficial LNs should be further evaluated in more studies in the future. Because significant heterogeneity was present in this analysis, our study explored factors that may be the source of heterogeneity by meta-regression analysis. Although 11 specific covariates of patient and study were examined, none was found that could affect ES measurement accuracy or SR measurement accuracy. We could not explain this finding. Specifically, LNs position, which might influence the stiffness value, was not significant in the metaregression analysis. Despite its potential, there were some technique limitations of RTE. The RTE technique is labor intensive and time consuming, requiring additional time for off-line computer processing. In addition, some artifacts would be caused by no standardized compression load applied with freehand RTE and motion of surrounding tissues and vessels during compression scanning [5,18]. Therefore, investigators in future studies need to analyze inter and intra-observer variability. Some limitations of this study should be taken into consideration: first, in our meta-analysis, the results were pooled for LNs with different positions. The different depth of the LNs could influence the stiffness value. The perception of the pressure and frequency of some deep LNs such as LNs in the supraclavicular or auxiliary areas will be affected because of signal attenuating [15]. We would encourage investigators to be rigorous in their patient selection in future studies. In addition, significant heterogeneity was also
2583
present in the evaluation of ES measurement and SR measurement accuracy in these studies, and therefore interpretation should be cautious. In conclusion, our meta-analysis shows that RTE could be used as a good identification tool for malignant LNs, with an exceeded 80% disease probability following a “positive” measurement. A “negative” measurement is also accurate and informative for SR measurement, with disease being present in only 13% of patients, but not for ES measurement. It is necessary to further evaluate the potential role of RTE in a large, prospective, international, multicenter study. Conflicts of interest No conflicts of interest exist for all authors. Acknowledgement The authors thank Ke-Qing Shi, MD, Department of Infection and Liver Diseases, Liver Research Center, the First Affiliated Hospital of Wenzhou Medical College, Wenzhou 325000, China, for his valuable statistical assistance. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.ejrad.2011.10.026. References [1] Lecuru F, Mathevet P, Querleu D, et al. Bilateral negative sentinel nodes accurately predict absence of lymph node metastasis in early cervical cancer: results of the SENTICOL study. J Clin Oncol 2011;29(13):1686–91. [2] Choi HJ, Ju W, Myung SK, et al. Diagnostic performance of computer tomography, magnetic resonance imaging, and positron emission tomography or positron emission tomography/computer tomography for detection of metastatic lymph nodes in patients with cervical cancer: meta-analysis. Cancer Sci 2010;101(6):1471–9. [3] Gritzmann N, Hollerweger A, Macheiner P, et al. Sonography of soft tissue masses of the neck. J Clin Ultrasound 2002;30(6):356–73. [4] Tan SM, Teh HS, Mancer JF, et al. Improving B mode ultrasound evaluation of breast lesions with real-time ultrasound elastography: a clinical approach. Breast 2008;17(3):252–7. [5] Lyshchik A, Higashi T, Asato R, et al. Cervical lymph node metastases: diagnosis at sonoelastography: initial experience. Radiology 2007;243(1): 258–67. [6] Landoni V, Francione V, Marzi S, et al. Quantitative analysis of elastography images in the detection of breast cancer. Eur J Radiol 2011, doi:10.1016/j.ejrad.2011.04.012. [7] Bhatia KS, Cho CC, Yuen YH, et al. Real-time qualitative ultrasound elastography of cervical lymph nodes in routine clinical practice: interobserver agreement and correlation with malignancy. Ultrasound Med Biol 2010;36(12):1990–7. [8] Whiting P, Rutjes AW, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003;3:25. [9] Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol 2005;58(9):882–93. [10] Hellmich M, Lehmacher W. A ruler for interpreting diagnostic test results. Methods Inf Med 2005;44(1):124–6. [11] Alam F, Naito K, Horiguchi J, et al. Accuracy of sonographic elastography in the differential diagnosis of enlarged cervical lymph nodes: comparison with conventional B-mode sonography. AJR Am J Roentgenol 2008;191(2): 604–10. [12] Rubaltelli L, Stramare R, Tregnaghi A, et al. The role of sonoelastography in the differential diagnosis of neck nodules. J Ultrasound 2009;12(3):93–100. [13] Shi GH, Wang XM, Ou GC, et al. Comparative study of ultrasonic elastography with conventional ultrasonography in cervical lymph nodes [in Chinese]. Chin J Ultrasound Med 2010;26(8):730–3. [14] Zhang YR, Lv Q, Yin YH, et al. The value of ultrasound elastography in differential diagnosis of superficial lymph nodes. Front Med China 2009;3(3):368–74. [15] Tan R, Xiao Y, He Q. Ultrasound elastography: its potential role in assessment of cervical lymphadenopathy. Acad Radiol 2010;17(7):849–55. [16] Choi JJ, Kang BJ, Kim SH, et al. Role of sonographic elastography in the differential diagnosis of axillary lymph nodes in breast cancer. J Ultrasound Med 2011;30(4):429–36.
2584
L. Ying et al. / European Journal of Radiology 81 (2012) 2576–2584
[17] Taylor K, O‘Keeffe S, Britton PD, et al. Ultrasound elastography as an adjuvant to conventional ultrasound in the preoperative assessment of axillary lymph nodes in suspected breast cancer: a pilot study. Clin Radiol 2011;66(11):1064–71.
[18] Havre RF, Elde E, Gilja OH, et al. Freehand real-time elastography: impact of scanning parameters on image quality and in vitro intra- and interobserver validations. Ultrasound Med Biol 2008;34(10):1638–50.