Original Research n Musculoskeletal
Imaging
Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.
Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?1 Fang Zhao, MD Shivani Ahlawat, MD Sahar J. Farahani, MBBS Kristy L. Weber, MD Elizabeth A. Montgomery, MD John A. Carrino, MD, MPH Laura M. Fayad, MD
1
From the Shandong Medical Imaging Research Institute, Shandong University, Jinan, Shandong, China (F.Z.); and Russell H. Morgan Department of Radiology and Radiological Science, The Johns Hopkins Medical Institutions, 601 N Caroline St, Baltimore, MD 21287 (S.A., S.J.F., K.L.W., E.A.M., J.A.C., L.M.F.). Received August 19, 2013; revision requested September 27; revision received December 17; accepted January 10, 2014; final version accepted January 31. Address correspondence to L.M.F. (e-mail: lfayad1@ jhmi.edu).
Purpose:
To identify the magnetic resonance (MR) imaging features that can be used to differentiate high-grade from lowgrade soft-tissue sarcoma (STS).
Materials and Methods:
Institutional review board approval was obtained, and informed consent was waived. Patients with STS who had undergone MR imaging with T1-weighted, T2-weighted, and contrast material–enhanced sequences prior to neoadjuvant therapy and surgery were included retrospectively. Tumor grade (grades 1–3) was recorded from the histologic specimen for each STS. Images were evaluated by two observers for tumor size and MR features (signal intensity, heterogeneity, margin, and perilesional characteristics) on images obtained with each sequence. Descriptive statistics for low-grade (grade 1) and high-grade (grades 2 and 3) STS were recorded, and the accuracy of individual features was determined. A multivariate logistic regression model was developed to identify features that were independently predictive of a high-grade tumor.
Results:
Ninety-five patients (48 female [mean age, 55.8 years; age range, 7–96 years] and 47 male [mean age, 55.3 years; age range, 1–87 years]) with STS (16 patients with grade 1 STS, 34 patients with grade 2 STS, and 45 patients with grade 3 STS) were included. High-grade STS differed from low-grade STS in size (.5 cm, P = .004), tumor margin (partly or poorly defined margin on T1-weighted images, P = .002; with other sequences, P , .001), internal signal intensity composition (heterogeneous signal intensity on T2-weighted images, P = .009), and peritumoral characteristics (peritumoral high signal intensity on T2-weighted images, P = .025; peritumoral enhancement on contrastenhanced T1-weighted images, P , .001). The logistic regression model showed that peritumoral contrast enhancement is the strongest independent indicator of highgrade status (odds ratio, 13.6; 95% confidence interval: 2.9, 64.6).
Conclusion:
Among several MR imaging features that aid in the discrimination of high-grade from low-grade sarcomas, the presence of peritumoral contrast enhancement is a feature that may be solely used to diagnose high-grade STS. RSNA, 2014
q
RSNA, 2014
q
192
radiology.rsna.org n Radiology: Volume 272: Number 1—July 2014
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
S
oft-tissue sarcoma (STS) includes a heterogeneous group of malignant neoplasms that often carry a high mortality rate (1). Various prognostic factors have been evaluated for the purpose of treatment planning and forecasting outcome, with histopathologic grade being one of the most important factors in predicting the prognosis of patients with STS (2–5). Neoadjuvant chemotherapy is usually included in the treatment plans
Advances in Knowledge nn High-grade soft-tissue sarcoma (STS) differed significantly from low-grade STS in size (high-grade STS tumors were more commonly at least 5 cm, P = .004), internal signal intensity characteristics (high-grade tumors were more heterogeneous at T2-weighted imaging, P = .009), tumor margin (high-grade tumors were more likely to have partly or poorly defined margins with all nonenhanced and contrast-enhanced T1-weighted sequences, P .002), peritumoral high signal intensity on T2-weighted images (high-grade tumors were more likely to have peritumoral high signal intensity, P = .025), and peritumoral contrast enhancement (high-grade tumors exhibited peritumoral enhancement more frequently, P , .001). nn In deep STS, differences in tumor size, margin, and perilesional characteristics remained significant, while in superficial STS, high- and low-grade tumor features were more similar. nn Intravenous contrast material administration for MR imaging provides useful information for characterization of the grade of STS.
for high-grade STS, while it is not for low-grade tumors (6,7). Hence, establishing the tumor grade accurately is critical when deciding on subsequent therapeutic action. To plan treatment and decide on chemotherapy preoperatively, the standard of care is based on the assignment of tumor grade by means of percutaneous biopsy (8). However, tumor grade may occasionally be nondiagnostic or mischaracterized in a percutaneous sample (8–10) because of an insufficient specimen or sampling error, which could lead to less appropriate treatment and a potentially poorer prognosis. Although not a frequent problem, in 66 STS cases studied by Yang et al (8), five had a change in grade between percutaneous biopsy and final histologic examination after surgical resection. Magnetic resonance (MR) imaging is a well-established tool for the staging of STS, primarily for the determination of tumor extent (11–13). The relationship between MR features and pathologic findings has been studied for specific types of tumors. However, the correlation of MR imaging features and pathologic grade of STS has been assessed to a very limited degree (14,15), with reports that peritumoral signal intensity changes and peripheral growth pattern are related to tumor grade (14). The hypothesis of the current study was that MR features exist that can be used to predict a highgrade STS. Hence, the purpose of this study was to explore an assortment of MR imaging features of STS and identify features that allow differentiation of highgrade STS from low-grade STS.
Materials and Methods Overview Institutional review board approval was obtained, and informed consent
nn Peritumoral contrast material enhancement on contrastenhanced T1-weighted MR images can be used as a strong independent predictor for characterization of STS as high grade, with an odds ratio of 13.6 (95% confidence interval: 2.9, 64.6). Radiology: Volume 272: Number 1—July 2014 n radiology.rsna.org
Implication for Patient Care nn Since treatment decisions for STS depend heavily on the assignment of tumor grade, MR imaging features that can be used to predict a high tumor grade can serve to supplement biopsy results.
Zhao et al
was waived for this retrospective study. No industry support was given for this project. MR images acquired in 95 patients (prior to any treatment) with subsequent histologically proven STS were reviewed by two observers who recorded the MR features of each tumor. The accuracy (sensitivity and specificity) of each feature was determined in the differentiation of highgrade from low-grade histologic findings. A multivariate regression model was developed that incorporated the most sensitive features to uncover the imaging characteristics that were most predictive of high-grade STS.
Patients Consecutive patient records were obtained from an institutional review board–approved database generated for patients treated in an orthopedic oncology clinic between January 2010 and January 2013. Inclusion criteria were (a) patients with STS who had undergone MR imaging prior to biopsy, neoadjuvant therapy, and surgery, whose MR images were available for interpretation, and (b) patients for whom a definitive pathologic grade could be determined from the resection specimen. Exclusion criteria were patients with inadequate pathologic findings for determination of tumor grade, patients without available MR images acquired prior to preoperative neoadjuvant therapy, and patients with image quality
Published online before print 10.1148/radiol.14131871 Content codes: Radiology 2014; 272:192–201 Abbreviation: STS = soft-tissue sarcoma Author contributions: Guarantors of integrity of entire study, S.A., L.M.F.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, F.Z., S.J.F., E.A.M., L.M.F.; clinical studies, F.Z., S.J.F., E.A.M., L.M.F.; experimental studies, S.J.F., L.M.F.; statistical analysis, S.J.F., J.A.C., L.M.F.; and manuscript editing, all authors Conflicts of interest are listed at the end of this article.
193
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
deemed to be nondiagnostic after a review of the images.
Histologic Analysis The histologic diagnosis for each tumor was assigned by a pathologist (E.A.M.) with 22 years of experience in STS analysis and was based on pathologic findings in surgical specimens from our institution in 91 of 95 patients (96%) and on pathologic findings in histologic samples obtained from outside institutions but reviewed in our institution in four of 95 patients (4%). Lesions were graded according to the French Federation of Cancer Centers Sarcoma Group system (grades 1–3) (16). MR Imaging Protocol The MR imaging protocol included nonenhanced and contrast material–enhanced MR sequences and was slightly variable because of varied locations within the body and minor adjustments to the parameters. All patients underwent MR imaging with T1-weighted (repetition time msec/echo time msec, 450–716/8–14) and fluid-sensitive (T2-weighted 2750–4583/58–126) sequences or short tau inversion recovery (5610/39; inversion time msec, 150), with 94 of 95 patients undergoing imaging with fat-suppressed T2-weighted or short tau inversion-recovery sequences and one undergoing imaging with T2-weighted sequences without fat suppression. Eighty-two of 95 patients underwent imaging with contrast-enhanced T1-weighted sequences (spin echo 425–676/9–13); of these, 75 of 82 underwent fat-suppressed contrast-enhanced T1-weighted imaging, and seven of 82 underwent contrastenhanced T1-weighted imaging without fat suppression. MR Image Analysis All images were reviewed by two experienced radiologists in consensus (L.M.F. and S.A., with 1 and 11 years of postresidency musculoskeletal imaging experience, respectively). First, the quality of images obtained with each MR imaging sequence (T1weighted, T2-weighted, and contrastenhanced T1-weighted) was recorded 194
(nondiagnostic, nondiagnostic with 25%–50% artifacts, diagnostic with ,25% artifacts, or diagnostic with no substantial artifacts). Second, lesion location (neck, chest, abdomen, pelvis, thigh, calf, foot, arm, forearm, or hand), tissue layer (superficial or subcutaneous, deep intramuscular, deep intermuscular, or deep mixed [tumor partially subcutaneous and partially intramuscular]), and lesion size (greatest linear dimension of tumor, categorized as ,5 cm or 5 cm) were recorded. Next, the following features were observed on the nonenhanced images (T1-weighted and T2-weighted): signal intensity (hypointense, isointense, or hyperintense in comparison to muscle), signal intensity heterogeneity (homogeneous, ,25% heterogeneous, 25%– 75% heterogeneous, or .75% heterogeneous), and tumor margin (poorly defined [.75% of margin not clear], mixed margin [10%–25% of margin not clear], or well defined [.90% of margin clear]). The presence or absence of a peritumoral capsule sign (17,18), internal low-signal-intensity septations on T2-weighted images, peritumoral high signal intensity on T2-weighted images, a peritumoral fat capsule sign, and periosteal reaction, as well as cortex, marrow, and joint extension, were recorded. Finally, on the contrast-enhanced images, the presence or absence of neurovascular encasement, vascular occlusion, tumor enhancement, and peritumoral enhancement were recorded; the tumor margin (same categories as those for T1- and T2-weighted images) and the percentage of contrast enhancement and/or heterogeneity in the tumor (,25% enhancement, 25%–50% enhancement, 50%–75% enhancement, or .75% enhancement) were observed.
Statistical Analysis Statistical analysis was performed by using STATA/SE 10.0 software (StataCorp, College Station, Tex). First, descriptive statistical analyses were performed. In our study, grade 1 tumors were considered low-grade STS, and grades 2 and 3 were considered highgrade STS (19). Nonparametric tests
Zhao et al
were used for categorical variables. Differences in age between male and female patients were analyzed by using the Wilcoxon rank sum test; the x2 test was used to examine any significant difference in the MR features of highand low-grade sarcomas, and summary measures of sensitivity and specificity were generated. The Pearson correlation coefficient was used to examine the correlation between MR imaging features. Second, the accuracy of each individual MR feature in the differentiation of low- and high-grade tumors was determined. For each MR feature, a category of the feature associated with aggressive behavior was selected (for example, the presence of neurovascular encasement, rather than its absence, or a poorly defined or mixed-definition margin rather than a well-defined margin was chosen to test for accuracy of the MR feature). Finally, multivariate logistic regression modeling was developed to incorporate available imaging features and patient characteristics (age, patient sex) to identify features that were independently predictive of a high-grade tumor. The variables that demonstrated a significant association with high-grade status were entered into the model as a forward stepwise method. The final model was selected on the basis of the variables with P values less than .05 or the ones that improved the model according to the likelihood ratio. The odds ratio was used as a measure of the relative magnitude of an association between predictor variables and high-grade status.
Results Of 156 consecutively retrieved patient records, 95 patients with STS met the inclusion criteria. Patients were excluded for having inadequate pathologic findings (n = 9), absence of available MR images prior to treatment (n = 46), and nondiagnostic image quality (n = 6). Overall mean age of the patients was 55.5 years, with an age range of 1–95 years; there was no significant difference in the ages of female and male patients (P = .89) (48 females [mean age, 55.8 years; age range, 7–96 years]
radiology.rsna.org n Radiology: Volume 272: Number 1—July 2014
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Zhao et al
Table 1 Descriptive Statistics and Distribution of Features according to Tumor Grade Feature Tissue layer (n = 95) Subcutaneous Intramuscular Intermuscular Mixed Size (n = 95) ,5 cm 5 cm Signal intensity on T1-weighted images (n = 95) Isointense Hypointense Hyperintense Signal intensity heterogeneity on T1-weighted images (n = 95) Homogeneous ,25% heterogeneous 25%–75% heterogeneous .75% heterogeneous Margin on T1-weighted images (n = 95) Poorly defined Mixed definition Well defined Peritumoral capsule sign (n = 95) Signal intensity on T2-weighted images (n = 94)‡ Isointense Hypointense Hyperintense Signal intensity heterogeneity on T2-weighted images (n = 94) Homogeneity ,25% heterogeneity 25%–75% heterogeneity .75% heterogeneity Margin on T2-weighted images (n = 94) Poorly defined Mixed definition Well defined Internal low-signal-intensity septations (n = 94) Peritumoral high signal intensity on T2-weighted images (n = 94) Peritumoral fat capsule sign (n = 94) Periosteal reaction (n = 94) Cortex extension (n = 94) Marrow extension (n = 94) Joint extension (n = 94) Neurovascular encasement (n = 94) Vascular occlusion (n = 92)§ Margin on contrast-enhanced T1-weighted images (n = 82)|| Poorly defined Mixed definition Well defined
Grade 1
Grade 2
Grade 3
Low Grade (%)
High Grade (%)*
5 8 1 2
11 15 3 5
12 22 8 3
31 (5/16) 69 (11/16) 69 (11/16) 69 (11/16)
29 (23/79) 71 (56/79) 71 (56/79) 71 (56/79)
9 7
8 26
8 37
56 (9/16) 44 (7/16)
20 (16/79) 80 (63/79)
12 0 4
18 3 13
23 1 21
75 (12/16) 25 (4/16) 25 (4/16)
52 (41/79) 48 (38/79) 48 (38/79)
6 4 4 2
8 10 13 3
14 11 8 12
38 (6/16) 62 (10/16) 62 (10/16) 62 (10/16)
28 (22/79) 72 (57/79) 72 (57/79) 72 (57/79)
4 1 11 9
6 20 8 17
4 27 14 25
31 (5/16) 31 (5/16) 69 (11/16) 56 (9/16)
72 (57/79) 72 (57/79) 28 (22/79) 53 (42/79)
1 0 14
3 1 30
0 1 44
7 (1/15) 93 (14/15) 93 (14/15)
4 (3/79) 96 (76/79) 96 (76/79)
4 2 2 7
3 6 12 13
1 7 19 28
27 (4/15) 73 (11/15) 73 (11/15) 73 (11/15)
5 (4/79) 95 (75/79) 95 (75/79) 95 (75/79)
2 2 11 10 12 6 0 0 0 0 1 1
1 23 10 26 33 5 5 0 0 2 4 0
3 33 9 35 43 19 8 5 3 1 9 1
27 (4/15) 27 (4/15) 73 (11/15) 67 (10/15) 80 (12/15) 40 (6/15) 0 0 0 0 7 (1/15) 7 (1/15)
76 (60/79) 76 (60/79) 24 (19/79) 77 (61/79) 96 (76/79) 30 (24/79) 16 (13/79) 6 (5/79) 4 (3/79) 4 (3/79) 16 (13/79) 1 (1/77)
1 1 12
2 19 10
2 27 8
14 (2/14) 14 (2/14) 86 (12/14)
74 (50/68) 74 (50/68) 26 (18/68)
P Value† .30
.004
.09
.44
.002
.82 .65
.009
,.001
.40 .025 .51 .08 .30 .42 .43 .29 .19 ,.001
Table 1 (continues)
Radiology: Volume 272: Number 1—July 2014 n radiology.rsna.org
195
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Zhao et al
Table 1 (continued) Descriptive Statistics and Distribution of Features according to Tumor Grade Feature
Grade 1
Enhancement percentage (n = 82) ,25% 25%–50% 55%–75% .75% Peritumoral enhancement (n = 82)
Grade 2
Grade 3
Low Grade (%)
High Grade (%)*
1 10 13 7 27
5 16 10 6 35
14 (2/14) 86 (12/14) 86 (12/14) 86 (12/14) 43 (6/14)
10 (7/68) 90 (61/68) 90 (61/68) 90 (61/68) 91 (62/68)
P Value† .53
2 3 2 7 6
,.001
Note.—Data are numbers of patients, unless indicated otherwise. Numbers in parentheses are raw data. * High grade is defined as grades 2 and 3 combined. †
Low grade versus high grade.
‡
One set of T2-weighted images was excluded for nondiagnostic quality.
§
Three patients were excluded because of the inability to assess vascular occlusion, owing to lack of contrast-enhanced T1-weighted images.
||
Eighty-two of 95 patients had contrast-enhanced T1-weighted images.
and 47 males [mean age, 55.3 years; age range, 1–87 years]). There were 16 patients with grade 1 STS, 34 patients with grade 2 STS, and 45 patients with grade 3 STS. The histologic types of STS were undifferentiated pleomorphic sarcoma and/or malignant fibrous histiocytoma (n = 16), liposarcoma (n = 15, five low grade), pleomorphic sarcoma (n = 11), myxofibrosarcoma (n = 9), synovial sarcoma (n = 9), fibrosarcoma (n = 5, two low grade), malignant peripheral nerve sheath tumor (n = 5), leiomyosarcoma (n = 5, one low grade), rhabdomyosarcoma (n = 4), fibromyxoid sarcoma (n = 3, all low grade), spindle cell sarcoma (n = 3), myxoid sarcoma (n = 2), myxoinflammatory fibroblastic sarcoma (n = 2, all low grade), hemangioendothelioma (n = 2, all low grade), epithelioid angiosarcoma (n = 2), pleomorphic hyalinizing angiectatic tumor (n = 1, low grade), and Ewing sarcoma (n = 1). The quality of the MR images was rated as diagnostic in most cases: For T1-weighted images, quality was diagnostic in 95 of 95 patients (89 with no substantial artifacts, six with ,25% artifacts). For T2-weighted imaging, images acquired in 94 of 95 patients were diagnostic (88 with no substantial artifacts, six with ,25% artifacts, and one nondiagnostic case with 25%–50% artifacts [which was subsequently excluded]). For contrast-enhanced T1weighted imaging, images in 82 of 82 patients were diagnostic (76 with no 196
substantial artifacts, six with ,25% artifacts). The locations of the STS were in the neck (n = 1), chest (n = 6), abdomen (n = 1), pelvis (n = 6), thigh (n = 40), calf (n = 22), foot (n = 6), arm (n = 8), and forearm (n = 5). Table 1 shows the descriptive statistics for tumor size and MR features of the study population, detailed by tumor grade. High-grade STS differed significantly from low-grade STS in size (high-grade STS was more commonly 5 cm, P , .01), internal signal intensity characteristics (high-grade tumors were more heterogeneous on T2-weighted images, P , .01), tumor margin (high-grade tumors were more likely to have partly or poorly defined margins on all nonenhanced and contrast-enhanced T1-weighted images, P , .01), peritumoral high signal intensity on T2-weighted images (high-grade tumors were more likely to have peritumoral high signal intensity, P , .05) and peritumoral contrast enhancement (high-grade tumors exhibited peritumoral enhancement more frequently, P , .01) (Fig 1). High- and low-grade STS were found in both deep (Figs 1, 2) and superficial (subcutaneous) (Fig 3) locations. In deep STS, differences in tumor size, margin, and perilesional characteristics remained significant, with high-grade tumors being larger (5 cm) and having a partly or poorly defined margin and peritumoral enhancement on
contrast-enhanced T1-weighted images compared with low-grade tumors (Figs 1, 2). In superficial STS, however, highand low-grade tumor features were more similar, and differences in tumor size and margin in particular were not maintained; only perilesional characteristics (peritumoral enhancement on contrast-enhanced T1-weighted images) were more common in high- than in low-grade superficial STS (P = .05) (Fig 3). Table 2 displays the accuracy (sensitivity and specificity) of each individual MR feature for differentiation of highand low-grade STS. A poorly or partly defined tumor margin offered the highest combination of sensitivity and specificity on all nonenhanced and contrastenhanced images for characterization of a sarcoma as high grade (72% and 69%, 76% and 73%, and 74% and 86% for T1-weighted, T2-weighted, and contrast-enhanced T1-weighted images, respectively). Although rarely present, invasion into adjacent structures (periosteum, cortex, marrow, and neurovascular structures) was a specific feature for characterization of a tumor as high grade. Of the factors studied in the logistic regression model, peritumoral enhancement on contrast-enhanced T1-weighted images was the strongest independent indicator of a high-grade sarcoma (odds ratio of 13.6; 95% confidence interval: 2.9, 64.6), with a
radiology.rsna.org n Radiology: Volume 272: Number 1—July 2014
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Zhao et al
Figure 1
Figure 1: Images in a 65-year-old man with grade 3 deep intramuscular sarcoma (undifferentiated pleomorphic sarcoma). The MR images show typical features of a high-grade tumor, including poor margin definition on images obtained with all sequences, internal signal intensity heterogeneity, perilesional signal intensity abnormalities, and large size. (a) Axial T1-weighted MR image (466/16) shows a large 10.5 3 3.4 3 5.8-cm heterogeneously hyperintense mass with indistinct margins replacing the lateral gastrocnemius (arrow). (b) Axial T2-weighted MR image (fat suppressed 3380/60) shows internal signal intensity heterogeneity, with indistinct margins and perilesional edema (arrows) in the posterior calf. (c) Axial contrast-enhanced T1-weighted image (fat suppressed 475/12) shows heterogeneous enhancement within the mass (long arrow) with perilesional muscular, fascial, and subcutaneous enhancement (short arrows). (d) In this microscopic image (hematoxylin-eosin stain; original magnification, 320), the tumor is seen infiltrating subcutaneous adipose tissue. There is a cuff of lymphoplasmacytic inflammation at the interface of the pleomorphic malignant cells and the adipose tissue, possibly accounting for the perilesional edema and enhancement. (e) In this microscopic image (hematoxylin-eosin stain; original magnification, 320), the malignant pleomorphic cells are seen infiltrating skeletal muscle in a ragged pattern.
sensitivity and specificity of 91% and 57%, respectively, while accounting for other features (tumor size, tissue layer, tumor margin on images obtained with all sequences, signal heterogeneity with other sequences, peritumoral high signal intensity on T2-weighted images, age, and patient sex) in the model. (Table 3 shows the results of the final logistic regression model.) The tumor margin definition and peritumoral
contrast enhancement were correlated; however, there was no interaction between these variables in the logistic regression model. Figure 3 emphasizes the importance of MR imaging characteristics in the assignment of grade, as it is an example of a sarcoma for which there was a change in tumor grade between the percutaneous biopsy results (which showed a low-grade tumor) and the final surgical specimen after
Radiology: Volume 272: Number 1—July 2014 n radiology.rsna.org
resection (which showed a high-grade tumor); peritumoral enhancement was a clue to the presence of a high-grade sarcoma in this case.
Discussion For a patient with STS, the histologic tumor grade is the most critical piece of information needed for treatment planning, as it is heavily tied to the patient’s 197
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Zhao et al
Figure 2
Figure 2: Images in a 48-year-old woman with a grade 1 deep intramuscular myxoid liposarcoma. The MR images depict typical features of a low-grade tumor, including good margin definition with all imaging sequences, relatively homogeneous internal signal intensity, lack of perilesional signal intensity on T2-weighted and contrast-enhanced T1-weighted images, and small size. (a) Axial T1-weighted MR image (755/8.2) shows a homogeneous mass that is relatively isointense compared to skeletal muscle located in the medial soleus (arrow). (b) Axial T2-weighted MR image (fat suppressed 2836/124) shows the mass to be relatively homogeneously hyperintense, with no perilesional edema (arrow). The mass has low-signal-intensity internal septations. (c) Axial contrast-enhanced T1-weighted MR image (250/3.9) shows only mildly heterogeneous internal enhancement, without perilesional enhancement (arrow). (d) In this low-power microscopic image (hematoxylineosin stain; original magnification, 34), note the sharply demarcated tumor margin, with pseudocapsule. The upper portion of the field shows skeletal muscle with mixed adipose tissue, a result of peritumoral muscle atrophy. (e) In the high-power microscopic image (hematoxylin-eosin stain; original magnification, 360), the cellularity is monotonous and uniform, and nuclei are evenly spaced. Note the lipoblast in the center of the field, showing lipid droplets that crisply indent the nucleus.
risk of metastasis and overall survival. The pathologic grade of a sarcoma is the parameter used for selecting patients for whom preoperative chemotherapy should be considered (6,7), as high-grade STS may be treated with neoadjuvant chemotherapy, while lowgrade STS is not. To our knowledge, investigators in only one study have attempted to describe differences in MR features of low- and high-grade sarcomas with nonenhanced T1-weighted and T2-weighted imaging. Our study advances prior work by demonstrating a comprehensive assessment of the accuracy of all available MR features of 198
STS (on nonenhanced and contrast-enhanced images) for prediction of a highgrade tumor and developing a model to identify peritumoral contrast enhancement as the strongest independent predictor of a high-grade STS. At presentation, a patient with a soft-tissue mass typically undergoes MR imaging for assessment of tumor extent, although MR imaging characteristics have been described for various histologic soft-tissue tumor types, and some features have been described that can be used to distinguish benign and malignant soft-tissue tumors. When features are rendered indeterminate, a
biopsy of the mass is performed. Subsequently, when a sarcoma is identified by means of biopsy results, the determination of histologic grade is the next step in deciding on neoadjuvant treatment. Because biopsy results are occasionally erroneous with regard to tumor grade (8–10), imaging features that aid in the prediction of a high-grade state can be used as a supplement for biopsy results (when inconclusive or discordant with the MR features) and subsequent treatment. Few prior investigators have identified the differentiating features of lowand high-grade sarcomas. Liu et al (14)
radiology.rsna.org n Radiology: Volume 272: Number 1—July 2014
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Figure 3
Figure 3: MR images in a 65-year-old woman with grade 3 sarcoma. Percutaneous biopsy of this mass initially indicated a grade 1 leiomyosarcoma; however, tumor grade was changed after surgical resection to be a grade 3 leiomyosarcoma. (a) Coronal fat-suppressed T2-weighted MR image (3100/60) shows moderate perilesional hyperintensity along the proximal tumor margin (arrow). (b) Coronal contrast-enhanced T1-weighted image (fat suppressed 560/10) shows mildly heterogeneous enhancement within the mass and minimal perilesional enhancement (arrow). With the results of the current study, after percutaneous biopsy, a high-grade diagnosis could have been considered in this case in which a grade 1 sarcoma was diagnosed, given the perilesional signal intensity changes that are more common with high-grade tumors. At the time of diagnosis, review of the MR images did not prospectively indicate the presence of a high-grade lesion or necessitate repeat biopsy with a different target.
studied only nonenhanced T1-weighted and T2-weighted imaging in the investigation of 59 sarcomas and showed that the histologic grade of STS was significantly related to the definition of tumor margin and peripheral growth pattern, with high-grade tumors more commonly having a poorly defined margin, while low-grade tumors had a well-defined margin on nonenhanced images; peritumoral fluid signal intensity was more common in high-grade tumors, and a low peritumoral signal intensity capsule sign was more common in low-grade
tumors. Our study confirms the importance of tumor margin in predicting grade: Generally, a poorly or partly defined tumor margin on images obtained with any sequence indicates that tumor cells have infiltrated the surrounding tissues and shows the invasive, aggressive nature of the tumor. Indeed, the peripheral tumor growth pattern in sarcomas has been demonstrated to be an important prognostic factor for the development of local recurrence and metastasis (5,20). Fernebro et al studied the peritumoral signal intensity changes
Radiology: Volume 272: Number 1—July 2014 n radiology.rsna.org
Zhao et al
of sarcomas at MR imaging and correlated them with histologic results, confirming that MR imaging provides information about the peripheral growth important to prognosis and the risk of metastasis, although the relationship of this feature to tumor grade was not assessed directly (15). As for a poorly defined margin, our study confirmed that peritumoral fluid signal intensity was significantly more common in highthan in low-grade STS. However, unlike in the study of Liu et al (14), the presence of a capsule sign was not significantly different in low- and high-grade tumors, perhaps because of the larger number of patients investigated in our study population. Intravenous contrast material administration is routinely used for the evaluation of STS, and the present study offers new insights into sarcoma imaging features after contrast enhancement. Most significantly, a partly or poorly defined tumor margin and peritumoral enhancement on contrast-enhanced images were sensitive features associated with highgrade STS, with the multivariate regression model demonstrating peritumoral enhancement to be the strongest independent feature for prediction of highgrade STS. The latter findings corroborate prior work that showed malignant tumors to have increased vascularity at the periphery (21) and reflect the more aggressive peripheral growth pattern of high-grade tumors. Though the signal intensity heterogeneity on T2-weighted images was significantly different between high-grade and low-grade STS, the degree of contrast enhancement within a tumor was not different, with both types of tumors being equally likely to enhance heterogeneously. The latter observation was unexpected, as highergrade tumors usually have increased necrosis (16,22). Tumor depth has also been found to be an important predictor of prognosis and the development of distant metastases (2,3,23). In the present study, both low- and high-grade sarcomas occurred superficially, as well as in the deep tissue layers. When observed independently, we found that the MR imaging features of high- and 199
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Table 2 Sensitivity and Specificity of Each MR Imaging Feature for Prediction of High-Grade Sarcoma Feature Deep tissue layer Size 5 cm Hypointensity or hyperintensity on T1-weighted images Signal intensity heterogeneity on T1-weighted images Poorly defined or mixed-definition margin on T1-weighted images Absence of peritumoral capsule sign Hypointensity or hyperintensity on T2-weighted images Signal intensity heterogeneity on T2-weighted images Poorly defined and mixed-definition margin on T2-weighted images Internal low-signal-intensity septations Peritumoral high signal intensity on T2-weighted images Absence of peritumoral fat cap Periosteal reaction Cortex extension Marrow extension Joint extension Neurovascular encasement Vascular occlusion Poorly defined and mixed-definition margin on contrast-enhanced T1-weighted images Enhancement and/or heterogeneity . 25% Peritumoral contrast enhancement†
Sensitivity (%)*
Specificity (%)*
71 (56/79) 80 (63/79) 48 (38/79) 72 (57/79) 72 (57/79) 47 (37/79) 96 (76/79) 95 (75/79) 76 (60/79) 77 (61/79) 96 (76/79) 70 (55/79) 16 (13/79) 6 (5/79) 4 (3/79) 4 (3/79) 16 (13/79) 1 (1/77) 74 (50/68)
31 (5/16) 56 (9/16) 75 (12/16) 38 (6/16) 69 (11/16) 56 (9/16) 7 (1/15) 27 (4/15) 73 (11/15) 33 (5/15) 20 (3/15) 40 (6/15) 100 (15/15) 100 (15/15) 100 (15/15) 100 (15/15) 93 (14/15) 93 (14/15) 86 (12/14)
90 (61/68) 91 (62/68)
14 (2/14) 57 (8/14)
Note.—Numbers in parentheses are raw data. * The denominators are variable because contrast-enhanced images were only available in 82 of 95 patients, and images acquired with fluid-sensitive sequences with fat-suppressed T2-weighted imaging or short inversion time inversion-recovery imaging were available in 94 of 95 patients. For assessment of vascular occlusion, images in two patients were deemed inadequate for optimal occlusion assessment. †
Peritumoral contrast enhancement is an independent predictor of high-grade status, with an odds ratio of 13.6 (95% confidence interval: 2.9, 64.6).
Table 3 Results of Final Multivariate Regression Model Variable Peritumoral contrast enhancement Poorly defined or mixed-definition margin on T1-weighted images Poorly defined and mixed-definition margin on contrast-enhanced T1-weighted images Hypointensity or hyperintensity on T1-weighted images Absence of peritumoral fat cap
low-grade superficial STS were similar, including such features as the definition of tumor margin and tumor size. One possible explanation for the similarity in low- and high-grade superficial STS may lie with the fact that subcutaneous 200
Odds Ratio
95% Confidence Interval
P Value
13.6 4.4
2.9, 64.6 0.9, 21.7
.001 ..05
5.9
0.7, 49.1
..05
4.3 0.6
0.1, 22.6 0.1, 22.4
..05 ..05
tumors can be detected sooner than deep lesions (24), and, hence, highgrade subcutaneous tumors can be detected earlier than those found in the deep tissues, and the MR features may not be as developed as those of deep
Zhao et al
high-grade sarcomas. In addition, highgrade subcutaneous sarcomas reportedly have a better prognosis than their deeper counterparts (25,26). Tumor size has been established as one of the important factors related to clinical outcome and is included in prognostic systems in many studies, with a traditional threshold of 5 cm (27,28): According to the literature, tumors at least 5 cm in diameter have lower overall survival and a higher risk of recurrence and distant metastases (3,5,29). Hence, as expected, high-grade tumors were more likely to be larger than 5 cm in our study, while low-grade tumors were smaller than 5 cm. Our study has limitations. As a retrospective study, the MR imaging protocol, while containing T1-weighted, T2-weighted, and contrast-enhanced T1-weighted imaging, was not uniform, with some sequences having fat suppression and others none. Second, there were relatively few low-grade cases of STS as compared with high-grade tumors, but this proportion accurately reflects the population distribution of low- and high-grade STS in the community (30). Third, all histologic types were assessed communally; there may be differentiating MR features among the different histologic subtypes of STS that confound the results of this study, but a subgroup analysis was not possible with the number of patients that had specific histologic types in this study. In addition, since the premise of our study was the characterization of grade in patients with known STS, these study results cannot be generalized to the population of patients with an indeterminate soft-tissue mass (benign or malignant); benign soft-tissue masses with aggressive features, such as myositis ossificans, may share MR imaging features that mimic high-grade sarcomas, but our study was not designed to investigate the distinguishing imaging features of benign and malignant soft-tissue masses. In conclusion, while tumor size, tumor margin, heterogeneous signal intensity on T2-weighted images, and peritumoral high signal intensity on T2-weighted images can be used to
radiology.rsna.org n Radiology: Volume 272: Number 1—July 2014
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
differentiate high- and low-grade STS, the presence of peritumoral contrast enhancement is a feature that may be solely used to diagnose high-grade STS. Along with percutaneous biopsy results, MR imaging features provide an additional tool for determining tumor grade. Disclosures of Conflicts of Interest: F.Z. No relevant conflicts of interest to disclose. S.A. No relevant conflicts of interest to disclose. S.J.F. No relevant conflicts of interest to disclose. K.L.W. No relevant conflicts of interest to disclose. E.A.M. No relevant conflicts of interest to disclose. J.A.C. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: author received grants from Siemens, Carestream, and Toshiba and received personal fees from Siemens, BioClinica, Pfizer, Medtronic, and General Electric for consulting and/or lectures. Other relationships: none to disclose. L.M.F. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: author received grants from Siemens Medical Systems and GERRAF for MR spectroscopy. Other relationships: none to disclose.
References 1. Kolovich GG, Wooldridge AN, Christy JM, Crist MK, Mayerson JL, Scharschmidt TJ. A retrospective statistical analysis of highgrade soft tissue sarcomas. Med Oncol 2012;29(2):1335–1344. 2. Pisters PW, Leung DH, Woodruff J, Shi W, Brennan MF. Analysis of prognostic factors in 1,041 patients with localized soft tissue sarcomas of the extremities. J Clin Oncol 1996;14(5):1679–1689. 3. Stefanovski PD, Bidoli E, De Paoli A, et al. Prognostic factors in soft tissue sarcomas: a study of 395 patients. Eur J Surg Oncol 2002;28(2):153–164.
7. Menendez LR, Ahlmann ER, Savage K, Cluck M, Fedenko AN. Tumor necrosis has no prognostic value in neoadjuvant chemotherapy for soft tissue sarcoma. Clin Orthop Relat Res 2007;455(455):219–224. 8. Yang J, Frassica FJ, Fayad L, Clark DP, Weber KL. Analysis of nondiagnostic results after image-guided needle biopsies of musculoskeletal lesions. Clin Orthop Relat Res 2010;468(11):3103–3111. 9. Strauss DC, Qureshi YA, Hayes AJ, Thway K, Fisher C, Thomas JM. The role of core needle biopsy in the diagnosis of suspected soft tissue tumours. J Surg Oncol 2010;102(5):523–529. 10. Mitsuyoshi G, Naito N, Kawai A, et al. Accurate diagnosis of musculoskeletal lesions by core needle biopsy. J Surg Oncol 2006;94(1):21–27. 11. Totty WG, Murphy WA, Lee JK. Softtissue tumors: MR imaging. Radiology 1986;160(1):135–141. 12. Walker EA, Salesky JS, Fenton ME, Mur phey MD. Magnetic resonance imaging of malignant soft tissue neoplasms in the adult. Radiol Clin North Am 2011;49(6):1219– 1234, vi. 13. Kransdorf MJ, Jelinek JS, Moser RP Jr, et al. Soft-tissue masses: diagnosis using MR imaging. AJR Am J Roentgenol 1989;153(3):541–547. 14. Liu QY, Li HG, Chen JY, Liang BL. Correlation of MRI features to histopathologic grade of soft tissue sarcoma [in Chinese]. Ai Zheng 2008;27(8):856–860. 15. Fernebro J, Wiklund M, Jonsson K, et al. Focus on the tumour periphery in MRI evaluation of soft tissue sarcoma: infiltrative growth signifies poor prognosis. Sarcoma 2006;2006:21251.
4. Zagars GK, Ballo MT, Pisters PW, et al. Prognostic factors for patients with localized soft-tissue sarcoma treated with conservation surgery and radiation therapy: an analysis of 1225 patients. Cancer 2003;97(10):2530–2543.
16. Guillou L, Coindre JM, Bonichon F, et al. Comparative study of the National Cancer Institute and French Federation of Cancer Centers Sarcoma Group grading systems in a population of 410 adult patients with soft tissue sarcoma. J Clin Oncol 1997;15(1):350–362.
5. Mandard AM, Petiot JF, Marnay J, et al. Prognostic factors in soft tissue sarcomas. A multivariate analysis of 109 cases. Cancer 1989;63(7):1437–1451.
17. Rydholm A. Surgical margins for soft tis sue sarcoma. Acta Orthop Scand Suppl 1997;273:81–85.
6. Lucas DR, Kshirsagar MP, Biermann JS, et al. Histologic alterations from neoadjuvant chemotherapy in high-grade extremity soft tissue sarcoma: clinicopathological correlation. Oncologist 2008;13(4):451–458.
18. Takahashi M, Sato K, Miura T. MR imaging of musculoskeletal sarcomas: the clinical significance of peritumoral low signal intensity lines in planning surgical margins. Nippon Seikeigeka Gakkai Zasshi 1993;67(10): 881–896.
Radiology: Volume 272: Number 1—July 2014 n radiology.rsna.org
Zhao et al
19. Reynoso D, Subbiah V, Trent JC, et al. Neoadjuvant treatment of soft-tissue sarcoma: a multimodality approach. J Surg Oncol 2010;101(4):327–333. 20. Engellau J, Bendahl PO, Persson A, et al. Improved prognostication in soft tissue sarcoma: independent information from vascular invasion, necrosis, growth pattern, and immunostaining using whole-tumor sections and tissue microarrays. Hum Pathol 2005;36(9):994–1002. 21. Ma LD, Frassica FJ, McCarthy EF, Bluemke DA, Zerhouni EA. Benign and malignant musculoskeletal masses: MR imaging differentiation with rim-to-center differential enhancement ratios. Radiology 1997;202(3):739–744. 22. Coindre JM, Trojani M, Contesso G, et al. Reproducibility of a histopathologic grading system for adult soft tissue sarcoma. Cancer 1986;58(2):306–309. 23. Gaynor JJ, Tan CC, Casper ES, et al. Refinement of clinicopathologic staging for localized soft tissue sarcoma of the extremity: a study of 423 adults. J Clin Oncol 1992;10(8):1317–1329. 24. Sawamura C, Matsumoto S, Shimoji T, Tanizawa T, Ae K. What are risk factors for local recurrence of deep high-grade softtissue sarcomas? Clin Orthop Relat Res 2012;470(3):700–705. 25. Collin C, Godbold J, Hajdu S, Brennan M. Localized extremity soft tissue sarcoma: an analysis of factors affecting survival. J Clin Oncol 1987;5(4):601–612. 26. Peabody TD, Monson D, Montag A, Schell MJ, Finn H, Simon MA. A comparison of the prognoses for deep and subcutaneous sarcomas of the extremities. J Bone Joint Surg Am 1994;76(8):1167–1173. 27. Rööser B, Attewell R, Berg NO, Rydholm A. Prognostication in soft tissue sarcoma. A model with four risk factors. Cancer 1988;61(4):817–823. 28. Levine EA. Prognostic factors in soft tissue sarcoma. Semin Surg Oncol 1999;17(1): 23–32. 29. Saddegh MK, Lindholm J, Lundberg A, Nilsonne U, Kreicbergs A. Staging of softtissue sarcomas. Prognostic analysis of clinical and pathological features. J Bone Joint Surg Br 1992;74(4):495–500. 30. Carneiro A, Bendahl PO, Engellau J, et al. A prognostic model for soft tissue sarcoma of the extremities and trunk wall based on size, vascular invasion, necrosis, and growth pattern. Cancer 2011;117(6):1279–1287.
201
Imaging
Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.
Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?1 Fang Zhao, MD Shivani Ahlawat, MD Sahar J. Farahani, MBBS Kristy L. Weber, MD Elizabeth A. Montgomery, MD John A. Carrino, MD, MPH Laura M. Fayad, MD
1
From the Shandong Medical Imaging Research Institute, Shandong University, Jinan, Shandong, China (F.Z.); and Russell H. Morgan Department of Radiology and Radiological Science, The Johns Hopkins Medical Institutions, 601 N Caroline St, Baltimore, MD 21287 (S.A., S.J.F., K.L.W., E.A.M., J.A.C., L.M.F.). Received August 19, 2013; revision requested September 27; revision received December 17; accepted January 10, 2014; final version accepted January 31. Address correspondence to L.M.F. (e-mail: lfayad1@ jhmi.edu).
Purpose:
To identify the magnetic resonance (MR) imaging features that can be used to differentiate high-grade from lowgrade soft-tissue sarcoma (STS).
Materials and Methods:
Institutional review board approval was obtained, and informed consent was waived. Patients with STS who had undergone MR imaging with T1-weighted, T2-weighted, and contrast material–enhanced sequences prior to neoadjuvant therapy and surgery were included retrospectively. Tumor grade (grades 1–3) was recorded from the histologic specimen for each STS. Images were evaluated by two observers for tumor size and MR features (signal intensity, heterogeneity, margin, and perilesional characteristics) on images obtained with each sequence. Descriptive statistics for low-grade (grade 1) and high-grade (grades 2 and 3) STS were recorded, and the accuracy of individual features was determined. A multivariate logistic regression model was developed to identify features that were independently predictive of a high-grade tumor.
Results:
Ninety-five patients (48 female [mean age, 55.8 years; age range, 7–96 years] and 47 male [mean age, 55.3 years; age range, 1–87 years]) with STS (16 patients with grade 1 STS, 34 patients with grade 2 STS, and 45 patients with grade 3 STS) were included. High-grade STS differed from low-grade STS in size (.5 cm, P = .004), tumor margin (partly or poorly defined margin on T1-weighted images, P = .002; with other sequences, P , .001), internal signal intensity composition (heterogeneous signal intensity on T2-weighted images, P = .009), and peritumoral characteristics (peritumoral high signal intensity on T2-weighted images, P = .025; peritumoral enhancement on contrastenhanced T1-weighted images, P , .001). The logistic regression model showed that peritumoral contrast enhancement is the strongest independent indicator of highgrade status (odds ratio, 13.6; 95% confidence interval: 2.9, 64.6).
Conclusion:
Among several MR imaging features that aid in the discrimination of high-grade from low-grade sarcomas, the presence of peritumoral contrast enhancement is a feature that may be solely used to diagnose high-grade STS. RSNA, 2014
q
RSNA, 2014
q
192
radiology.rsna.org n Radiology: Volume 272: Number 1—July 2014
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
S
oft-tissue sarcoma (STS) includes a heterogeneous group of malignant neoplasms that often carry a high mortality rate (1). Various prognostic factors have been evaluated for the purpose of treatment planning and forecasting outcome, with histopathologic grade being one of the most important factors in predicting the prognosis of patients with STS (2–5). Neoadjuvant chemotherapy is usually included in the treatment plans
Advances in Knowledge nn High-grade soft-tissue sarcoma (STS) differed significantly from low-grade STS in size (high-grade STS tumors were more commonly at least 5 cm, P = .004), internal signal intensity characteristics (high-grade tumors were more heterogeneous at T2-weighted imaging, P = .009), tumor margin (high-grade tumors were more likely to have partly or poorly defined margins with all nonenhanced and contrast-enhanced T1-weighted sequences, P .002), peritumoral high signal intensity on T2-weighted images (high-grade tumors were more likely to have peritumoral high signal intensity, P = .025), and peritumoral contrast enhancement (high-grade tumors exhibited peritumoral enhancement more frequently, P , .001). nn In deep STS, differences in tumor size, margin, and perilesional characteristics remained significant, while in superficial STS, high- and low-grade tumor features were more similar. nn Intravenous contrast material administration for MR imaging provides useful information for characterization of the grade of STS.
for high-grade STS, while it is not for low-grade tumors (6,7). Hence, establishing the tumor grade accurately is critical when deciding on subsequent therapeutic action. To plan treatment and decide on chemotherapy preoperatively, the standard of care is based on the assignment of tumor grade by means of percutaneous biopsy (8). However, tumor grade may occasionally be nondiagnostic or mischaracterized in a percutaneous sample (8–10) because of an insufficient specimen or sampling error, which could lead to less appropriate treatment and a potentially poorer prognosis. Although not a frequent problem, in 66 STS cases studied by Yang et al (8), five had a change in grade between percutaneous biopsy and final histologic examination after surgical resection. Magnetic resonance (MR) imaging is a well-established tool for the staging of STS, primarily for the determination of tumor extent (11–13). The relationship between MR features and pathologic findings has been studied for specific types of tumors. However, the correlation of MR imaging features and pathologic grade of STS has been assessed to a very limited degree (14,15), with reports that peritumoral signal intensity changes and peripheral growth pattern are related to tumor grade (14). The hypothesis of the current study was that MR features exist that can be used to predict a highgrade STS. Hence, the purpose of this study was to explore an assortment of MR imaging features of STS and identify features that allow differentiation of highgrade STS from low-grade STS.
Materials and Methods Overview Institutional review board approval was obtained, and informed consent
nn Peritumoral contrast material enhancement on contrastenhanced T1-weighted MR images can be used as a strong independent predictor for characterization of STS as high grade, with an odds ratio of 13.6 (95% confidence interval: 2.9, 64.6). Radiology: Volume 272: Number 1—July 2014 n radiology.rsna.org
Implication for Patient Care nn Since treatment decisions for STS depend heavily on the assignment of tumor grade, MR imaging features that can be used to predict a high tumor grade can serve to supplement biopsy results.
Zhao et al
was waived for this retrospective study. No industry support was given for this project. MR images acquired in 95 patients (prior to any treatment) with subsequent histologically proven STS were reviewed by two observers who recorded the MR features of each tumor. The accuracy (sensitivity and specificity) of each feature was determined in the differentiation of highgrade from low-grade histologic findings. A multivariate regression model was developed that incorporated the most sensitive features to uncover the imaging characteristics that were most predictive of high-grade STS.
Patients Consecutive patient records were obtained from an institutional review board–approved database generated for patients treated in an orthopedic oncology clinic between January 2010 and January 2013. Inclusion criteria were (a) patients with STS who had undergone MR imaging prior to biopsy, neoadjuvant therapy, and surgery, whose MR images were available for interpretation, and (b) patients for whom a definitive pathologic grade could be determined from the resection specimen. Exclusion criteria were patients with inadequate pathologic findings for determination of tumor grade, patients without available MR images acquired prior to preoperative neoadjuvant therapy, and patients with image quality
Published online before print 10.1148/radiol.14131871 Content codes: Radiology 2014; 272:192–201 Abbreviation: STS = soft-tissue sarcoma Author contributions: Guarantors of integrity of entire study, S.A., L.M.F.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, F.Z., S.J.F., E.A.M., L.M.F.; clinical studies, F.Z., S.J.F., E.A.M., L.M.F.; experimental studies, S.J.F., L.M.F.; statistical analysis, S.J.F., J.A.C., L.M.F.; and manuscript editing, all authors Conflicts of interest are listed at the end of this article.
193
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
deemed to be nondiagnostic after a review of the images.
Histologic Analysis The histologic diagnosis for each tumor was assigned by a pathologist (E.A.M.) with 22 years of experience in STS analysis and was based on pathologic findings in surgical specimens from our institution in 91 of 95 patients (96%) and on pathologic findings in histologic samples obtained from outside institutions but reviewed in our institution in four of 95 patients (4%). Lesions were graded according to the French Federation of Cancer Centers Sarcoma Group system (grades 1–3) (16). MR Imaging Protocol The MR imaging protocol included nonenhanced and contrast material–enhanced MR sequences and was slightly variable because of varied locations within the body and minor adjustments to the parameters. All patients underwent MR imaging with T1-weighted (repetition time msec/echo time msec, 450–716/8–14) and fluid-sensitive (T2-weighted 2750–4583/58–126) sequences or short tau inversion recovery (5610/39; inversion time msec, 150), with 94 of 95 patients undergoing imaging with fat-suppressed T2-weighted or short tau inversion-recovery sequences and one undergoing imaging with T2-weighted sequences without fat suppression. Eighty-two of 95 patients underwent imaging with contrast-enhanced T1-weighted sequences (spin echo 425–676/9–13); of these, 75 of 82 underwent fat-suppressed contrast-enhanced T1-weighted imaging, and seven of 82 underwent contrastenhanced T1-weighted imaging without fat suppression. MR Image Analysis All images were reviewed by two experienced radiologists in consensus (L.M.F. and S.A., with 1 and 11 years of postresidency musculoskeletal imaging experience, respectively). First, the quality of images obtained with each MR imaging sequence (T1weighted, T2-weighted, and contrastenhanced T1-weighted) was recorded 194
(nondiagnostic, nondiagnostic with 25%–50% artifacts, diagnostic with ,25% artifacts, or diagnostic with no substantial artifacts). Second, lesion location (neck, chest, abdomen, pelvis, thigh, calf, foot, arm, forearm, or hand), tissue layer (superficial or subcutaneous, deep intramuscular, deep intermuscular, or deep mixed [tumor partially subcutaneous and partially intramuscular]), and lesion size (greatest linear dimension of tumor, categorized as ,5 cm or 5 cm) were recorded. Next, the following features were observed on the nonenhanced images (T1-weighted and T2-weighted): signal intensity (hypointense, isointense, or hyperintense in comparison to muscle), signal intensity heterogeneity (homogeneous, ,25% heterogeneous, 25%– 75% heterogeneous, or .75% heterogeneous), and tumor margin (poorly defined [.75% of margin not clear], mixed margin [10%–25% of margin not clear], or well defined [.90% of margin clear]). The presence or absence of a peritumoral capsule sign (17,18), internal low-signal-intensity septations on T2-weighted images, peritumoral high signal intensity on T2-weighted images, a peritumoral fat capsule sign, and periosteal reaction, as well as cortex, marrow, and joint extension, were recorded. Finally, on the contrast-enhanced images, the presence or absence of neurovascular encasement, vascular occlusion, tumor enhancement, and peritumoral enhancement were recorded; the tumor margin (same categories as those for T1- and T2-weighted images) and the percentage of contrast enhancement and/or heterogeneity in the tumor (,25% enhancement, 25%–50% enhancement, 50%–75% enhancement, or .75% enhancement) were observed.
Statistical Analysis Statistical analysis was performed by using STATA/SE 10.0 software (StataCorp, College Station, Tex). First, descriptive statistical analyses were performed. In our study, grade 1 tumors were considered low-grade STS, and grades 2 and 3 were considered highgrade STS (19). Nonparametric tests
Zhao et al
were used for categorical variables. Differences in age between male and female patients were analyzed by using the Wilcoxon rank sum test; the x2 test was used to examine any significant difference in the MR features of highand low-grade sarcomas, and summary measures of sensitivity and specificity were generated. The Pearson correlation coefficient was used to examine the correlation between MR imaging features. Second, the accuracy of each individual MR feature in the differentiation of low- and high-grade tumors was determined. For each MR feature, a category of the feature associated with aggressive behavior was selected (for example, the presence of neurovascular encasement, rather than its absence, or a poorly defined or mixed-definition margin rather than a well-defined margin was chosen to test for accuracy of the MR feature). Finally, multivariate logistic regression modeling was developed to incorporate available imaging features and patient characteristics (age, patient sex) to identify features that were independently predictive of a high-grade tumor. The variables that demonstrated a significant association with high-grade status were entered into the model as a forward stepwise method. The final model was selected on the basis of the variables with P values less than .05 or the ones that improved the model according to the likelihood ratio. The odds ratio was used as a measure of the relative magnitude of an association between predictor variables and high-grade status.
Results Of 156 consecutively retrieved patient records, 95 patients with STS met the inclusion criteria. Patients were excluded for having inadequate pathologic findings (n = 9), absence of available MR images prior to treatment (n = 46), and nondiagnostic image quality (n = 6). Overall mean age of the patients was 55.5 years, with an age range of 1–95 years; there was no significant difference in the ages of female and male patients (P = .89) (48 females [mean age, 55.8 years; age range, 7–96 years]
radiology.rsna.org n Radiology: Volume 272: Number 1—July 2014
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Zhao et al
Table 1 Descriptive Statistics and Distribution of Features according to Tumor Grade Feature Tissue layer (n = 95) Subcutaneous Intramuscular Intermuscular Mixed Size (n = 95) ,5 cm 5 cm Signal intensity on T1-weighted images (n = 95) Isointense Hypointense Hyperintense Signal intensity heterogeneity on T1-weighted images (n = 95) Homogeneous ,25% heterogeneous 25%–75% heterogeneous .75% heterogeneous Margin on T1-weighted images (n = 95) Poorly defined Mixed definition Well defined Peritumoral capsule sign (n = 95) Signal intensity on T2-weighted images (n = 94)‡ Isointense Hypointense Hyperintense Signal intensity heterogeneity on T2-weighted images (n = 94) Homogeneity ,25% heterogeneity 25%–75% heterogeneity .75% heterogeneity Margin on T2-weighted images (n = 94) Poorly defined Mixed definition Well defined Internal low-signal-intensity septations (n = 94) Peritumoral high signal intensity on T2-weighted images (n = 94) Peritumoral fat capsule sign (n = 94) Periosteal reaction (n = 94) Cortex extension (n = 94) Marrow extension (n = 94) Joint extension (n = 94) Neurovascular encasement (n = 94) Vascular occlusion (n = 92)§ Margin on contrast-enhanced T1-weighted images (n = 82)|| Poorly defined Mixed definition Well defined
Grade 1
Grade 2
Grade 3
Low Grade (%)
High Grade (%)*
5 8 1 2
11 15 3 5
12 22 8 3
31 (5/16) 69 (11/16) 69 (11/16) 69 (11/16)
29 (23/79) 71 (56/79) 71 (56/79) 71 (56/79)
9 7
8 26
8 37
56 (9/16) 44 (7/16)
20 (16/79) 80 (63/79)
12 0 4
18 3 13
23 1 21
75 (12/16) 25 (4/16) 25 (4/16)
52 (41/79) 48 (38/79) 48 (38/79)
6 4 4 2
8 10 13 3
14 11 8 12
38 (6/16) 62 (10/16) 62 (10/16) 62 (10/16)
28 (22/79) 72 (57/79) 72 (57/79) 72 (57/79)
4 1 11 9
6 20 8 17
4 27 14 25
31 (5/16) 31 (5/16) 69 (11/16) 56 (9/16)
72 (57/79) 72 (57/79) 28 (22/79) 53 (42/79)
1 0 14
3 1 30
0 1 44
7 (1/15) 93 (14/15) 93 (14/15)
4 (3/79) 96 (76/79) 96 (76/79)
4 2 2 7
3 6 12 13
1 7 19 28
27 (4/15) 73 (11/15) 73 (11/15) 73 (11/15)
5 (4/79) 95 (75/79) 95 (75/79) 95 (75/79)
2 2 11 10 12 6 0 0 0 0 1 1
1 23 10 26 33 5 5 0 0 2 4 0
3 33 9 35 43 19 8 5 3 1 9 1
27 (4/15) 27 (4/15) 73 (11/15) 67 (10/15) 80 (12/15) 40 (6/15) 0 0 0 0 7 (1/15) 7 (1/15)
76 (60/79) 76 (60/79) 24 (19/79) 77 (61/79) 96 (76/79) 30 (24/79) 16 (13/79) 6 (5/79) 4 (3/79) 4 (3/79) 16 (13/79) 1 (1/77)
1 1 12
2 19 10
2 27 8
14 (2/14) 14 (2/14) 86 (12/14)
74 (50/68) 74 (50/68) 26 (18/68)
P Value† .30
.004
.09
.44
.002
.82 .65
.009
,.001
.40 .025 .51 .08 .30 .42 .43 .29 .19 ,.001
Table 1 (continues)
Radiology: Volume 272: Number 1—July 2014 n radiology.rsna.org
195
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Zhao et al
Table 1 (continued) Descriptive Statistics and Distribution of Features according to Tumor Grade Feature
Grade 1
Enhancement percentage (n = 82) ,25% 25%–50% 55%–75% .75% Peritumoral enhancement (n = 82)
Grade 2
Grade 3
Low Grade (%)
High Grade (%)*
1 10 13 7 27
5 16 10 6 35
14 (2/14) 86 (12/14) 86 (12/14) 86 (12/14) 43 (6/14)
10 (7/68) 90 (61/68) 90 (61/68) 90 (61/68) 91 (62/68)
P Value† .53
2 3 2 7 6
,.001
Note.—Data are numbers of patients, unless indicated otherwise. Numbers in parentheses are raw data. * High grade is defined as grades 2 and 3 combined. †
Low grade versus high grade.
‡
One set of T2-weighted images was excluded for nondiagnostic quality.
§
Three patients were excluded because of the inability to assess vascular occlusion, owing to lack of contrast-enhanced T1-weighted images.
||
Eighty-two of 95 patients had contrast-enhanced T1-weighted images.
and 47 males [mean age, 55.3 years; age range, 1–87 years]). There were 16 patients with grade 1 STS, 34 patients with grade 2 STS, and 45 patients with grade 3 STS. The histologic types of STS were undifferentiated pleomorphic sarcoma and/or malignant fibrous histiocytoma (n = 16), liposarcoma (n = 15, five low grade), pleomorphic sarcoma (n = 11), myxofibrosarcoma (n = 9), synovial sarcoma (n = 9), fibrosarcoma (n = 5, two low grade), malignant peripheral nerve sheath tumor (n = 5), leiomyosarcoma (n = 5, one low grade), rhabdomyosarcoma (n = 4), fibromyxoid sarcoma (n = 3, all low grade), spindle cell sarcoma (n = 3), myxoid sarcoma (n = 2), myxoinflammatory fibroblastic sarcoma (n = 2, all low grade), hemangioendothelioma (n = 2, all low grade), epithelioid angiosarcoma (n = 2), pleomorphic hyalinizing angiectatic tumor (n = 1, low grade), and Ewing sarcoma (n = 1). The quality of the MR images was rated as diagnostic in most cases: For T1-weighted images, quality was diagnostic in 95 of 95 patients (89 with no substantial artifacts, six with ,25% artifacts). For T2-weighted imaging, images acquired in 94 of 95 patients were diagnostic (88 with no substantial artifacts, six with ,25% artifacts, and one nondiagnostic case with 25%–50% artifacts [which was subsequently excluded]). For contrast-enhanced T1weighted imaging, images in 82 of 82 patients were diagnostic (76 with no 196
substantial artifacts, six with ,25% artifacts). The locations of the STS were in the neck (n = 1), chest (n = 6), abdomen (n = 1), pelvis (n = 6), thigh (n = 40), calf (n = 22), foot (n = 6), arm (n = 8), and forearm (n = 5). Table 1 shows the descriptive statistics for tumor size and MR features of the study population, detailed by tumor grade. High-grade STS differed significantly from low-grade STS in size (high-grade STS was more commonly 5 cm, P , .01), internal signal intensity characteristics (high-grade tumors were more heterogeneous on T2-weighted images, P , .01), tumor margin (high-grade tumors were more likely to have partly or poorly defined margins on all nonenhanced and contrast-enhanced T1-weighted images, P , .01), peritumoral high signal intensity on T2-weighted images (high-grade tumors were more likely to have peritumoral high signal intensity, P , .05) and peritumoral contrast enhancement (high-grade tumors exhibited peritumoral enhancement more frequently, P , .01) (Fig 1). High- and low-grade STS were found in both deep (Figs 1, 2) and superficial (subcutaneous) (Fig 3) locations. In deep STS, differences in tumor size, margin, and perilesional characteristics remained significant, with high-grade tumors being larger (5 cm) and having a partly or poorly defined margin and peritumoral enhancement on
contrast-enhanced T1-weighted images compared with low-grade tumors (Figs 1, 2). In superficial STS, however, highand low-grade tumor features were more similar, and differences in tumor size and margin in particular were not maintained; only perilesional characteristics (peritumoral enhancement on contrast-enhanced T1-weighted images) were more common in high- than in low-grade superficial STS (P = .05) (Fig 3). Table 2 displays the accuracy (sensitivity and specificity) of each individual MR feature for differentiation of highand low-grade STS. A poorly or partly defined tumor margin offered the highest combination of sensitivity and specificity on all nonenhanced and contrastenhanced images for characterization of a sarcoma as high grade (72% and 69%, 76% and 73%, and 74% and 86% for T1-weighted, T2-weighted, and contrast-enhanced T1-weighted images, respectively). Although rarely present, invasion into adjacent structures (periosteum, cortex, marrow, and neurovascular structures) was a specific feature for characterization of a tumor as high grade. Of the factors studied in the logistic regression model, peritumoral enhancement on contrast-enhanced T1-weighted images was the strongest independent indicator of a high-grade sarcoma (odds ratio of 13.6; 95% confidence interval: 2.9, 64.6), with a
radiology.rsna.org n Radiology: Volume 272: Number 1—July 2014
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Zhao et al
Figure 1
Figure 1: Images in a 65-year-old man with grade 3 deep intramuscular sarcoma (undifferentiated pleomorphic sarcoma). The MR images show typical features of a high-grade tumor, including poor margin definition on images obtained with all sequences, internal signal intensity heterogeneity, perilesional signal intensity abnormalities, and large size. (a) Axial T1-weighted MR image (466/16) shows a large 10.5 3 3.4 3 5.8-cm heterogeneously hyperintense mass with indistinct margins replacing the lateral gastrocnemius (arrow). (b) Axial T2-weighted MR image (fat suppressed 3380/60) shows internal signal intensity heterogeneity, with indistinct margins and perilesional edema (arrows) in the posterior calf. (c) Axial contrast-enhanced T1-weighted image (fat suppressed 475/12) shows heterogeneous enhancement within the mass (long arrow) with perilesional muscular, fascial, and subcutaneous enhancement (short arrows). (d) In this microscopic image (hematoxylin-eosin stain; original magnification, 320), the tumor is seen infiltrating subcutaneous adipose tissue. There is a cuff of lymphoplasmacytic inflammation at the interface of the pleomorphic malignant cells and the adipose tissue, possibly accounting for the perilesional edema and enhancement. (e) In this microscopic image (hematoxylin-eosin stain; original magnification, 320), the malignant pleomorphic cells are seen infiltrating skeletal muscle in a ragged pattern.
sensitivity and specificity of 91% and 57%, respectively, while accounting for other features (tumor size, tissue layer, tumor margin on images obtained with all sequences, signal heterogeneity with other sequences, peritumoral high signal intensity on T2-weighted images, age, and patient sex) in the model. (Table 3 shows the results of the final logistic regression model.) The tumor margin definition and peritumoral
contrast enhancement were correlated; however, there was no interaction between these variables in the logistic regression model. Figure 3 emphasizes the importance of MR imaging characteristics in the assignment of grade, as it is an example of a sarcoma for which there was a change in tumor grade between the percutaneous biopsy results (which showed a low-grade tumor) and the final surgical specimen after
Radiology: Volume 272: Number 1—July 2014 n radiology.rsna.org
resection (which showed a high-grade tumor); peritumoral enhancement was a clue to the presence of a high-grade sarcoma in this case.
Discussion For a patient with STS, the histologic tumor grade is the most critical piece of information needed for treatment planning, as it is heavily tied to the patient’s 197
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Zhao et al
Figure 2
Figure 2: Images in a 48-year-old woman with a grade 1 deep intramuscular myxoid liposarcoma. The MR images depict typical features of a low-grade tumor, including good margin definition with all imaging sequences, relatively homogeneous internal signal intensity, lack of perilesional signal intensity on T2-weighted and contrast-enhanced T1-weighted images, and small size. (a) Axial T1-weighted MR image (755/8.2) shows a homogeneous mass that is relatively isointense compared to skeletal muscle located in the medial soleus (arrow). (b) Axial T2-weighted MR image (fat suppressed 2836/124) shows the mass to be relatively homogeneously hyperintense, with no perilesional edema (arrow). The mass has low-signal-intensity internal septations. (c) Axial contrast-enhanced T1-weighted MR image (250/3.9) shows only mildly heterogeneous internal enhancement, without perilesional enhancement (arrow). (d) In this low-power microscopic image (hematoxylineosin stain; original magnification, 34), note the sharply demarcated tumor margin, with pseudocapsule. The upper portion of the field shows skeletal muscle with mixed adipose tissue, a result of peritumoral muscle atrophy. (e) In the high-power microscopic image (hematoxylin-eosin stain; original magnification, 360), the cellularity is monotonous and uniform, and nuclei are evenly spaced. Note the lipoblast in the center of the field, showing lipid droplets that crisply indent the nucleus.
risk of metastasis and overall survival. The pathologic grade of a sarcoma is the parameter used for selecting patients for whom preoperative chemotherapy should be considered (6,7), as high-grade STS may be treated with neoadjuvant chemotherapy, while lowgrade STS is not. To our knowledge, investigators in only one study have attempted to describe differences in MR features of low- and high-grade sarcomas with nonenhanced T1-weighted and T2-weighted imaging. Our study advances prior work by demonstrating a comprehensive assessment of the accuracy of all available MR features of 198
STS (on nonenhanced and contrast-enhanced images) for prediction of a highgrade tumor and developing a model to identify peritumoral contrast enhancement as the strongest independent predictor of a high-grade STS. At presentation, a patient with a soft-tissue mass typically undergoes MR imaging for assessment of tumor extent, although MR imaging characteristics have been described for various histologic soft-tissue tumor types, and some features have been described that can be used to distinguish benign and malignant soft-tissue tumors. When features are rendered indeterminate, a
biopsy of the mass is performed. Subsequently, when a sarcoma is identified by means of biopsy results, the determination of histologic grade is the next step in deciding on neoadjuvant treatment. Because biopsy results are occasionally erroneous with regard to tumor grade (8–10), imaging features that aid in the prediction of a high-grade state can be used as a supplement for biopsy results (when inconclusive or discordant with the MR features) and subsequent treatment. Few prior investigators have identified the differentiating features of lowand high-grade sarcomas. Liu et al (14)
radiology.rsna.org n Radiology: Volume 272: Number 1—July 2014
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Figure 3
Figure 3: MR images in a 65-year-old woman with grade 3 sarcoma. Percutaneous biopsy of this mass initially indicated a grade 1 leiomyosarcoma; however, tumor grade was changed after surgical resection to be a grade 3 leiomyosarcoma. (a) Coronal fat-suppressed T2-weighted MR image (3100/60) shows moderate perilesional hyperintensity along the proximal tumor margin (arrow). (b) Coronal contrast-enhanced T1-weighted image (fat suppressed 560/10) shows mildly heterogeneous enhancement within the mass and minimal perilesional enhancement (arrow). With the results of the current study, after percutaneous biopsy, a high-grade diagnosis could have been considered in this case in which a grade 1 sarcoma was diagnosed, given the perilesional signal intensity changes that are more common with high-grade tumors. At the time of diagnosis, review of the MR images did not prospectively indicate the presence of a high-grade lesion or necessitate repeat biopsy with a different target.
studied only nonenhanced T1-weighted and T2-weighted imaging in the investigation of 59 sarcomas and showed that the histologic grade of STS was significantly related to the definition of tumor margin and peripheral growth pattern, with high-grade tumors more commonly having a poorly defined margin, while low-grade tumors had a well-defined margin on nonenhanced images; peritumoral fluid signal intensity was more common in high-grade tumors, and a low peritumoral signal intensity capsule sign was more common in low-grade
tumors. Our study confirms the importance of tumor margin in predicting grade: Generally, a poorly or partly defined tumor margin on images obtained with any sequence indicates that tumor cells have infiltrated the surrounding tissues and shows the invasive, aggressive nature of the tumor. Indeed, the peripheral tumor growth pattern in sarcomas has been demonstrated to be an important prognostic factor for the development of local recurrence and metastasis (5,20). Fernebro et al studied the peritumoral signal intensity changes
Radiology: Volume 272: Number 1—July 2014 n radiology.rsna.org
Zhao et al
of sarcomas at MR imaging and correlated them with histologic results, confirming that MR imaging provides information about the peripheral growth important to prognosis and the risk of metastasis, although the relationship of this feature to tumor grade was not assessed directly (15). As for a poorly defined margin, our study confirmed that peritumoral fluid signal intensity was significantly more common in highthan in low-grade STS. However, unlike in the study of Liu et al (14), the presence of a capsule sign was not significantly different in low- and high-grade tumors, perhaps because of the larger number of patients investigated in our study population. Intravenous contrast material administration is routinely used for the evaluation of STS, and the present study offers new insights into sarcoma imaging features after contrast enhancement. Most significantly, a partly or poorly defined tumor margin and peritumoral enhancement on contrast-enhanced images were sensitive features associated with highgrade STS, with the multivariate regression model demonstrating peritumoral enhancement to be the strongest independent feature for prediction of highgrade STS. The latter findings corroborate prior work that showed malignant tumors to have increased vascularity at the periphery (21) and reflect the more aggressive peripheral growth pattern of high-grade tumors. Though the signal intensity heterogeneity on T2-weighted images was significantly different between high-grade and low-grade STS, the degree of contrast enhancement within a tumor was not different, with both types of tumors being equally likely to enhance heterogeneously. The latter observation was unexpected, as highergrade tumors usually have increased necrosis (16,22). Tumor depth has also been found to be an important predictor of prognosis and the development of distant metastases (2,3,23). In the present study, both low- and high-grade sarcomas occurred superficially, as well as in the deep tissue layers. When observed independently, we found that the MR imaging features of high- and 199
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
Table 2 Sensitivity and Specificity of Each MR Imaging Feature for Prediction of High-Grade Sarcoma Feature Deep tissue layer Size 5 cm Hypointensity or hyperintensity on T1-weighted images Signal intensity heterogeneity on T1-weighted images Poorly defined or mixed-definition margin on T1-weighted images Absence of peritumoral capsule sign Hypointensity or hyperintensity on T2-weighted images Signal intensity heterogeneity on T2-weighted images Poorly defined and mixed-definition margin on T2-weighted images Internal low-signal-intensity septations Peritumoral high signal intensity on T2-weighted images Absence of peritumoral fat cap Periosteal reaction Cortex extension Marrow extension Joint extension Neurovascular encasement Vascular occlusion Poorly defined and mixed-definition margin on contrast-enhanced T1-weighted images Enhancement and/or heterogeneity . 25% Peritumoral contrast enhancement†
Sensitivity (%)*
Specificity (%)*
71 (56/79) 80 (63/79) 48 (38/79) 72 (57/79) 72 (57/79) 47 (37/79) 96 (76/79) 95 (75/79) 76 (60/79) 77 (61/79) 96 (76/79) 70 (55/79) 16 (13/79) 6 (5/79) 4 (3/79) 4 (3/79) 16 (13/79) 1 (1/77) 74 (50/68)
31 (5/16) 56 (9/16) 75 (12/16) 38 (6/16) 69 (11/16) 56 (9/16) 7 (1/15) 27 (4/15) 73 (11/15) 33 (5/15) 20 (3/15) 40 (6/15) 100 (15/15) 100 (15/15) 100 (15/15) 100 (15/15) 93 (14/15) 93 (14/15) 86 (12/14)
90 (61/68) 91 (62/68)
14 (2/14) 57 (8/14)
Note.—Numbers in parentheses are raw data. * The denominators are variable because contrast-enhanced images were only available in 82 of 95 patients, and images acquired with fluid-sensitive sequences with fat-suppressed T2-weighted imaging or short inversion time inversion-recovery imaging were available in 94 of 95 patients. For assessment of vascular occlusion, images in two patients were deemed inadequate for optimal occlusion assessment. †
Peritumoral contrast enhancement is an independent predictor of high-grade status, with an odds ratio of 13.6 (95% confidence interval: 2.9, 64.6).
Table 3 Results of Final Multivariate Regression Model Variable Peritumoral contrast enhancement Poorly defined or mixed-definition margin on T1-weighted images Poorly defined and mixed-definition margin on contrast-enhanced T1-weighted images Hypointensity or hyperintensity on T1-weighted images Absence of peritumoral fat cap
low-grade superficial STS were similar, including such features as the definition of tumor margin and tumor size. One possible explanation for the similarity in low- and high-grade superficial STS may lie with the fact that subcutaneous 200
Odds Ratio
95% Confidence Interval
P Value
13.6 4.4
2.9, 64.6 0.9, 21.7
.001 ..05
5.9
0.7, 49.1
..05
4.3 0.6
0.1, 22.6 0.1, 22.4
..05 ..05
tumors can be detected sooner than deep lesions (24), and, hence, highgrade subcutaneous tumors can be detected earlier than those found in the deep tissues, and the MR features may not be as developed as those of deep
Zhao et al
high-grade sarcomas. In addition, highgrade subcutaneous sarcomas reportedly have a better prognosis than their deeper counterparts (25,26). Tumor size has been established as one of the important factors related to clinical outcome and is included in prognostic systems in many studies, with a traditional threshold of 5 cm (27,28): According to the literature, tumors at least 5 cm in diameter have lower overall survival and a higher risk of recurrence and distant metastases (3,5,29). Hence, as expected, high-grade tumors were more likely to be larger than 5 cm in our study, while low-grade tumors were smaller than 5 cm. Our study has limitations. As a retrospective study, the MR imaging protocol, while containing T1-weighted, T2-weighted, and contrast-enhanced T1-weighted imaging, was not uniform, with some sequences having fat suppression and others none. Second, there were relatively few low-grade cases of STS as compared with high-grade tumors, but this proportion accurately reflects the population distribution of low- and high-grade STS in the community (30). Third, all histologic types were assessed communally; there may be differentiating MR features among the different histologic subtypes of STS that confound the results of this study, but a subgroup analysis was not possible with the number of patients that had specific histologic types in this study. In addition, since the premise of our study was the characterization of grade in patients with known STS, these study results cannot be generalized to the population of patients with an indeterminate soft-tissue mass (benign or malignant); benign soft-tissue masses with aggressive features, such as myositis ossificans, may share MR imaging features that mimic high-grade sarcomas, but our study was not designed to investigate the distinguishing imaging features of benign and malignant soft-tissue masses. In conclusion, while tumor size, tumor margin, heterogeneous signal intensity on T2-weighted images, and peritumoral high signal intensity on T2-weighted images can be used to
radiology.rsna.org n Radiology: Volume 272: Number 1—July 2014
MUSCULOSKELETAL IMAGING: Can MR Imaging Be Used to Predict Tumor Grade in Soft-Tissue Sarcoma?
differentiate high- and low-grade STS, the presence of peritumoral contrast enhancement is a feature that may be solely used to diagnose high-grade STS. Along with percutaneous biopsy results, MR imaging features provide an additional tool for determining tumor grade. Disclosures of Conflicts of Interest: F.Z. No relevant conflicts of interest to disclose. S.A. No relevant conflicts of interest to disclose. S.J.F. No relevant conflicts of interest to disclose. K.L.W. No relevant conflicts of interest to disclose. E.A.M. No relevant conflicts of interest to disclose. J.A.C. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: author received grants from Siemens, Carestream, and Toshiba and received personal fees from Siemens, BioClinica, Pfizer, Medtronic, and General Electric for consulting and/or lectures. Other relationships: none to disclose. L.M.F. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: author received grants from Siemens Medical Systems and GERRAF for MR spectroscopy. Other relationships: none to disclose.
References 1. Kolovich GG, Wooldridge AN, Christy JM, Crist MK, Mayerson JL, Scharschmidt TJ. A retrospective statistical analysis of highgrade soft tissue sarcomas. Med Oncol 2012;29(2):1335–1344. 2. Pisters PW, Leung DH, Woodruff J, Shi W, Brennan MF. Analysis of prognostic factors in 1,041 patients with localized soft tissue sarcomas of the extremities. J Clin Oncol 1996;14(5):1679–1689. 3. Stefanovski PD, Bidoli E, De Paoli A, et al. Prognostic factors in soft tissue sarcomas: a study of 395 patients. Eur J Surg Oncol 2002;28(2):153–164.
7. Menendez LR, Ahlmann ER, Savage K, Cluck M, Fedenko AN. Tumor necrosis has no prognostic value in neoadjuvant chemotherapy for soft tissue sarcoma. Clin Orthop Relat Res 2007;455(455):219–224. 8. Yang J, Frassica FJ, Fayad L, Clark DP, Weber KL. Analysis of nondiagnostic results after image-guided needle biopsies of musculoskeletal lesions. Clin Orthop Relat Res 2010;468(11):3103–3111. 9. Strauss DC, Qureshi YA, Hayes AJ, Thway K, Fisher C, Thomas JM. The role of core needle biopsy in the diagnosis of suspected soft tissue tumours. J Surg Oncol 2010;102(5):523–529. 10. Mitsuyoshi G, Naito N, Kawai A, et al. Accurate diagnosis of musculoskeletal lesions by core needle biopsy. J Surg Oncol 2006;94(1):21–27. 11. Totty WG, Murphy WA, Lee JK. Softtissue tumors: MR imaging. Radiology 1986;160(1):135–141. 12. Walker EA, Salesky JS, Fenton ME, Mur phey MD. Magnetic resonance imaging of malignant soft tissue neoplasms in the adult. Radiol Clin North Am 2011;49(6):1219– 1234, vi. 13. Kransdorf MJ, Jelinek JS, Moser RP Jr, et al. Soft-tissue masses: diagnosis using MR imaging. AJR Am J Roentgenol 1989;153(3):541–547. 14. Liu QY, Li HG, Chen JY, Liang BL. Correlation of MRI features to histopathologic grade of soft tissue sarcoma [in Chinese]. Ai Zheng 2008;27(8):856–860. 15. Fernebro J, Wiklund M, Jonsson K, et al. Focus on the tumour periphery in MRI evaluation of soft tissue sarcoma: infiltrative growth signifies poor prognosis. Sarcoma 2006;2006:21251.
4. Zagars GK, Ballo MT, Pisters PW, et al. Prognostic factors for patients with localized soft-tissue sarcoma treated with conservation surgery and radiation therapy: an analysis of 1225 patients. Cancer 2003;97(10):2530–2543.
16. Guillou L, Coindre JM, Bonichon F, et al. Comparative study of the National Cancer Institute and French Federation of Cancer Centers Sarcoma Group grading systems in a population of 410 adult patients with soft tissue sarcoma. J Clin Oncol 1997;15(1):350–362.
5. Mandard AM, Petiot JF, Marnay J, et al. Prognostic factors in soft tissue sarcomas. A multivariate analysis of 109 cases. Cancer 1989;63(7):1437–1451.
17. Rydholm A. Surgical margins for soft tis sue sarcoma. Acta Orthop Scand Suppl 1997;273:81–85.
6. Lucas DR, Kshirsagar MP, Biermann JS, et al. Histologic alterations from neoadjuvant chemotherapy in high-grade extremity soft tissue sarcoma: clinicopathological correlation. Oncologist 2008;13(4):451–458.
18. Takahashi M, Sato K, Miura T. MR imaging of musculoskeletal sarcomas: the clinical significance of peritumoral low signal intensity lines in planning surgical margins. Nippon Seikeigeka Gakkai Zasshi 1993;67(10): 881–896.
Radiology: Volume 272: Number 1—July 2014 n radiology.rsna.org
Zhao et al
19. Reynoso D, Subbiah V, Trent JC, et al. Neoadjuvant treatment of soft-tissue sarcoma: a multimodality approach. J Surg Oncol 2010;101(4):327–333. 20. Engellau J, Bendahl PO, Persson A, et al. Improved prognostication in soft tissue sarcoma: independent information from vascular invasion, necrosis, growth pattern, and immunostaining using whole-tumor sections and tissue microarrays. Hum Pathol 2005;36(9):994–1002. 21. Ma LD, Frassica FJ, McCarthy EF, Bluemke DA, Zerhouni EA. Benign and malignant musculoskeletal masses: MR imaging differentiation with rim-to-center differential enhancement ratios. Radiology 1997;202(3):739–744. 22. Coindre JM, Trojani M, Contesso G, et al. Reproducibility of a histopathologic grading system for adult soft tissue sarcoma. Cancer 1986;58(2):306–309. 23. Gaynor JJ, Tan CC, Casper ES, et al. Refinement of clinicopathologic staging for localized soft tissue sarcoma of the extremity: a study of 423 adults. J Clin Oncol 1992;10(8):1317–1329. 24. Sawamura C, Matsumoto S, Shimoji T, Tanizawa T, Ae K. What are risk factors for local recurrence of deep high-grade softtissue sarcomas? Clin Orthop Relat Res 2012;470(3):700–705. 25. Collin C, Godbold J, Hajdu S, Brennan M. Localized extremity soft tissue sarcoma: an analysis of factors affecting survival. J Clin Oncol 1987;5(4):601–612. 26. Peabody TD, Monson D, Montag A, Schell MJ, Finn H, Simon MA. A comparison of the prognoses for deep and subcutaneous sarcomas of the extremities. J Bone Joint Surg Am 1994;76(8):1167–1173. 27. Rööser B, Attewell R, Berg NO, Rydholm A. Prognostication in soft tissue sarcoma. A model with four risk factors. Cancer 1988;61(4):817–823. 28. Levine EA. Prognostic factors in soft tissue sarcoma. Semin Surg Oncol 1999;17(1): 23–32. 29. Saddegh MK, Lindholm J, Lundberg A, Nilsonne U, Kreicbergs A. Staging of softtissue sarcomas. Prognostic analysis of clinical and pathological features. J Bone Joint Surg Br 1992;74(4):495–500. 30. Carneiro A, Bendahl PO, Engellau J, et al. A prognostic model for soft tissue sarcoma of the extremities and trunk wall based on size, vascular invasion, necrosis, and growth pattern. Cancer 2011;117(6):1279–1287.
201
