| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Musculoskeletal Imaging |
1 From the Departments of Diagnostic Imaging (D.A.E., L.M.W., D.J.S.), Pathology and Laboratory Medicine (R.A.K.), and Medical Imaging and Public Health Sciences (G.T.), and University Musculoskeletal Oncology Unit (R.S.B., J.S.W.), Mount Sinai Hospital and the University Health Network, University of Toronto, 600 University Ave, Toronto, Ontario, Canada M5G 1X5. From the 2001 RSNA scientific assembly. Received March 28, 2002; revision requested June 10; final revision received January 16, 2003; accepted January 27. Address correspondence to L.M.W. (e-mail: lwhite@mtsinai.on.ca).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Preoperative MR images (1.5 T, transverse and longitudinal planes, T1 and T2 weighted) of 56 osseous sites in 51 patients who underwent bone resection at surgery for soft-tissue sarcoma were assessed retrospectively for signs of osseous invasion, by consensus of two readers who were blinded to clinical and histopathologic findings. Delay between MR imaging and surgery averaged 34 days (range, 2–112 days). MR signs assessed included osseous abutment by tumor, maximal diameter of osseous abutment, extent of circumferential abutment of long bones (<25%, 25%–50%, >50%), cortical destruction, and cortical and medullary signal intensity change on T1- and T2-weighted images. Imaging findings were correlated with histopathologic findings. Sensitivities, specificities, positive and negative predictive values (PPV, NPV), and P values were calculated.
RESULTS: Eleven sites (20%) showed osseous invasion histologically (two, cortical; nine, both cortical and medullary). Tumor abutted bone at 44 lesion sites (sensitivity, 100%; specificity, 27%). Maximal diameter of osseous abutment and extent of circumferential abutment did not significantly affect osseous invasion (P = .09 and .11, respectively). On T1- and T2-weighted images, 13 lesion sites showed cortical signal intensity change (sensitivity, 100%; specificity, 96%) and 10 showed cortical destruction (sensitivity, 82%; specificity, 98%). Eleven sites showed decreased medullary T1 signal intensity (sensitivity, 100%; specificity, 96%), and 12 showed increased medullary T2 signal intensity (sensitivity, 100%; specificity, 94%). MR imaging overall had a sensitivity of 100%, specificity of 93%, PPV of 79%, and NPV of 100% for detection of osseous invasion on the basis of any finding of cortical destruction or cortical or medullary signal intensity change on T1- or T2-weighted images (P < .001).
CONCLUSION: On T1- and T2-weighted MR images, findings of cortical and medullary signal intensity change and cortical destruction were sensitive and specific for detection of osseous invasion by soft-tissue sarcoma.
© RSNA, 2003
Index terms: Bone neoplasms, MR, 40.12141 • Bone neoplasms, secondary, 40.37 • Sarcoma
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Studies have suggested that pathologic evidence of osseous invasion is associated with a shortened survival following diagnosis (1,8). Furthermore, in patients treated at the University Musculoskeletal Oncology Unit at the University of Toronto, osseous invasion has been associated with a greater proportion of high-grade versus low-grade soft-tissue sarcomas. According to the system of soft-tissue sarcoma staging used by Enneking et al (9), the grade and dimensions of the tumor and the presence or absence of metastases are key prognostic factors. In this staging system, the extent of sarcoma is characterized as either intra- or extracompartmental. Intracompartmental sarcoma is bounded by natural barriers to tumor extension, such as cartilage, fascia, synovium, periosteum, or bone; osseous invasion constitutes extracompartmental tumor extension.
Accurate radiologic assessment of the extent of the soft-tissue sarcoma provides critical anatomic information for planning both the surgical approach and the treatment field for adjuvant radiation therapy (5,10). Magnetic resonance (MR) imaging has emerged as the preferred modality for evaluation of soft-tissue sarcoma (4,11–15). Our aim in this study was to assess the overall accuracy of MR imaging and of specific MR imaging signs for the detection of osseous invasion in soft-tissue sarcoma, with histopathologic analysis used as the standard of reference.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
The study group included 51 patients (22 men, 29 women) with a mean age of 55.7 years (range, 17–88 years), all of whom had given prior consent for the use of their medical records and images in research. Institutional review board approval also was obtained for the review of patient records and images.
Imaging and Image Evaluation
MR imaging of all patients was performed with a 1.5-T imager (Signa; GE Medical Systems, Milwaukee, Wis). T1- and T2-weighted images were obtained in the transverse plane and at least one longitudinal plane (coronal or sagittal) by using either a surface coil or a body coil. For T1-weighted imaging, a conventional spin-echo pulse sequence was used (repetition time msec/echo time msec, 400–680/8–16); for T2-weighted imaging, a fast spin-echo pulse sequence was used (3,500–5,867/80–105; echo train length, eight). All T2-weighted images were acquired with fat saturation except those from six patients in whom a conventional spin-echo pulse sequence was used without fat suppression. The matrix size, section thickness, and field of view varied according to the anatomic location and size of each lesion.
All images were evaluated retrospectively and by consensus of two readers (L.M.W., D.J.S.) who were blinded to clinical and histopathologic findings and who reviewed the images together. After the anatomic location of each lesion was identified, the images were assessed as follows:
1. A determination of osseous abutment by the tumor was made on the basis of evidence on T1-weighted images of a loss of normal soft-tissue interface between the tumor mass and the adjacent cortex. Osseous abutment was considered absent if a completely normal tissue interface was observed between the tumor and adjacent bone. If no osseous abutment was evident on T1-weighted images, T2-weighted images of the same patient were evaluated to determine whether peritumoral edema or reactive change was present that extended to the cortical surface. Peritumoral edema and reactive change were identified on T2-weighted images as regions of increased soft-tissue signal intensity with ill-defined margins and without mass effect or distortion of the soft-tissue interface without corresponding abnormality on T1-weighted images.
2. If osseous abutment was present, its extent was determined by measuring its maximal diameter on MR images. Additionally, in lesions involving long bones, the maximal extent of circumferential abutment was evaluated as less than 25%, 25%–50%, or more than 50% of the total circumference of the adjacent bone.
3. Cortical signal intensity change was defined as increased signal intensity in the normally hypointense cortical bone adjacent to the tumor. Cortical signal intensity change was assessed on both transverse and longitudinal images to avoid false-positive diagnoses resulting from partial volume averaging along obliquely imaged cortical margins, as well as to assist in the identification of nutrient vessels.
4. Cortical destruction was defined as a defect in the cortical bone adjacent to the tumor. All images on which cortical destruction was identified also showed cortical signal intensity change, whereas not all images that showed cortical signal intensity change also indicated cortical destruction, because some signal intensity changes were due to malignant cell infiltration of the cortex without obvious cortical erosion or destruction.
5. Medullary signal intensity change on T1-weighted images was defined as focal replacement of the normally high signal intensity of bone marrow by low signal intensity in the medulla adjacent to the tumor.
6. Medullary signal intensity change on T2-weighted images was defined as focal replacement of the normal bone marrow signal intensity (which may be intermediate or low, depending on whether fat suppression was used) by high signal intensity in the medulla adjacent to the tumor.
Surgery and Histologic Analysis
The mean time interval between MR imaging and surgical resection was 34 days (range, 2–112 days), but for four of 51 patients the delay was greater than 8 weeks. Thirty-one (61%) of 51 patients received adjuvant preoperative radiation therapy.
All osseous tissue resected from the tumor sites was subjected to histologic analysis for evidence of tumor invasion of the cortex and medulla. For each lesion, histologic evaluation was performed by an orthopedic pathologist (R.A.K.) with extensive experience and training in sarcoma evaluation. Resected osseous tissue specimens were evaluated visually or with specimen radiography. Bone was sampled at the sites in closest proximity to the tumor and at sites of apparent abnormality on specimen radiographs. The bone was decalcified and embedded in paraffin. Slices with a thickness of 5 µm were cut, stained with hematoxylin-eosin, and examined with light microscopy. Bone was considered to be involved when tumor cells were seen in the cortex or the medulla. Bone was considered not to be involved when tumor cells were seen in the periosteum but not in the cortex or medulla, or when subperiosteal new bone formation was apparent in which no tumor cells were present.
Statistical Analysis
Statistical analysis was performed by using histopathologic findings in surgically resected bone as the standard of reference. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated separately for the presence of osseous abutment, cortical and medullary signal intensity change, and cortical destruction. A two-tailed Fisher exact test was performed, and P values were calculated for each imaging sign. P < .05 was considered to indicate a statistically significant difference.
Again using histopathologic findings as the standard of reference, the overall performance of MR imaging was assessed for the detection of osseous invasion on the basis of any finding of cortical destruction or cortical or medullary signal intensity change on T1- or T2-weighted images. The previously described statistics were recalculated for this combination of MR imaging signs.
Logistic regression analysis was performed, with presence or absence of osseous invasion as the outcome and with maximal diameter of osseous abutment as the predictor. In addition, a logistic regression model with presence or absence of osseous invasion as the outcome was used to test for a linear trend across lesion groups defined by circumferential percentage of osseous abutment.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
The anatomic locations of the sites were as follows: lower extremity (34 [61%] of 56 sites), shoulder girdle (nine [16%] of 56), upper extremity (seven [13%] of 56), pelvic girdle (four [7%] of 56), and trunk (two [4%] of 56).
The lesions were identified at histologic analysis as malignant fibrous histiocytoma (20 [36%] of 56 sites), synovial sarcoma (eight [14%] of 56), leiomyosarcoma (seven [13%] of 56), liposarcoma (seven [13%] of 56), extraosseous chondrosarcoma (three [5%] of 56), fibrosarcoma (three [5%] of 56), clear cell sarcoma (two [4%] of 56), and sarcoma of miscellaneous or unspecified type (six [11%] of 56).
Histologically proved neoplastic osseous invasion had occurred in 11 (20%) of the 56 osseous sites; involvement of the cortex alone was evident in two sites, and involvement of both the cortex and medulla was seen in nine (Table 1). No evidence of either cortical or medullary involvement was found at histologic evaluation of tissue samples from the other 45 sites.
|
|
|
There were 35 sites in which osseous abutment by the tumor involved long bones. The maximal circumferential extent of tumor abutment involving long bones was less than 25% in 17 of the 35 osseous sites, 25%–50% in 11, and more than 50% in seven (Fig 2). The numbers of sites with histologically proved osseous invasion in each measurement group were as follows: two (12%) of 17 sites in the less-than-25% group, four (36%) of 11 sites in the 25%–50% group, and three (43%) of seven sites in the more-than-50% group. Results of a statistical test for linear trend in logistic regression showed no significant association between circumferential extent of osseous abutment and presence or absence of osseous invasion (P = .11).
|
|
|
|
|
Medullary Signal Intensity Change
Medullary signal intensity changes on T1-weighted images and T2-weighted images were correlated with histologic findings of medullary involvement and of osseous invasion as a whole (both medullary and cortical involvement). In the nine sites in which intramedullary invasion was histologically proved (Fig 4), coexisting cortical involvement was also confirmed histologically, reflective of direct tumor extension through the cortex and into the medulla.
|
|
Twelve sites were considered to have medullary signal intensity change on T2-weighted images (sensitivity, 100%; specificity, 94%; PPV, 75%; NPV, 100%; P < .001). All nine sites with histologically proved medullary involvement had corresponding focal increased medullary signal intensity on T2-weighted images. There were three false-positive findings of medullary involvement, based on signal intensity change on T2-weighted images; two of these findings were supported also by signal intensity change on T1-weighted images (described in the preceding paragraph). The third false-positive finding based on medullary signal intensity change on T2-weighted images was not confirmed by findings on T1-weighted images. In this case, histologic analysis revealed reactive change only. The three lesions involved in these false-positive findings had been subjected to preoperative radiation therapy.
Overall Accuracy of MR Imaging Signs
Fourteen (25%) of the 56 sites were considered positive for osseous invasion on the basis of observation of at least one of the following: cortical signal intensity change on T1-weighted images or T2-weighted images, cortical destruction, or medullary signal intensity change on T1-weighted images or T2-weighted images. No false-negative findings and only three false-positive findings of osseous invasion were made on the basis of this combination of MR imaging signs (sensitivity, 100%; specificity, 93%; PPV, 79%; NPV, 100%; P < .001) (Table 3).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
We defined osseous abutment by the tumor as loss of a normal soft-tissue interface between the tumor mass and osseous cortex on T1-weighted images. For each of the 12 sites in which T1-weighted images showed no osseous abutment, T2-weighted images were evaluated for peritumoral edema and reactive change extending to the cortical surface. It can be difficult to differentiate between tumor tissue and peritumoral edema or reactive change, but features favoring the latter include areas of increased signal intensity in soft tissue on T2-weighted images, with ill-defined margins, no mass effect, no distortion of the soft-tissue interface, and no corresponding signal intensity change on T1-weighted images. Peritumoral edema or reactive change may indicate a vascular or inflammatory response with or without tumor cell infiltration (15). The reactive zone around a sarcoma commonly is removed en bloc with the tumor during limb salvage surgery (12). Furthermore, radiation therapy may contribute to peritumoral changes of increased signal intensity on T2-weighted images (17). In our study, nine of the 12 sites in which osseous abutment was not evident on T1-weighted images had peritumoral edema extending to the bone surface. At histologic analysis, none of these sites were found to have cortical or medullary invasion by tumor. Hence, our study shows that observation of a completely preserved soft-tissue interface on T1-weighted images, even in the presence of peritumoral edema and/or reactive changes extending to the bone surface, has a NPV of 100% for osseous invasion.
Of the 44 sites found positive for osseous abutment on the basis of lesion appearance on MR images, only 11 (25%) showed histologic evidence of osseous invasion. It might be expected that a greater extent of osseous abutment by the tumor would correspond to an increased probability of osseous invasion, but the results of our study do not confirm this. Although there was a trend of increased likelihood of osseous invasion with increasing maximal diameter of osseous abutment, this trend was nonsignificant (P = .09 for logistic regression analysis). Furthermore, an analysis of data for lesions that abutted long bones, based on lesion subcategorization according to maximal circumferential abutment of bone into groups with less than 25%, 25%–50%, and more than 50% abutment, showed no statistically significant association between increasing circumferential abutment and likelihood of osseous invasion (P = .11 for test for linear trend in logistic regression). These findings suggest that bone invasion may be more dependent on the biological grade of sarcoma than on the extent of osseous abutment, although the effect of biologic grade was not evaluated in our study. It is also possible that the number of lesions in our study that were identified as positive for osseous invasion by soft-tissue sarcoma was too small to allow the results of such an analysis to reach statistical significance.
Cortical signal intensity change both on T1-weighted images and on T2-weighted images correlated strongly with histologic evidence of cortical involvement. All 11 lesions found at histologic analysis to have cortical involvement also showed cortical signal intensity changes on T1- or T2-weighted MR images (sensitivity, 100%; NPV, 100%). There were two false-positive findings based on cortical signal intensity changes observed on T1- or T2-weighted images (specificity, 96%; PPV, 85%). Cortical destruction was observed in 10 sites, with two false-negative findings (sensitivity, 82%; NPV, 96%) and one false-positive finding (specificity, 98%; PPV, 90%). Cortical destruction may be less sensitive than cortical signal intensity change for detection of cortical involvement, because tumor cells can permeate the cortex (creating cortical signal intensity change) without causing appreciable cortical destruction. Conversely, cortical destruction would be expected to be more specific because reactive change alone is unlikely to cause cortical structural remodeling.
Focal decrease in medullary signal intensity on T1-weighted images and focal increase in medullary signal intensity on T2-weighted images strongly correlated with histologic evidence of medullary involvement. All nine sites found to have medullary involvement at histologic analysis (all with coexistent cortical involvement) were characterized by decreased medullary signal intensity on T1-weighted images and increased medullary signal intensity on T2-weighted images (sensitivity, 100%; NPV, 100%).
There were two false-positive findings of medullary involvement on the basis of decreased medullary signal intensity on T1-weighted images (specificity, 96%; PPV, 82%) and three false-positive findings on the basis of increased medullary signal intensity on T2-weighted images (specificity, 94%; PPV, 75%). The slightly higher specificity of medullary signal intensity change on T1-weighted images in our study confirms prior reports, primarily concerning tumors with intramedullary involvement, that increased signal intensity on T2-weighted images lacks specificity (because it may be due to peritumoral edema, as well as to osseous invasion) and can lead to overestimation of tumor size and extent (4,5,11,12,15,18). For this reason, T1-weighted images are more accurate for the assessment of intramedullary involvement (11).
In all sites with false-positive findings based on cortical or medullary signal intensity changes on T1- or T2-weighted images, reactive change was found at histologic analysis. All lesions with false-positive findings based on altered intramedullary signal intensity seen on T1- or T2-weighted images had been subjected to preoperative radiation therapy. In these lesions, tumor cells that were present initially in the medulla may have been eliminated by irradiation, leaving behind only reparative tissue. Irradiation also can induce morphologic changes in bone marrow that may affect medullary signal intensity on MR images (19,20) and thus lead to false-positive findings of osseous invasion.
Another potential source of discrepancies between imaging findings and histologic findings with respect to osseous invasion are various limitations inherent in the process of histologic examination. The pathologist is dependent on the surgical description of specimen orientation, and normal anatomic landmarks may be obscured by resection. Moreover, because not all resected bone is subjected to histologic assessment, osseous invasion may be recognized on MR images and theoretically missed at histologic examination (5).
There are potential pitfalls also in the interpretation of the MR imaging signs of osseous invasion. Difficulties may arise in the assessment of osseous abutment by the tumor at sites of muscle or tendon insertion into bone (eg, at the linea aspera of the femur). At such sites, the absence of the soft-tissue interface normally seen between muscle or tendon and bone could be misinterpreted as tumor abutment, because muscle and tumor may have similar signal intensities on T1-weighted images. Next, normal nutrient vessels or islands of hematopoietic marrow may cause focal cortical or medullary signal intensity change on MR images, which could be confused with tumor-related changes. Finally, partial volume averaging may cause the cortical border to appear ill defined on images, leading to a false-positive finding of cortical destruction. This tends to occur especially on transverse images of the tubular long bones, at sites of metadiaphyseal flaring, because the cortex is thin and obliquely oriented to the imaging plane. Correlation of transverse images with longitudinal images is important to avoid this pitfall.
The results of our study indicate that the overall accuracy of MR imaging for the detection of osseous invasion by soft-tissue sarcoma is high with findings of cortical or medullary signal intensity change or cortical destruction on T1- and T2-weighted images (Table 3). Our study, however, had several limitations. Due to the relatively few cases of histologically proved osseous invasion, caution must be exercised in interpreting the absolute statistical values obtained for each MR imaging sign. The relatively small numbers reflect the known resistance of bone to invasion by soft-tissue sarcoma. In addition, selection bias may have been introduced by our inclusion in this study only of subjects in whom bone was resected at surgery. We did not examine images of subjects with soft-tissue sarcoma who did not undergo bone resections, because any imaging findings could not be correlated with histopathologic findings. This accounts for the apparent high rate (20%) of osseous invasion of soft-tissue sarcoma in our study subjects. Nevertheless, we did find an overall NPV of 100% for MR assessment of osseous invasion, and this NPV should be no lower for the general population of soft-tissue sarcoma patients, including those who did not undergo bone resection, since they are even less likely to have osseous invasion than were our study subjects. Therefore, the issue of selection bias does not invalidate the applicability of this key finding to other sarcoma patient populations.
Clustering is another potential limitation, since in three of our 51 study subjects more than one site was assessed for osseous invasion, such that a total of 56 osseous sites were evaluated (two patients had three separate osseous sites resected, and one patient had two separate osseous sites resected). The potential effect of five clustered measurements on our findings in 56 evaluated sites is small, however, and an assessment of the actual effect of clustering with a random effects model is not feasible without a larger set of clustered data.
The delay between MR imaging and surgical resection, and the use of radiation therapy in the interim in some cases, are further potential limitations. These factors may account for some of the discrepancies between imaging findings and histologic findings. The delay between MR imaging and surgical resection in some patients might be expected to have caused an increase in the number of false-negative findings in our study; patients without osseous invasion at MR imaging could have developed osseous invasion by the time of resection. The false-negative rate of 0% for overall MR imaging findings, however, indicates that this potential limitation did not affect our study results.
Other limitations include the fact that we did not assess images for possible periosteal reaction; we considered only imaging findings of direct neoplastic involvement of cortical bone to indicate osseous invasion. In addition, there was a possibility of bias in the interpretation of individual MR imaging signs, because readers could not be blinded to additional signs present on the same image. Furthermore, the MR imaging parameters used in each case were not uniform; matrix size, field of view, and section thickness varied, depending on the anatomic location and size of the lesion. In the earlier MR studies, conventional spin-echo T2-weighted sequences without fat suppression were used, whereas in later studies we used fast spin-echo T2-weighted sequences with fat saturation. Because intravenous administration of gadolinium is not routinely used at our institution for the assessment of soft-tissue sarcoma at MR imaging, we did not have access to contrast material–enhanced images for our assessment.
In conclusion, cortical and medullary signal intensity changes and cortical destruction observed on T1- and T2-weighted MR images are highly sensitive and highly specific signs of osseous invasion by soft-tissue sarcoma, when compared with findings at histopathologic evaluation. Increasing maximal diameter and increasing circumference of osseous abutment by the tumor did not result in a statistically significant increase in the likelihood of osseous invasion.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, L.M.W., D.A.E., D.J.S.; study concepts, L.M.W., R.A.K., D.A.E.; study design, L.M.W., D.J.S., D.A.E.; literature research, L.M.W., D.J.S., D.A.E.; clinical studies, L.M.W., R.A.K., D.J.S., D.A.E., J.S.W., R.S.B.; data acquisition, L.M.W., R.A.K., D.J.S., D.A.E.; data analysis/interpretation, L.M.W., R.A.K., D.J.S., D.A.E., G.T.; statistical analysis, D.A.E., G.T.; manuscript preparation, all authors; manuscript definition of intellectual content, L.M.W., D.A.E., D.J.S.; manuscript editing, revision/review, and final version approval, D.A.E., L.M.W., R.A.K., R.S.B., J.S.W.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
