Bone scintigraphy remains the method of choice in the evaluation of bone metastases because of its accessibility, reasonable cost, and ability to show the entire skeletal system. However, MR imaging has been advocated as a useful method for the evaluation of bone marrow diseases, including bone metastases [1,2,3,4,5,6,7,8]. Nonetheless, the detection rates of vertebral metastases can be different between these two modalities and may cause a clinical dilemma in patient treatment. Previous studies have reported that positive MR imaging findings and negative findings on bone scans can occur in patients with vertebral body metastases [4, 9,10,11,12,13,14,15,16]. To our knowledge, the cause of such a difference remains undetermined. Many possible explanations have been proposed [4, 7, 11, 15,16,17]. To our knowledge, none has been proven. The purpose of this study was to investigate whether lesion size and, in particular, its spatial relationship to the bony cortex can influence the differences in detection rates of vertebral body metastases between MR imaging and bone scintigraphy.

From the clinical charts of 450 patients with a diagnosis of widely disseminated spinal metastases in the past 10 years at our institution, 74 patients who had undergone both bone scintigraphy and unenhanced MR studies of the spine within a 4-week period (mean, 6.2 days) were selected for retrospective evaluation. There were 42 males and 32 females between the ages of 9 and 88 years (mean age, 59 years). The primary tumors included carcinomas of breast (n = 22), lung (n = 21), prostate (n = 13), genitourinary tract (n = 5), gastrointestinal tract (n = 2), and other carcinomas (n = 11). Bone scintigraphy was performed with technetium-99m methyl diphosphonate (Tc-99m MDP). The dose of Tc-99m MDP was 22 mCi (814 MBq), and a high-resolution collimator was used. Because single-photon emission computed tomography (SPECT) was not performed in all patients, only planar imaging was evaluated. Metastases were primarily examined on axial and sagittal T1-weighted MR images (spin echo; TR range/TE range, 316-650/11-20), and T2-weighted images (spin echo; TR/TE range, 2000/90-100) were available for reference. The MR images and bone scintigrams were reviewed independently by three radiologists. Lesions were included as metastatic foci only if they were hypointense on T1-weighted imaging and hyperintense on T2-weighted imaging. Schmorl's nodes, benign bone islands, and hemangiomas were excluded on the basis of their characteristic radiologic findings. Differences in interpretation were solved by consensus. Histopathologic confirmation of each metastatic focus was not available because most of the patients had end-stage disease.

The size of each lesion was measured at its greatest dimension on MR images and was categorized as small (<2 cm) or large (≥2 cm). Coalescent lesions were included in the large-lesion group. We classified the relationship between the lesion and cortical bone on the basis of the sagittal and axial T1-weighted images by determining whether the lesion was distant from (intramedullary), had contact with (subcortical), or had invaded the bony cortex (transcortical). The three radiologists classified lesion-cortical bone relationships without knowledge of the scintigraphic results. MR imaging findings were compared with those of planar bone scintigraphy. The Pearson's chisquare test and two-tailed t test were used for statistical comparisons.

The probability of positive findings on a bone scan was dependent on lesion size (small versus large) and location (intramedullary, subcortical, and transcortical). Both the Pearson's chi-square and the two-tailed t tests confirmed a statistically significant correlation between cortical involvement and the size of the lesion (p < 0.0001), cortical involvement and bone scintigraphic results (p < 0.0001), and lesion size and bone scintigraphic results (p < 0.0001).

Bone metastases are by far the most common malignant bone tumors seen in adults. The prevalence of bone metastases in patients with known primary cancer is about 70% of the patients with metastases [18]. Bone metastases may occur with almost all malignancies, but they are most common in carcinomas of the breast (47-85%), lung (32%), prostate (54-85%), kidney (33-40%), or thyroid (28-60%) [18, 19].

The spine is the most common site of skeletal metastases (39%) because of the abundant vascularization and red bone marrow [5, 13, 15]. Most bone metastases are hematogenous in origin [5, 20]. The initial seeding of metastatic deposits via hematogenous spread is typically localized in the hematopoietic (red) marrow. This location explains the predominance of metastatic bone lesions in the axial skeleton (>90% of metastatic bone lesions) [11].

Back pain is the most frequent initial complaint in patients with spinal metastatic disease [9, 21, 22] because lesions invade the richly innervated periosteum with or without detected invasion, possibly through the cortex via the haversian canals. The imaging of spinal metastatic disease may include conventional radiography, myelography, radionuclide bone scintigraphy, CT, and MR imaging. Radiography, CT, and bone scans assess mainly the bony abnormality, particularly the cortex, whereas MR imaging examines bone marrow, in which the early metastatic deposits frequently occur [20]. Findings of abnormal bone on radiography, bone scintigraphy, and CT mainly result from tumor invasion of the cortical bone rather than of the medullary bony matrix. Although conventional radiography is useful for determining the integrity of cortical bone [10], at least 50% of the cortical bone mass must be destroyed before a lesion can be clearly seen on conventional radiography [23]. CT is sensitive in detecting subtle cortical invasion but is less sensitive for medullary bone or bone marrow involvement [10, 13]. Bone destruction can be difficult to detect on CT in the presence of osteoporosis or degenerative changes, which are common in elderly adults with cancer [9].

The mechanism of abnormal Tc-99m MDP uptake shown in bone scanning is complex. Abnormal radionuclide uptake is generally believed to increase with regional bone-blood flow, bone remodeling, formation of new bone, and enhanced bone matrix turnover [12, 24,25,26,27]. Bone scintigraphy is sensitive for detecting areas of bone remodeling, particularly cortex, associated with metastatic deposits as little as 5-10% in the lesion-to-normal bone ratio, for abnormal focus to be appreciated on planar scintigraphy [4, 11]. Scintigraphy may reveal bone metastases up to 18 months before radiography shows them and has a 50-80% greater sensitivity [28]. However, abnormal increase of uptake in the medullary cavity because of tumor destruction of the medullary bony matrix may not be as obvious because of the relatively small amount of medullary bone and relatively high uptake of the normal cortical bone.

MR imaging is a sensitive method of detecting intramedullary metastases to those bones with large marrow cavities such as vertebral bodies [10, 20]. With increased capability for scanning the whole body with fast imaging techniques, screening the bone marrow cavities can be cost-effective [29, 30]. Early results suggest that the skeletal system of the entire body can be scanned within 30-45 min on turbo short tau inversion recovery (STIR) imaging and within 6 min with total body echo-planar imaging, and the detectability of metastatic tumor with these imaging techniques seems to be better than that with bone scintigraphy. However, turbo STIR imaging and total body echo-planar imaging are not widely used methods and are suboptimal for the metastatic workup of the entire skeletal system on classic MR sequences. MR imaging is not cost-effective in examining bones with small cavities such as ribs because it cannot globally examine the entire skeletal system as bone scintigraphy can [10]. Therefore, in general, the role of MR imaging in the examination of suspected bone metastases is limited to those bones with large bone marrow cavities such as the spine or proximal extremities and is complementary to other imaging methods [1, 8, 12].

Because the vertebral body has a relatively large marrow cavity, early or small metastases tend to be intramedullary lesions without cortical involvement [20] and may not cause sufficient bony remodeling to be detected on bone scans. In our study, positive findings on bone scans were always associated with MR evidence of cortical involvement. Findings on bone scans were always negative for intramedullary lesions (without cortical involvement), although only one large lesion did not have cortical involvement [11, 17]. Small lesions or lesions localized away from the cortex are, therefore, likely to be undiagnosed on bone scintigraphy, despite destruction of trabecular bone. Even if most of the bone marrow has been infiltrated with metastases, the uptake of radioactive tracers caused by the destruction of the relatively small amount of medullary bony matrix remains low and, therefore, may not be easily appreciated when the uptake is contrasted with that of the normal cortex.

There are limitations in our study. Our purpose was to investigate possible causes for the difference in detection rates of bone metastases on MR imaging and bone scintigraphy, and we did not attempt to compare the sensitivity and specificity of these two methods. Our results did not prove or disprove that cortical involvement is the cause of the difference in the detection rates. In addition, only planar bone scans were used because not all patients underwent SPECT bone scintigraphy. The patients in our study were limited to those with widely disseminated metastatic disease detected on MR imaging, and pathologic confirmation was not available in this group of patients.

In conclusion, our findings suggest that cortical involvement is likely a key factor contributing to the difference in detection rates of vertebral body metastases on MR imaging and bone scintigraphy. Findings of early and small vertebral metastatic deposits in the marrow cavity may be negative on bone scans. Although MR imaging is useful in the detection of early metastases that are localized completely in the bone marrow cavity, bone scintigraphy remains the most cost-effective method for examination of the entire skeleton. MR imaging can be useful in cases in which bone scan findings remain negative and vertebral involvement is suspected on the basis of clinical findings. However, MR imaging is inadequate in assessing cortical involvement [31, 32].

Address correspondence to W.T.C. Yuh.