Does Magnetic Resonance Imaging of the Spine Have a Role in the Staging of Prostate Cancer?
Article Outline
Abstract
Aims
Magnetic resonance imaging (MRI) is an effective method for evaluating the spine in patients with a high risk of metastatic disease. The aim of this study was to compare MRI spine with radionuclide bone scan in detecting spinal metastases for staging prostate cancer patients.
Materials and methods
A cohort of 99 patients with locally advanced prostate cancer at high risk of skeletal metastasis (prostate-specific antigen
>10ng/ml, composite Gleason score≥8) or equivocal findings on bone scan were included in the retrospective study, and their MRI spine and bone scans were analysed.
Results
Ten patients were detected to have definite spinal metastasis by bone scan, whereas 12 patients had definite skeletal metastasis by MRI spine. Compared with the ‘gold standard’, derived from clinical and radiological follow-up, the sensitivities for radionuclide bone scan and that for MRI spine for detecting skeletal metastasis were 71.4 and 85.7%, respectively (P=0.023), whereas the specificities were 96.5 and 97.7%, respectively (P=0.95). Of the 34 individual metastatic lesions in the spine, 15 were concordantly positive on both scans, whereas five lesions were positive only by bone scan and 11 positive only by MRI. The addition of MRI spine in the staging for prostate cancer resulted in a change of stage and management plan in seven (7%) patients.
Conclusion
MRI spine has comparable specificity and slightly better sensitivity than bone scan to detect spinal metastasis from prostate cancer.
Key words: Magnetic resonance imaging, metastasis, prostate, spine, staging
Introduction
The single most important determinant of treatment options for prostate cancer is the accurate assessment of disease extent, especially skeletal metastasis [1]. The most frequent sites of metastasis from prostate cancer are the bones and commonly to the vertebrae 2, 3. Technetium 99m methylene diphosphonate (MDP) radionuclide bone scan is considered the investigation of choice to detect skeletal metastasis in staging patients with prostate cancer. Preliminary studies with small patient numbers have suggested that magnetic resonance imaging (MRI) seems to be more sensitive and probably more specific than bone scan for detecting skeletal metastasis in the spine 4, 5, especially in patients with prostate cancer 4, 6, 7. MRI may also be helpful in clarifying equivocal findings on radionuclide scan 2, 7, 8. Studies have suggested that staging with MRI spine may be useful in patients with high clinical suspicion of skeletal metastasis, especially if radionuclide bone scan is non-contributory, to plan optimal management by improving the selection of patients for curative management 2, 6.
The aim of our study was to evaluate the utility of MRI spine in detecting skeletal metastasis and in staging patients with high-risk locally advanced prostate cancer, in comparison with radionuclide bone scan.
Materials and Methods
A retrospective review of the case notes and images of 99 consecutive prostate cancer patients who reported to the Urology Unit of the Royal Marsden Hospital between January 2001 and October 2005 were analysed, if they were not on systemic treatment and satisfied the criteria of ‘high risk for skeletal metastasis’ characterised by composite Gleason score (Gl)≥8 or initial prostate-specific antigen (PSA)>10ng/ml or equivocal findings on radionuclide bone scan. Patients who had been on hormonal treatments for 2 weeks or more prior to the MRI were excluded from the study. The study was approved by the institutional Research Ethics committee. All patients reviewed in the study had radionuclide bone scan and MRI of the spine within an interval of less than 4 weeks.
The characteristics of the 99 patients included in the study are shown in Table 1.
Table 1. Patient characteristics
| Number | Range/percentage | |
|---|---|---|
| Number of patients | 99 | |
| Median age | 66 years | 44–83 years |
| T status (AJCC TNM) | ||
| 1 | 9 | 9% |
| 2 | 19 | 19% |
| 3 | 62 | 62% |
| 4 | 7 | 7% |
| x | 2 | 2% |
| Median initial PSA | 26.5ng/ml | 2–1600ng/ml |
| Gleason score | ||
| 3+3 | 17 | 17% |
| 3+4 | 18 | 18% |
| 4+3 | 16 | 16% |
| 4+4 | 16 | 16% |
| 4+5 | 23 | 23% |
| 5+5 | 7 | 7% |
| Median time between scans | 7 days | |
| Median follow-up | 1140 days | 384–1749 days |
Imaging Techniques
Radionuclide bone scans were carried out after intravenous administration of technetium 99m MDP. Total body images were obtained from 3h after injection and supplemented by spot views of the spine if indicated. MRI examination of the spine was carried out with a 1.5 Tesla system. The MRI scans of the spine were the sagittal T1- and STIR-weighted images. Thirty-nine patients in our study had whole body MRI with coronal STIR and coronal T1 images, with sagittal images of the spine. Radiographs and computed tomography of individual sites were carried out as indicated.
Image Analysis
On MRI, a skeletal lesion of high or low signal intensity on STIR images and low signal intensity on T1-weighted images was considered suspicious of metastases.
Areas of intense abnormal uptake of technetium 99m MDP on bone scan that could not be definitely explained by benign abnormalities, e.g. degenerative disease, were considered as suspicious for metastases. An area of uptake on bone scan was considered equivocal or indeterminate when the uptake could not be confidently categorised as either malignant or benign.
Metastatic lesions in the spine identified on radionuclide bone scan and MRI were tabulated according to the individual site of the lesion and spinal region of interest. The regions of the spine were defined as (1) cervical spine, (2) dorsal spine and (3) lumbosacral spine. Images of each modality were analysed independently by either the MRI radiologist or radionuclide imaging radiologist. The radiologists were unaware of the clinical and biochemical findings or of any other imaging studies concerning the patients, except the fact that they had prostate cancer. A patient was considered to have negative imaging when no bony lesions in the spine meeting the criteria of malignancy were identified on either imaging. The findings on MRI of the spine and the spinal region on radionuclide bone scan were compared to each other and also with the ‘gold standard’ arrived at from the study MRI and bone scan, other imaging including targeted X rays or computed tomography for correlation of equivocal findings, and subsequent follow-up radionuclide bone scans or MRI carried out to assess the response to treatment. For example, if patients showed a good biochemical and clinical response to hormonal therapy, reduction or resolution of lesions over a 6–24 month period, it was taken as an indication of the presence of metastatic disease.
Statistical Methods
The outcomes analysed were the radiological findings in individual patients and the number of individual metastatic skeletal lesions identified by each imaging modality. The specificity and sensitivity for the detection of skeletal metastasis was calculated for MRI spine and radionuclide bone scan independently, with reference to the ‘gold standard’. In patients with bone metastasis by ‘gold standard’, McNemar's chi-squared test was carried out to determine a difference in the sensitivity for the detection of skeletal metastasis between MRI and radionuclide bone scan. The analysis was carried out on a ‘per patient’ and ‘per region’ basis. The statistical analysis was conducted with SPSS 13.0 statistical software.
Results
Ten patients were detected to have definite spinal metastasis on radionuclide bone scan, whereas 12 patients had skeletal metastasis on MRI spine, as shown in Table 2. Compared with the ‘gold standard’ (14 patients with definite metastasis) arrived at by the two experts, the sensitivities for radionuclide bone scan and MRI spine for detecting skeletal metastasis were 71.4 and 85.7%, respectively (P=0.023), whereas the specificities were 96.5 and 97.7%, respectively (P=0.95).
Table 2. Predictive values of detecting spinal metastasis by radionuclide bone scan and magnetic resonance imaging (MRI) of the spine for individual patients compared with the ‘gold standard’ (14 patients had spinal metastasis)
| MRI of the spine | Spine metastasis on bone scan | |
|---|---|---|
| True positive | 12 | 10 |
| True negative | 83 | 82 |
| False negative | 2 | 4 |
| False positive | 2 | 3 |
| Sensitivity | 85.7% | 71.4% |
| Specificity | 97.65% | 96.5% |
| Positive predictive value | 85.7% | 76.9% |
| Negative predictive value | 97.65% | 95.4% |
Four patients with positive findings for skeletal metastasis on MRI spine had normal spine on radionuclide bone scan and two patients with spinal metastasis on bone scan had a normal MRI spine, whereas positive findings were concordant in eight patients. As shown in Table 3, 19 patients had indeterminate/equivocal findings in the spine on radionuclide bone scan that were not clarified on radiographs. These indeterminate findings were clarified as non-metastatic by MRI spine. Indeterminate MRI findings in two patients were clarified as benign on bone scan.
Table 3. Initial comparison of findings in the spine on radionuclide bone scan and magnetic resonance imaging (MRI) of the spine for individual patients
| MRI spine | Total | ||||
|---|---|---|---|---|---|
| Metastasis | Equivocal | No metastasis | |||
| Radionuclide bone scan | Metastasis | 8 | – | 5 | 13 (3 false positive) |
| Equivocal | – | – | 19 | 19 | |
| No metastasis | 6 | 2 | 59 | 67 | |
| Total | 14 (2 false positive) | 2 | 83 | 99 | |
Bone scan detected metastasis in 14 spinal regions, whereas MRI spine detected metastasis in 18 spinal regions. MRI could detect more metastatic lesions in the thoracic spine (seven) than radionuclide bone scan (four), whereas the detection rates in the other spinal regions were similar. Of the 34 individual metastatic lesions in the spine by the ‘gold standard’, 15 were concordantly positive on both bone scan and MRI spine, whereas five were positive only by radionuclide bone scan and 11 positive only by MRI spine.
None of the patients in the study had non-skeletal distant metastasis. There was a change of stage and altered management plan with the introduction of MRI spine in seven patients (7%) compared with staging by bone scan alone. Four patients were upstaged as metastatic (clinical features: T2, Gl 7 with PSA 158ng/ml; T4, Gl 9 with PSA 40ng/ml; T2, Gl 8 with PSA 215ng/ml and T2, Gl 9 with PSA 42ng/ml) and the bone scan findings in three patients (clinical features: T3, Gl 6 with PSA 38ng/ml, T2, Gl 7 with PSA 11ng/ml and T3, Gl 7 with PSA 25ng/ml) were re-classified as benign skeletal changes.
Discussion
We conducted this study in patients with prostate cancer who clinically had a high risk of metastasis, to examine whether MRI spine would improve the results achieved with radionuclide bone scan. In our study, MRI of the spine could detect more skeletal spinal metastatic lesions than bone scan in patients with prostate cancer. MRI spine clarified the diagnosis in 19 patients with indeterminate/equivocal bone scan findings. MRI spine was more sensitive for detecting spinal metastasis than bone scan with comparable specificity, suggesting that it may be a better imaging modality for staging patients with high-risk prostate cancer.
In previous studies with small patient numbers, MRI has been shown to detect more metastatic lesions in the spine than radionuclide bone scan in various malignancies [4]. In the study by Sauvage et al. [9] in non-small cell lung cancer, MRI of the spine was not found to be useful as a screening procedure before treatment, but offered advantages over radionuclide bone scan in symptomatic patients, when bone scan detected abnormal foci and provided a better analysis of the extent of metastatic involvement within vertebrae. In this study, the sensitivity of MRI to detect vertebral involvement (92%) was superior to that of radionuclide bone scan (67%) with similar specificity (94%) [9]. From our study, these findings might also be true for sclerotic bone metastasis from prostate cancer.
In a prospective study in advanced prostate cancer, Lecouvet et al. [6] have shown MRI of the axial skeleton to be more sensitive (100%) than any other imaging, including bone scan. Our results are not comparable with the impressive results reported by Lecouvet et al. [6], which could be due to the fact that our study focussed only on untreated patients who had imaging for initial staging, whereas this subgroup formed only 39% of patients in the study by Lecouvet et al. 6, 10. In the study by Lecouvet et al. [6], 42% (28/66) of patients were already on systemic treatment, where the utility of bone scan to assess response or to detect new metastatic skeletal lesions is limited. In this study, MRI helped modify the treatment plan in 33% of patients with advanced prostate cancer [6]. In our study, the contribution from MRI spine to accurate initial staging was only in 7% of patients, who had a change of stage or management plan.
Our study revealed instances with discordant findings on radionuclide bone scan and MRI spine, including false-positive findings on both imaging modalities. Gosfield et al. [4], in their comparative study of radionuclide bone scan and MRI in 35 patients with various malignancies, found the maximum proportion of discordant readings occurring in lumbar regions and more frequently in patients with prostate cancer. In our study, discordance was more in the thoracic region, although this may have been a random variation. Our results suggest that staging by either one of the two imaging modalities is not comprehensive and ideally it may be beneficial to have both MRI spine and bone scan for accurate staging of patients with high-risk prostate cancer.
The limitations of our study include the retrospective nature of the study and low patient numbers. The proportion of patients detected to have skeletal metastasis in our study was low, suggesting that future studies may have to include patients with a higher risk profile than those in this study, probably with a higher Gleason score, possibly Gl≥9 and a higher PSA cut-off, possibly≥40ng/ml. Fifty-five patients in our study belonged to this high-risk subgroup, of whom 10 had spinal metastasis. MRI could detect metastasis in all 10 patients, compared with metastasis detection in only six patients by bone scan. A selection bias may have occurred in our study by the fact that a higher proportion of patients with a normal or indeterminate bone scan were included in the study, as patients detected to have skeletal metastasis on initial bone scan were started immediately on hormonal treatment routinely by clinicians. Difficulty in obtaining MRI scans within weeks of the bone scan and before initiation of systemic treatment is a common scenario in busy hospitals and this delay may have resulted in a higher instance of patients with a negative bone scan and a positive MRI. A further bias in patient selection may have occurred in this study as patients with retained metallic foreign bodies and patients who were intolerant to MRI due to claustrophobia may have been excluded from the study.
MRI spine has a comparable specificity and a slightly better sensitivity than radionuclide bone scan to detect spinal metastasis from prostate cancer. MRI spine for staging may be beneficial in patients with equivocal bone scan findings and in patients at a very high risk, i.e. high composite Gleason score and high PSA. Although the results from this study seem promising, whether it would warrant inclusion of MRI spine as part of routine staging procedure needs to be ascertained by prospective studies with a larger population and a health economic analysis.
References
- . The role of computerized tomography, magnetic resonance imaging, bone scan, and monoclonal antibody nuclear scan for prognosis prediction in prostate cancer. Semin Urol Oncol. 1998;16(3):145–152
- . Preliminary results of a bone marrow magnetic resonance imaging protocol for patients with high-risk prostate cancer. Urology. 1999;54(1):118–123
- Detection of occult spinal cord compression with magnetic resonance imaging of the spine. Clin Oncol (R Coll Radiol). 2007;19(7):528–531
- . Comparison of radionuclide bone scans and magnetic resonance imaging in detecting spinal metastases. J Nucl Med. 1993;34(12):2191–2198
- Detection of vertebral metastases: comparison between MR imaging and bone scintigraphy. Radiographics. 1991;11(2):219–232
- Magnetic resonance imaging of the axial skeleton for detecting bone metastases in patients with high-risk prostate cancer: diagnostic and cost-effectiveness and comparison with current detection strategies. J Clin Oncol. 2007;25(22):3281–3287
- . Magnetic resonance imaging for the diagnosis of prostate cancer metastatic to bone. Br J Urol. 1995;75(1):54–58
- . Magnetic resonance imaging for detection of prostate cancer metastatic to bone. J Urol. 1993;149(6):1482–1484
- [MRI in the early diagnosis of spinal metastases of bronchial cancer], J Radiol. 1996;77(3):185–190
- . MRI or bone scan or both for staging of prostate cancer?. J Clin Oncol. 2007;25(36):5837–5838[author reply 5838–5839]
PII: S0936-6555(08)00401-9
doi:10.1016/j.clon.2008.09.006
© 2008 The Royal College of Radiologists. Published by Elsevier Inc. All rights reserved.
