To optimize selection of a radiotherapy schedule for patients with spinal metastases, the authors analyzed prognostic factors and developed a scoring system to predict survival in such patients.
Wiley Online Library is migrating to a new platform powered by Atypon, the leading provider of scholarly publishing platforms. The new Wiley Online Library will be migrated over the weekend of February 24 & 25 and will be live on February 26, 2018. For more information, please visit our migration page: http://www.wileyactual.com/WOLMigration/Remove maintenance message
To optimize selection of a radiotherapy schedule for patients with spinal metastases, the authors analyzed prognostic factors and developed a scoring system to predict survival in such patients.
Five-hundred forty-four patients with spinal metastases received radiotherapy at Shizuoka Cancer Center Hospital between September 2002 and November 2006. Prognostic factors for survival were studied using a Cox proportional hazards model, and a scoring system was developed based on regression coefficients: Three points were given for an unfavorable tumor type and bad performance status (≥3); 2 points were given for hypercalcemia, visceral metastases, and previous chemotherapy; and 1 point was given for multiple bone metastases and age ≥71 years.
The overall survival rates after 6 months, 12 months, and 24 months were 49%, 32%, and 19%, respectively, and the median survival was 5.9 months (95% confidence interval, 4.9-6.8 months). In total, 503 patients (93%) were followed for ≥12 months or until death. These patients were separated into Groups A, B, and C based on scores of 0 to 4, 5 to 9, and 10 to 14, respectively. These groups included 24%, 57%, and 19% of patients, respectively; and the mean median survival for Groups A, B, and C was 27.1 months, 5.4 months, and 1.8 months, respectively. Overall survival rates after 6 months, 12 months, and 24 months were 89%, 77%, and 54% in Group A; 46%, 22%, and 9% in Group B; and 7%, 4%, and 0% in Group C, respectively (P < .001).
The scoring system was able to predict the survival of patients with spinal metastases and may be useful for selecting an optimal radiotherapy schedule. Cancer 2008. © 2008 American Cancer Society.
Bone metastasis is a common cause of cancer-related pain and neurologic disturbance. Palliative radiotherapy for painful bone metastases is well established,1–3 and a total dose of 30 grays (Gy) in 10 fractions is considered a standard and effective schedule. In recent years, several randomized trials have produced a similar outcomes between long-course radiotherapy (30 Gy in 10 fractions) and short-course radiotherapy (8 Gy in 1 fraction, 20 Gy in 5 fractions).4–10 A short course is recommended if survival is expected to be relatively short, because this approach reduces the treatment period and associated costs.11, 12 However, in-field recurrence rates are higher for short-course radiotherapy; therefore, a longer course is recommended if relatively favorable survival is expected.9, 10
Evaluating prognostic factors for survival is difficult in patients who have metastases to the spinal column. Prognostic factors for life expectancy and a scoring system for predicting survival in such patients have been discussed in a few studies,13–15 with the site of the primary lesion, performance status (PS), presence of visceral metastases, previous chemotherapy, multiple skeletal metastases, interval from tumor diagnosis to metastatic spinal cord compression (MSCC), and ambulatory status before radiotherapy identified as prognostic factors for survival. In our hospital, we select an optimal radiotherapy schedule according to the prognostic scoring system described by Katagiri et al13 in which 3 points are assigned for an unfavorable tumor type, 2 points are assigned for visceral metastases, and 1 point is assigned for multiple bone metastases, bad PS, and previous chemotherapy. According to this system, the 1-year survival rate is 11% for patients with total scores ≥6 and 89% for patients with total scores ≤2. However, these results included patients who received both radiotherapy and surgery, and the survival periods were calculated from the time when bone metastases were diagnosed. In this report, we focus on survival from the time radiotherapy was administration. On the basis of the results, we developed a new scoring system for predicting the prognosis of patients who received radiotherapy to allow the selection of optimal radiotherapy schedules.
Five-hundred forty-four patients with metastases in the spinal column received radiotherapy at our hospital between September 2002 and November 2006. The patients comprised 287 men and 257 women, and their median age was 63 years (range, 19-94 years). The primary tumor sites were as follows; lung, 151 patients (28%); breast, 115 patients (21%); gastrointestinal tract, 94 patients (17%); prostate, 34 patients (6%); liver, 24 patients (4%); and others, 126 patients (23%). Sixty-nine patients had an Eastern Cooperative Oncology Group (ECOG) PS of 0, 141 patients had an ECOG PS of 1, 121 patients had an ECOG PS of 2, 133 patients had an ECOG PS of 3, and 80 patients had an ECOG PS 4 (Table 1).
|Characteristics||No. of Patients||%|
|Median age (range)||63 (19-94)|
|Distribution of primary tumor|
Irradiation was performed with 4- to 18-MV photons by using linear accelerators and was delivered mainly through a single posterior field or through parallel opposed fields. The radiation dose was delivered to the center of the spinal cord in a radiation port using a 3-dimensional treatment-planning system. The treatment volume usually encompassed 1 normal vertebra above and below the metastatic lesion. The radiation schedule was determined after a conference with the primary physician and orthopedists and was based on predictable prognosis and the purpose of the radiotherapy, which was mainly to relieve pain or to improve/prevent paralysis. In general, short-course radiotherapy was chosen for patients who had poor predicted survival, and long-course radiotherapy was selected for patients with good predicted survival or for the improvement/prevention of paralysis. Different radiation schedules were compared using the equivalent dose in 2-Gy fractions.16, 17 Regarding the schedules, 314 patients received 30 Gy in 10 fractions, 111 patients received 40 Gy in 20 fractions, 77 patients received 20 Gy in 5 fractions, 6 patients received 8 Gy in 1 fraction, and 36 patients received a schedule of >20 Gy in 4 fractions.
The database was analyzed using SAS software (SAS, Chicago, Ill). The log-rank test was used to test the difference between survival curves for each factor. The following patient characteristics were studied for their prognostic value for predicting survival: age, sex, ECOG PS, tumor type, presence of visceral metastases, presence of multiple bone metastases, previous chemotherapy, previous radiotherapy for other bone metastases, total serum calcium corrected for albumin level, interval from tumor diagnosis to radiotherapy, and the presence of paralytic symptoms. First, we fit the Cox proportional hazards model to the data and included all factors. On the basis of the results, the pairs of factors with more significant associations were picked up. Of the factors in those pairs, we eliminated nonsignificant factors from the model. The significant level was set at a 2-sided 5%. Next, we fit the Cox proportional hazards model to the data, including selected factors. If a factor had a P value > .05, then the prognostic score was allocated zero points. If P < .05, then the k-means clustering method was applied to the absolute values of estimated regression coefficients divided by their standard error to classify patients into 3 groups. The prognostic scores of 3, 2, and 1 were allocated, in the order of descending mean statistics in the 3 groups. The prognostic score for every patient was calculated by adding the allocated scores for each significant prognostic factor.
The following potential prognostic factors were evaluated with respect to overall survival: age (≤70 years vs s ≥ 71 years), sex, ECOG PS (0-2 vs 3-4), tumor type (favorable vs unfavorable; where breast cancer, prostate cancer, lymphoma/myeloma, and thyroid cancer (except anaplastic thyroid cancer) are considered favorable, because the median survival for patients with these tumors is expected to be >24 months18–24), the presence of visceral metastases, the presence of multiple bone metastases, previous chemotherapy, previous radiotherapy for other bone metastases, total serum calcium corrected for albumin level (mg/dL; experience indicates that hypercalcemia associated with cancer is poor prognostic factor), interval from tumor diagnosis to radiotherapy (<10 months vs ≥10 months; 10 months was the median for the cohort), and the presence of paralytic symptoms. Age was divided into 63 years, which was the median for the cohort, and 70 years (25% of the cohort was aged ≥71 years). Of these 2 ages, age 70 years showed a more significant difference, so we selected age 70 years as the point of division. Radiotherapy schedule was excluded from this analysis, because the schedule was selected based on survival predicted by the Katagiri et al scoring system. Survival was analyzed using the Kaplan-Meier method,25 and overall survival was measured from the first day of radiotherapy until death from any cause.
Univariate analysis and multivariate analysis with Cox regression models were performed using SAS software. In univariate analysis of each potential prognostic factors, unfavorable tumor type, elevated calcium (>10 mg/dL), poor PS, previous chemotherapy, visceral metastases, multiple bone metastases, older age, sex (men), and the presence of paralytic symptoms were associated with poor survival (Table 2).
|Variable||No. of Patients||MST, mo||1-Year Survival, %||2-Year Survival, %||P|
|Primary tumor type|
|Multiple bone metastases|
|Previous radiotherapy for other bone metastases|
|Total serum calcium corrected for albumin level, mg/dL*|
|Interval from tumor diagnosis to radiotherapy, mo†|
Initial multivariate analysis indicated that the pairs of factors that had the strongest association with survival were previous chemotherapy and visceral metastases, interval from tumor diagnosis to radiotherapy and tumor type, paralytic symptoms and PS, and sex and tumor type. For the first pair, both previous chemotherapy and visceral metastases were significant; for the second pair, only tumor type was significant; for the third pair, only PS was significant; and, for the fourth pair, only tumor type was significant. On the basis of these results, sex, paralytic symptoms, and interval from tumor diagnosis to radiotherapy were removed from the model. Then, we fit the Cox model to the patient data using the selected factors (Table 3). In this multivariate analysis, unfavorable tumor type, elevated calcium, poor PS, previous chemotherapy, visceral metastases, multiple bone metastases, and older age were associated strongly with poor survival, with hazard ratios of 3.3, 2.6, 2.2, 1.8, 1.7, 1.6, and 1.4, respectively.
|Variable||Regression Coefficient||SE||Chi-Square||P||HR||95 % CI|
|Age, y (≤70=0, >70=1)||0.35980||0.11642||9.5506||.002||1.433||1.14-1.80|
|ECOG PS (0-2=0, 3-4=1)||0.78429||0.10315||57.8090||<.001||2.191||1.79-2.68|
|Primary tumor (favorable=0, unfavorable=1)||1.20192||0.12141||98.0027||<.001||3.326||2.62-4.22|
|Multiple bone metastases*||0.49906||0.17440||8.1888||.004||1.647||1.17-2.32|
|Serum calcium level (normal=0, elevated=1)||0.94376||0.16185||34.0035||<.001||2.570||1.87-3.54|
A scoring system was developed based on regression coefficients in the multivariate analyses (Table 4). In this system, 3 points are allocated to an unfavorable tumor type and bad PS (≥3); 2 points is allocated for hypercalcemia, visceral metastases, and previous chemotherapy; and 1 point is allocated for multiple bone metastases and older age (≥71 years). For example, a patient with lung cancer (3 points) who has visceral metastases (2 points) and multiple bone metastases (1 point), aged 74 years (1 point), who has a normal serum calcium level (0 points) and a good PS (0 points) and had not received previous chemotherapy (0 points) would have a total score of 7 points (3 + 2 + 1 + 1 + 0 + 0 + 0 = 7). This score is associated with a 1-year survival rate after radiotherapy of 19%. Overall survival rates for each score are shown in Figure 1.
|Type of primary tumor|
|ECOG PS ≥3||3|
|Multiple bone metastases||1|
|Elderly (≥71 y)||1|
Of 544 patients, 503 (93%) were followed for a minimum of 12 months or until death; and, during that time, 41 patients (7%) were lost to follow-up. The median follow-up for survivors was 17.5 months. The overall survival rates after 6 months, 12 months, and 24 months were 49% (95% confidence interval [CI], 45%-53%), 32% (95% CI, 28%-36%), and 19% (95% CI, 15%-22%), respectively, and the median overall survival was 5.9 months (95% CI, 4.9-6.8 months).
The classification into good, moderate, and poor prognostic groups was conducted using log-rank test statistics (Fig. 2); that is, this classification was data-driven. Patients in Group A (24% of the total population) had total scores from 0 to 4, patients in Group B (57% of the total population) had total scores from 5 to 9, and patients in Group C (19% of the total population) had total scores from 10 to 14. The median survival was 27.1 months, 5.4 months, and 1.8 months for Group A, B, and C, respectively; and the overall survival rates after 6 months, 12 months, and 24 months were 89%, 77%, and 54%, respectively, in Group A; 46%, 22%, and 9%, respectively, in Group B; and 7%, 4%, and 0%, respectively in Group C (P < .001).
Several randomized trials have demonstrated that schedules of 8 Gy in 1 fraction, 20 Gy in 5 fractions, 30 Gy in 10 fractions, and 40 Gy in 20 fractions produce similar clinical outcomes.4–10 However, some reports indicate that in-field recurrence is more frequent in short-course radiotherapy (20 Gy in ≤5 fractions) than in long-course radiotherapy (>30 Gy in 10 fractions). Rades et al reported 1-year in-field recurrence-free rates after radiotherapy of about 74% and 90% in short- and long-course radiotherapy, respectively,10 and recommended short-course radiotherapy for patients with shorter predicted survival and long-course radiotherapy for other patients. However, the background of patients with metastases to the spinal column is very complicated, and this makes it difficult to predict an accurate prognosis. Katagiri et al derived a scoring system for survival from a diagnosis of bone metastases,13 with site of the primary lesion, visceral metastases, PS, previous chemotherapy, and multiple skeletal metastases included as prognostic factors for survival. Our results generally are consistent with that study, because we also observed that the primary lesion site, PS, visceral metastases, previous chemotherapy, and multiple bone metastases were prognostic factors. We also identified age and total serum calcium corrected for albumin level as new prognostic factors in our analysis. van der Linden et al studied the prognosis of 342 patients who received radiotherapy for metastases in the spinal column14 and observed that the primary lesion site, PS, and visceral metastases were prognostic factors. Primary lesion site, PS, and visceral metastases also were used as prognostic factors in our study. However, multiple bone metastases were not identified as a significant prognostic factor in the study by van der Linden et al, in contrast to the results reported by Katagiri et al and our results. This inconsistency may be because of bias in selection of the sites of primary lesions. Because, in the study by van der Linden et al, only patients with Harrington Class I or II disease (no significant neurologic involvement or involvement of bone without collapse or instability) who had no compression of the spinal cord26 were included, and patients with renal cell carcinoma and malignant melanoma were excluded (breast, prostrate, and lung cancers accounted for nearly 90% of all patients).
Rades et al have published several detailed studies of radiotherapy for patients with MSCC9, 10, 15, 27–31 and recently reported a new scoring system to predict overall survival in patients with MSCC32 based on the type of primary tumor, the interval between tumor diagnosis and MSCC, the presence of other bone metastases at the time of radiotherapy, visceral metastases at the time of radiotherapy, pretreatment ambulatory status (ambulatory and nonambulatory status are defined by an ECOG PS of 1-2 and 3-4, respectively), and the time to development of motor deficits before radiotherapy as prognostic factors for survival. The type of primary tumor, presence of other bone metastases at the time of radiotherapy, and visceral metastases at the time of radiotherapy also were used in our analysis. Compared with the report by Rades et al, the interval from tumor diagnosis to radiotherapy was not a prognostic factor in our study. This may be because of patient selection bias. Rades et al selected patients who had MSCC, whereas our cohort included patients who received radiotherapy for pain relief, and those with MSCC accounted for 133 of 544 patients (24%), indicating a difference between the patient populations of the 2 studies. In addition, we used PS as a prognostic factor rather than ambulatory status before radiotherapy, because PS is among the indispensable, important factors for selecting the optimal therapeutic strategy, such as chemotherapy, surgery, and radiotherapy, for patients with cancer. Multivariate analysis revealed a strong correlation between PS and the presence of paralytic symptoms; therefore, we used PS only as a prognostic factor, because it had a stronger influence on prognosis.
On the basis of the results from our study and as predicted by Katagiri et al, Rades et al, and van der Linden et al, primary lesion site, PS (Rades et al use ambulatory status before radiotherapy instead of PS), and visceral metastases were definite prognostic factors for survival. Previous chemotherapy was a significant prognostic factor in our analysis and according to Katagiri et al but was not included in the analyses of van der Linden et al and Rades et al, suggesting that additional investigation of this factor may be required. Furthermore, serum calcium level and age were not used in the previous studies; therefore, these factors also require further verification to confirm that they are prognostic factors for survival. These differences probably were caused by patient selection bias, as discussed above, or because the factors that were evaluated in each study were not same.
One problem in defining a scoring system is that a classification approach for the primary lesion has not been established. Several reports have suggested that breast cancer, prostate cancer, and lymphoma should be included in the good-prognosis group,9, 10, 13, 14, 32 because patients with these tumors are likely to have a more than 24-month median survival, even if they have distant metastases.19, 21, 24 Therefore, we included these cancers and thyroid cancer (except for anaplastic thyroid cancer) in the good-prognosis group.18–24
Rades et al also suggested an optimal radiotherapy schedule based on survival,32 and they recommended short-course radiotherapy for patients with 6-month and 1-year survival probabilities of <20% and <10%, respectively, because long-course radiotherapy did not appear to improve survival in these patients compared with a short course. In contrast, long-course radiotherapy was recommended for patients with 6-month and 1-year survival probabilities of >90% and >70%, respectively. For patients with a 1-year survival probability <25% (an intermediate zone between poor survival and good survival), there was no significant difference between short-course and standard-course radiotherapy; therefore, short-course radiotherapy was recommended. By using these criteria in our analysis, short-course radiotherapy is recommended for patients in Group C, who had total scores of 10 to 14 in our scoring system and 6-month and 1-year survival rates of 7% and 4%, respectively. In contrast, long-course radiotherapy is recommended for patients in Group A, who had total scores of 0 to 4 and 6-month and 1-year survival rates of 89% and 77%, respectively. Short-course radiotherapy may be recommended for patients in Group B, who had total scores of 5 to 9 and a 1-year survival rate of 22% (ie, <25%). Because dose-fractionation radiotherapy schedules were used arbitrarily in this study, the number of patients who received short-course radiotherapy was small, which made it difficult to conduct an effective analysis of radiologic therapeutic schedules. Thus, to avoid misjudgment, we decided to exclude these schedules from the analysis. Further investigation is needed to determine the optimal radiotherapy schedule for survival in patients with spinal metastases and to improve the accuracy of the scoring system.
In conclusion, our scoring system predicts the survival of patients with spinal metastases. Further prospective studies using this system will be needed to determine the optimal radiotherapy schedule appropriate for survival.