Performed on behalf of the Dutch Bone Metastasis Study Group. The Dutch Bone Metastasis Study Group consists of a steering committee (Jan Willem H. Leer, Yvette M. van der Linden, Hans van Houwelingen, Job Kievit, Wilbert B. van den Hout, Hanneke de Haes, and Elsbeth Steenland) and coordinators at participating institutions (Hendrik Martijn [Catharina Hospital, Eindhoven, The Netherlands], Bing Oei [Dr. Bernard Verbeeten Institute, Tilburg, The Netherlands], Ernest Vonk [Radiotherapy Institute Stedendriehoek and Omstreken, Deventer, The Netherlands], Elzbieta M. van der Steen-Banasik [Arnhem Radiotherapy Institute, Arnhem, The Netherlands], Ruud G.J. Wiggenraad [Westeinde Hospital, Den Haag, The Netherlands], Jaap Hoogenhout [University Hospital Nijmegen, Nijmegen, The Netherlands], Carla C. Wárlám-Rodenhuis (University Medical Center, Utrecht, The Netherlands], Geertjan van Tienhoven [University Medical Center, Amsterdam, The Netherlands], Rinus Wanders [Limburg Radiotherapy Institute, Heerlen, The Netherlands], Jacqueline Pomp [Reinier de Graaf Hospital, Delft, The Netherlands], Matthijs van Reijn [Twente Medical Spectrum, Enschede, The Netherlands], Ineke van Mierlo [Daniel Den Hoed Cancer Center, Rotterdam, The Netherlands], Ewald Rutten [Medical Center Alkmaar, Alkmaar, The Netherlands], Jan Metsaars [Leyenburg Hospital, Den Haag, The Netherlands], Gerrit Botke [Radiotherapeutic Institute Friesland, Leeuwarden, The Netherlands], and Ben G. Szabó [University Medical Center, Groningen, The Netherlands]).
Adequate prediction of survival is important in deciding on treatment for patients with symptomatic spinal metastases. The authors reviewed 342 patients with painful spinal metastases without neurologic impairment who were treated conservatively within a large, prospectively randomized radiotherapy trial. Response to radiotherapy and prognostic factors for survival were studied.
The data base of the Dutch Bone Metastasis Study was used. Response to treatment and prognostic factors for overall survival (OS) were studied using a Cox regression model. A scoring system was developed to predict OS.
Responses were noted in 73% of patients. In 3% of patients, spinal cord compression was reported a mean of 3.5 months after randomization. The median OS was 7 months, and significant predictors for survival were Karnofsky performance score, primary tumor (multivariate analysis; both P < 0.001), and the absence of visceral metastases (multivariate analysis; P = 0.02). A scoring system based on these predictors was developed, and 34% of patients were in Group A (median OS = 3.0 months), 48% of patients were in Group B (median OS = 9.0 months), and 18% of patients were in Group C (median OS = 18.7 months). Group C was comprised of patients with breast carcinoma, a good performance, and no visceral metastases.
Patients with cancer frequently develop bone metastases in the spinal column. Back pain is often the sole symptom reported; however, when the tumor mass or bone fragments compress the spinal cord or nerve roots, concomitant neurologic symptoms may occur. In general, the prognosis for a patient with a disseminated cancer is limited, and treatment should be directed toward optimal palliation with minimal treatment-related morbidity.1–4 Generally, radiotherapy is the first choice of treatment, although a surgical intervention sometimes is needed. In 1986, Harrington developed a strategy for the treatment of metastases in the spine.1 He divided patients into five classes, depending on the extent of neurologic compromise or bone destruction (Table 1). Primary radiotherapy was recommended for patients with Class I–III spinal metastases, and primary surgical intervention was recommended for patients with Class IV and V spinal metastases. Harrington noted that secondary surgery should be considered in patients with pain or neurologic symptoms that were refractory to radiotherapy or when spinal cord tolerance to radiation had been reached.1, 4 Several surgical techniques have been developed, ranging from minimally invasive methods, such as palliative decompression by laminectomy, to extensive procedures, such as radical en-bloc resection. The choice of a surgical technique depends on expected survival, treatment-related morbidity, and outcome after treatment. In general, the more extensive the surgical technique, the more prolonged the palliative effect.1 However, Harrington commented that many patients with vertebral collapse or instability, even if they were associated with severe local compromise, did not have a projected life expectancy sufficient to warrant such major surgical interventions. Adequate prediction of survival, therefore, is crucial.
Table 1. Harrington Classification of Metastases to the Spinal Columna
Involvement of bone without collapse or instability
Major neurologic impairment (sensory or motor) without significant involvement of bone
Vertebral collapse with pain due to mechanical causes or instability but without significant neurologic compromise
Vertebral collapse with pain due to mechanical causes or instability combined with major neurologic impairment
We studied a large group of 342 conservatively treated patients with Harrington Class I and II painful spinal metastases who received radiotherapy within the prospectively randomized Dutch Bone Metastasis Study (DBMS) to determine the value of a single radiotherapy dose of 8 grays (Gy) versus 24 Gy in 6 fractions.5, 6 We analyzed response to radiotherapy and prognostic factors for survival, and we developed a scoring system to use as a guideline for the treatment of patients with Harrington Class I and II spinal metastases.
MATERIALS AND METHODS
Patient Selection and Follow-Up
Details of the patient population and study design for the DBMS have been published elsewhere.5, 6 Briefly, between March 1996, and September 1998, 1157 Dutch patients with painful bone metastases from solid tumors were randomized between a group that received a single radiotherapy fraction of 8 Gy (n = 579 patients) and a group that received 6 radiotherapy fractions of 4 Gy (n = 578 patients). The objective of the study was to prove the equal effectiveness of a single fraction versus multiple fractions, and the endpoint of the study was response to pain. To be eligible for the study, patients had to have a maximum pain score during the preceding week of at least 2 on an 11-point pain scale ranging from 0 (no pain) to 10 (worst imaginable pain). The bone metastases had to be confined to an area that could be encompassed in a single radiation treatment field. Patients were excluded from the study if their metastases already had been irradiated, if they had metastases in the cervical spine, or if they had a pathologic fracture or compression of the spinal cord (only patients with Harrington Class I and II lesions were included). Patients also were excluded if they had renal cell carcinoma or malignant melanoma, because it was expected that these diseases would respond differently to radiotherapy. The Medical Ethics Committees of all participating institutions approved the study, and all patients signed informed consent forms. After randomization, intensive follow-up with 13 weekly questionnaires and, afterward, 13 monthly questionnaires on pain, treatment side effects,7 quality of life, and analgesic consumption was carried out up to a maximum of 2 years or until death. Data managers in the participating hospitals collected data on all events, such as death, retreatment, and occurrence of a fracture or spinal cord compression. In December, 1998, the follow-up on survival and events in all patients was updated, and the study was closed (maximum follow-up, 32 months). For the current study, all 342 patients with a spinal metastasis were selected, which included 30% of all randomized patients.
The data base was analyzed using SPSS software (version 11.0 for Windows; SPSS Inc., Chicago, IL). Response to radiotherapy was calculated in alignment with the Bone Metastases Consensus Working Party Guidelines.8 The Kaplan–Meier method was used for survival analyses. The following patient characteristics were studied for their prognostic value for predicting survival: Karnofsky performance score,9 primary tumor, visceral involvement, solitary versus multiple bone metastases, and response to radiotherapy. The Cox proportional hazards model was used for univariate and multivariate analyses. A scoring system for predicting survival was developed based on the results of the multivariate analyses using Karnofsky performance score, primary tumor, and visceral involvement as prognostic factors. The points allocated for each entered variable were derived from the hazard ratios from the univariate analyses. The scores were added together to produce a prognostic score. All reported P values are based on 2-sided tests, with P values < 0.05 considered significant.
Patient Characteristics, Response to Treatment and Follow-Up
Of the 342 trial patients who had spinal metastases, 53% of patients were male, and 47% were female. Most patients had breast carcinoma (42%), prostate carcinoma (24%), or lung carcinoma (21%). Thirteen percent had tumors located at other sites (3% colorectal, 2% bladder, 1% esophagus, 3% other sites, and 4% unknown primary). At the time of randomization, 60% of patients presented with more than 1 bone metastasis. The mean patient age at randomization was 66 years (range, 34–90 years). The mean Karnofsky performance score was 70. Forty-eight percent of patients were randomized to receive a single fraction of 8 Gy, and 52% were randomized to receive to 24 Gy in 6 fractions.
For all patients, the median overall survival was 7 months (mean overall survival, 11 months; 95% confidence interval [95%CI], 10–12 months). At the end of the study period, 75% of the patients had died.
After treatment, 73% of patients responded with lesser pain, with no differences in response between the single-fraction regimen and the multiple-fraction regimen (P = 0.52). Toxicity 1 month after radiotherapy was scored in approximately 78% of patients who had spinal metastases. Patients reported no nausea or only mild nausea (74%), emesis (84%), itching (94%), and painful skin (96%). Severe nausea, emesis, itching, and painful skin were reported in 9.5%, 5.3%, 0.7%, and 0.8% of patients, respectively.
During follow-up, 11% of patients received a second treatment for recurrent or continuing pain or for compression of the spinal cord. In total, 12 spinal cord compressions were reported (3%) after a mean of 3.5 months after randomization (range, from 3 days to 15 months). Nine patients with spinal cord compression received a second course of radiotherapy, and 3 patients were given best supportive care. Neurologic outcomes after second radiotherapy treatments were not reported in the follow-up questionnaires. No patient underwent surgery for spinal cord compression. After a diagnosis of spinal cord compression, median overall survival was 1 month (mean overall survival, 4 months; 95%CI, from 2 weeks to 8 months).
Prognostic Factors for Survival
Prognostic factors for survival were studied in the 342 trial patients (Table 2). Patients who had Karnofsky performance scores between 80 and 100 had a significantly prolonged survival compared with patients who had lower performance scores. Surprisingly, patients who had Karnofsky performance scores between 50 and 70 did not have a better prognosis compared with patients who had scores of only 10–40: Their median overall survival was 4.6 months versus 3.8 months, respectively (P = 0.55). Patients with breast carcinoma and prostate carcinoma had the best overall survival, with a median survival of 14.1 months and 9.2 months, respectively (P < 0.001). Twenty-five percent of patients had visceral metastases at the time of randomization. Their overall survival was significantly worse compared with patients who had no visceral metastases, with a median survival of 4.5 months versus 8.1 months, respectively (P < 0.001). The presence of other bone metastases in 60% of patients was not an adverse factor for survival (median overall survival, 7.6 months in patients with other bone metastases vs. 5.5 months in patients without other bone metastases; P = 0.23). Most patients with other bone metastases had breast carcinoma (50%) or prostate carcinoma (28%), and 74% of those patients received concomitant systemic therapy. Patients whose pain responded to radiotherapy had a better survival compared with nonresponders (median survival, 8.1 months vs. 3.4 months; P < 0.001). Figure 1 shows the survival curves for the different prognostic factors. The significantly predictive factors from the univariate analyses were tested in a multivariate analysis: Karnofsky performance score, primary tumor, and visceral involvement remained significantly predictive for survival (P < 0.001, P < 0.001, and P = 0.02, respectively). We excluded response to radiotherapy from the multivariate analysis to make the scoring system suitable for each patient who presented with a spinal metastasis, i.e., even for patients who had not yet received radiotherapy.
Table 2. Prognostic Factors for Predicting Survival in 342 Patients with Spinal Metastases Treated on the Dutch Bone Metastasis Study
Based on the outcomes of the univariate and multivariate analyses for survival, a scoring system for the prediction of survival was developed that included the Karnofsky performance score, primary tumor, and visceral involvement. The scoring system was structured as follows (Table 3): For Karnofsky performance score, 2, 1, and 0 points were allocated to scores of 80–100, 50–70, and 10–40, respectively; For primary tumor, 3, 2, 1, and 0 points were allocated to breast carcinoma, prostate carcinoma, lung carcinoma, and other types of malignancies, respectively; For visceral involvement, 1 and 0 points were allocated to absence and presence of visceral metastases, respectively.
Table 3. Design of a New Scoring System for Predicting Survival in 342 P with Spinal Metastases Treated on the Dutch Bone Metastasis Study
KPS: Karnofsky performance score.
The KPS is a conditional score ranging from 0% (= death) to 100% (normal situation, no complaints).
There was a minimum of 0 total points and a maximum of 6 total points. Three prognostic groups were formulated: Group A, with total scores of 0–3; Group B, with total scores of 4–5; and Group C, with total scores = 6. Table 4 lists the distribution of the 342 patients into the 3 prognostic groups, and Figure 2 shows the survival curves. The median overall survival was 3.0 months in Group A, 9.0 months in Group B, and 18.7 months in Group C. At the end of the 32-month study period, 95% of patients in Group A and 70% of patients in Group B had died. In Group C, 47% of patients remained alive. Surviving patients had a mean follow-up of 16 months (range, 3–32 months).
Table 4. Predicting Survival Using the New Scoring System in 342 Patients with Spinal Metastases Treated in the Dutch Bone Metastasis Study
A Cox proportional hazards model for univariate analysis was used to determine P values, HRs, and 95% CIs. The prognostic factors used were Karnofsky performance score, the type of primary tumor, and visceral involvement.
The current study showed that the median overall survival of patients in the DBMS who had Harrington Class I and II spinal metastases was limited (only 7 months). Noninvasive radiotherapy provided adequate palliation in these patients, with 73% of patients responding to pain and only 3% of patients reporting spinal cord palsies during follow-up. In addition, we were able to put together a scoring system in which survival was estimated accurately using the Karnofsky performance score, the type of primary tumor, and visceral involvement as prognostic factors for survival.
In general, the treatment of patients with symptomatic spinal metastases should be directed toward optimal palliation and a minimum of treatment-related morbidity. When deciding on treatment, physicians often refer to the scoring systems of Tokuhashi et al., Enkaoua et al., and Tomita et al. as guidelines for choosing the type of treatment to the spine.10–12 These scoring systems used retrospective surgical data and were based on survival in a limited number of patients with lesions varying from Harrington Class I to Class V (Table 1). Table 5 shows the three scoring systems: Points were allocated to a number of prognostic factors and were added up to produce a prognostic score. Mean overall survival varied between 3 months and 50 months (Table 6). In the studies by Enkaoua et al. and Tomita et al., as many as 50% of patients had a mean overall survival ≥ 24 months. In their discussions, the authors of all three scoring systems recommended that surgery should be undergone not only by the expected patients with a good prognosis but also by patients with a limited survival. Surprisingly, those authors suggested an important role for surgery in the primary treatment of patients with Class I–III metastases. The value of palliative radiotherapy for these patients was not discussed. Compared with the current study, their reported overall survival was much longer, probably due to the selection of patients for surgical treatment. Consequently, when these scoring systems were to be applied on every new patient who presented with a spinal metastasis, as those authors recommended, patients with a limited life expectancy would be assigned to undergo a surgical procedure with associated morbidity and even mortality. We believe the survival data and scoring system presented in the current report provide a more realistic indication of prognosis for patients who have spinal metastases.
Table 5. Design of Scoring Systems for the Surgical Treatment of Spinal Metastases Based on Life Expectancy
Table 6. Prognostic Groups of the Scoring Systems by Tokuhashi et al., Enkaoua et al., and Tomita et al. in Relation to Survival
Scoring system/prognostic group
No. of patients (%)
Mean OS in mos (range)
Suggested surgical method
OS: overall survival; SD: standard deviation.
a For the scoring systems of Tokuhashi et al. and Enkaoua et al., the prognostic factors used were the performance status, the number of extraspinal metastases, the number of metastases within the spinal body, the presence of visceral metastases, the primary tumor site, and the presence of spinal cord palsy. For the scoring system of Tomita et al., the prognostic factors used were the number of extraspinal metastases, the presence of visceral metastases, and the primary tumor site.
A remarkable finding in the scoring system reported by Tomita et al. was that the performance status of the patient was not incorporated into the scoring system (Table 5). Performance status generally is regarded as one of the strongest prognostic factors for survival, correlating with the ability to undergo medical treatment.13 In 2001, Chow et al. published a review on physicians' capabilities of predicting survival and concluded that the Karnofsky performance score was the most important factor.14 Although the Karnofsky performance score was one of the major factors predicting survival in the current study, when it was combined with the primary tumor type and the presence of visceral metastases, we were able to refine further the prediction of survival.
Obviously, along with survival, the expected treatment outcome is an important factor when deciding on treatment. Radiotherapy for patients with Harrington Class I–II metastases has been studied extensively. Several prospectively randomized trials and 2 recent meta-analyses on palliative radiotherapy in painful bone lesions, including spinal metastases, reported decreases in pain of 60–80% of patients.5, 6, 15–29 In patients with Class III metastases, a decrease in symptoms after radiotherapy doses of 16–24 Gy was reported in 10–90% of patients, depending on the severity of the pretreatment neurologic symptoms.30–34 Of the 3 reports on surgical scoring systems, Tokuhashi et al. did not discuss treatment outcomes, although 80% of their 64 patients had neurologic symptoms before surgery.10 Enkaoua et al. reported 79% relief of pain in 71 patients.11 In 25 patients with neurologic symptoms, the neurologic status improved in 14% of patients postoperatively, remained unchanged in 72% of patients, and deteriorated in 14% of patients. Tomita et al. only reported treatment outcomes in a second group of 61 patients who were treated prospectively in line with the proposed surgical strategy.12 Similar overall survival rates were seen compared with the original 67 patients, and 78% of patients had less or no pain postoperatively. In the 65% patients with neurologic symptoms before surgery, neurologic improvement of at least 1 Frankel grade35 was seen in 75% of patients.
A number of other reports on patient outcomes after spinal surgery have been published.36–40 Chataigner and Onimus studied 107 patients retrospectively who underwent surgery for Class I and II spinal metastases36 and reported improvements in pain of 98% postoperatively, but they also reported a 10% postoperative mortality rate. Those authors allocated their patients to the Tokuhashi prognostic groups and observed a mean overall survival of 2.0 months for Group A, 9.5 months for Group B, and 8.0 months for Group C. In another study, Hatrick et al. reported a limited median overall survival of < 5 months in 42 patients who underwent surgery after failure on radiotherapy.37 In that study, pain improved in 90% of the patients, and neurologic symptoms improved in 69% of patients. Hirabayashi et al. studied the medical records of 81 patients who underwent palliative spinal surgery.38 Fifty patients were nonambulatory preoperatively. After surgery, 70% of those patients were ambulatory; they had a median overall survival of 16.5 months and a median ambulation time of 13.8 months. Factors that influenced survival significantly survival were primary tumor (bone marrow, prostate carcinoma, or thyroid carcinoma vs. other sites) and postoperative ambulation (multivariate analysis [MV]; P < 0.001). The authors called for caution, because their study was restricted by its retrospective design: Patients who were selected for surgery generally were in better condition or had limited systemic disease.
Recently, at the 2003 annual meeting of the American Society for Therapeutic Radiation Oncology, Regine et al.41 presented the first results of a prospectively randomized trial on surgery plus radiotherapy versus radiotherapy alone for the treatment of spinal cord compression. Their patients had signs of spinal cord compression on magnetic resonance imaging scans either with or without neurologic symptoms. Regine et al. showed that 50 patients who were treated with surgery plus radiotherapy retained the ability to walk significantly longer than 51 patients who were treated with radiotherapy alone (median, 126 days vs. 35 days; P = 0.006). Unfortunately, their data have not yet matured.
Compared with the three surgical scoring systems, there are some remarks to be made about the current study. First, no patient with a spinal metastasis within the DBMS had signs of spinal cord compression at the time of randomization.5 During follow-up, only 3% spinal cord palsies were reported. Although the scoring systems of Tokuhashi et al.10 and Enkaoua et al.11 incorporated the degree of spinal cord palsy, the system proposed by Tomita et al.12 left it out. Like Tomita et al., we believe that symptoms of spinal cord palsy merely suggest the anatomic localization of the metastasis and the voluminous extent of the metastatic lesion and are not independent adverse prognostic factors. Second, patients with renal cell carcinoma, melanoma, and multiple myeloma were excluded from randomization because of the expected different biologic behavior of those entities. Therefore, no conclusions can be drawn from the current study for patients with those tumors. Finally, 48% of patients in the proposed scoring system Group C remained alive at the end of follow-up, with a mean follow-up of 16 months. If follow-up had been extended beyond the maximum of 32 months, then it is possible that the mean survival in these patients would have been even greater than 18 months.
In conclusion, we propose a scoring system for the prediction of survival in patients with Harrington Class I and II spinal metastases. We believe that radiotherapy is the primary treatment of choice in all patients with Class I–II lesions. Surgery should be considered only when pain is persistent despite radiotherapy or when spinal cord tolerance after radiotherapy has been reached; however, treating physicians should be aware of the limited life expectancy in the majority of patients. In patients with Class I and II lesions, extensive surgery, if any, must be reserved for those who have an expected good prognosis, perhaps for patients in the proposed scoring system Group C, i.e., those with primary breast carcinoma who have a good performance status and spinal metastasis without visceral involvement.