Cyberknife radiosurgery for breast cancer spine metastases

A matched-pair analysis




There are few options for breast cancer patients with spinal metastases recurrent within a previous radiation treatment field. CyberKnife radiosurgery has been used in our institution to treat such patients. To evaluate their outcomes, as there are no comparable radiation treatment options, the outcomes were compared between 18 patients with spinal metastases from breast cancer treated with CyberKnife stereotactic radiosurgery, 17 of which had prior radiotherapy to the involved spinal region and were progressing, and 18 matched patients who received conventional external beam radiotherapy (CRT) up-front for spinal metastases.


Radiosurgery was delivered in 3 to 5 fractions to doses ranging from 2100 to 2800 cGy. Women were matched to patients in a CRT group with respect to time from original diagnosis to diagnosis of metastases, estrogen receptor / progesterone receptor (ER/PR) status, presence or absence of visceral metastases, prior radiotherapy, and prior chemotherapy. Survival and complications were compared between treatment groups. Surviving patients were followed out to 24 months.


The CyberKnife and CRT groups were comparable along all matching dimensions and in performance status before treatment. Outcomes of treatment were similar for patients in both groups; ambulation, performance status, and pain worsened similarly across groups posttreatment. Survival and the number of complications appeared to favor the CyberKnife group, but the differences did not reach statistical significance.


The statistical comparability of the CyberKnife and CRT groups reflects the small sample size and stringent requirements for significance of the matched-pair analysis. Nevertheless, comparability in these difficult cases shows that salvage CyberKnife treatment is as efficacious as initial CRT without added toxicity. Cancer 2007. © 2007 American Cancer Society.

Metastatic breast cancer is marked by a propensity for bony sites.1, 2 The prevalence of bone metastases from breast cancer is high, with an estimated 217,440 patients suffering this sequelum each year, with a predilection for the axial skeleton.3 Up to 75% of patients have bone metastases on autopsy, and well over half of these involve the vertebrae.4 Spinal metastases are associated with significant symptoms, considerable neurologic risk, and further morbidity related to alteration of the structural, weight-bearing, and hematopoietic functions of the spine,1, 5 and are responsible for a large part of the decrease in quality of life experienced by breast cancer patients with metastatic disease. Unfortunately, treatment options for the spine are limited, largely due to the proximity of neighboring dose-limiting structures, including the spinal cord, nerve roots, peripheral nerves, esophagus, lungs, heart, kidneys, liver, and bowel.

There are few prospective studies of the palliative benefits of radiation therapy for treatment of the spine alone. In 1 study, Tombolini et al.6 treated 103 spinal sites from 95 patients with either a single fraction of 8 Gy, hypofractionation regimens of 20 Gy in 4–5 days, or conventional fractionation regimens to 30–40 Gy in 2–4 weeks. No differences between fractionation schemes were seen, with pain responses varying from 80% to 85%. A larger body of literature concerns palliation of bony sites, of which spinal sites typically constitute a large proportion. Several prospective series exploring external-beam palliative regimens of bone metastases have been reported.8–14 Most studies included varied primary histologies, although in the study by Steenland et al.14 a large proportion of patients (39%) had previous breast or prostate cancer. Their results are in agreement with those of other randomized trials, with rates of partial pain relief ranging from 55% to 89% and complete pain responses of 21% to 58%. The median duration of pain relief was typically less than 1 year and the median retreatment rate was 20% (in the studies that reported this datum), and was as high as 44%. Although standard external beam palliative regimens may be effective in patients with a limited life expectancy, these data raise the concern that they may be inadequate for patients with metastatic breast carcinoma, who are expected to have a median survival of 22.6 months,15 long enough to develop symptomatic recurrence within the spine.

These spinal recurrences pose a therapeutic challenge. The symptomatology, neurologic risk, and morbidity associated with recurrent spinal disease remain as significant as at the time of initial treatment and options are limited, as before. Surgery is more difficult because of postradiation fibrosis and poorer vascularity within the operative field. Furthermore, these critical structures have considerably lower radiation tolerances than at initial treatment time, which limits radiation options. The burden of morbidity carried by these patients and the lack of other durable palliative treatments necessitates an exploration of alternatives to standard palliative regimens.

The spine is an ideal site for radiosurgical treatment. The spine is relatively immobile, which assists in radiosurgical delivery and facilitates its imaging for confirmation of targeting. The high doses possible when radiosurgical techniques are used, a consequence of reliably excluding the spinal cord and other dose-limiting structures from the high-dose region, may also result in more durable control within the spine in a population of patients failing conventional radiation treatment.

The CyberKnife is a frameless, image-guided robotic radiosurgical system that can deliver radiation precisely to virtually any body site16–18 with submillimeter accuracy.19 Perhaps the leading principle of radiosurgery is that extreme conformity and accuracy can minimize the volume of normal tissue irradiated to such an extent that the need to deliver doses in multiple fractions is minimized, and there can be considerable escalation of dose. A therapeutic benefit is supposed on these grounds, and this has been amply demonstrated in cranial radiosurgery.20–24 Although technically possible outside the brain, the benefit of radiosurgical treatment of extracranial sites has not been demonstrated.

We consider retrospectively a series of women with breast cancer metastases to the spine who failed prior external beam radiation and who were subsequently treated with CyberKnife stereotactic radiosurgery. We used a matched-pair analysis to compare outcomes in this group to those in women with breast cancer metastases to the spine who were treated with conventional external-beam techniques (CRT). Women were matched with respect to time from original diagnosis to diagnosis of metastases, age, ambulatory level, and performance status. Most important, the group of women treated with CyberKnife radiosurgery represents a more difficult therapeutic challenge, as 17 of 18 of them had had prior radiotherapy within or immediately adjacent to the radiosurgery field and the remaining patient had prior radiation to within a few vertebral bodies of the active site, whereas no women in the conventionally treated group had such a restriction.


From March 12 2002 to January 1 2005, 18 women ≥18 years of age with histologically confirmed spinal metastases from breast carcinoma were treated at Georgetown University Hospital with CyberKnife radiosurgery. Patients were treated using fiducial guidance, with fiducial markers implanted into the spine before planning and treatment. Treatment doses varied according to clinical and technical factors such as proximity to spinal cord and previous irradiated dose. With small variations, patients were treated with 3 fractions: 3 × 800 cGy = 2400 cGy for patients who had not received prior irradiation to the tumor site, and 3 × 700 cGy = 2100 cGy for patients who had received prior irradiation to the involved spinal level. All patients underwent neurological assessment (by F.C.H.), pain assessment before and after irradiation using the Visual Analogue Scale (VAS), and routine follow-up with the radiation oncology and neurosurgery departments at 1, 3, 6, 12 months and on an as-needed basis at other times.

From the records of the Georgetown University Hospital Department of Radiation Medicine, 18 matched patients were found, all of whom completed a course of external beam radiation therapy for spinal metastases from breast cancer between 1995 and 2005. To be eligible for matching, these patients needed to have histologically confirmed metastases from breast carcinoma and to have lesions consistent with spine metastases.

Statistical Methods

Prognosis in the metastatic setting depends on several factors, such as interval from initial diagnosis to metastasis,33 bone vs. visceral metastases,34 tumor differentiation,35 number of metastases,36 location of metastases,37 prior radiotherapy,36 and performance status,36 among others. Several of these factors may by correlated, such as tumor differentiation and risk of visceral metastases35 or location of bony metastases and subsequent visceral metastasis risk.37 In this study, in an attempt to control for these factors, patients formerly treated with CRT were matched to CyberKnife-treated patients based primarily on time from original diagnosis to diagnosis of metastases, with values of 1 year or less counting as a match. Secondary matching variables included presence or absence of visceral metastases, stage at diagnosis, estrogen receptor / progesterone receptor (ER/PR) status of the primary tumor, prior radiotherapy (to the breast or chest wall), and prior chemotherapy. We attempted to match stage at diagnosis to within 1 stage level. The binary variables (presence or absence of visceral metastases, stage at diagnosis, ER/PR status of the primary tumor, prior radiotherapy to the breast or chest wall, and prior chemotherapy) were matched for identity as well as possible from the pool of consecutive external beam patients. The additional and negative prognostic factor of prior radiotherapy to the spinal area was clearly not matched and was distributed unequally between the 2 groups, being heavily and almost uniformly represented in the CyberKnife group.

The Karnofsky Performance Score (KPS) was used as a measure of performance status. Ambulatory level was scored as 1 for ambulatory with no deficits, 2 for ambulatory with a limp, 3 for use of a cane or walker, 4 for wheelchair requirement, and 5 for bedridden.

The distribution of baseline patient characteristics was compared between treatment groups using descriptive statistics. Chi-square tests for paired data were used to test for the difference in distributions between groups. Survival curves, dated from the time of CRT or CyberKnife spinal treatment, were calculated using Kaplan-Meier methods and compared using generalized Wilcoxon tests, because it is not known whether hazard functions are proportional between groups. Chi-square tests were used for comparing the incidence of complications (using the Radiation Therapy Oncology Group [RTOG] scoring system) between treatment groups.


Comparability of Patients

Table 1 compares the distributions of patient characteristics between the CyberKnife and CRT groups. Patients were paired primarily by time from diagnosis to metastases, so on that variable the distributions were comparable between groups (P = .49), the median time in both groups being greater than 3 years. Stage at diagnosis (P = .83; modal stage 2 in both groups) was also comparable as expected from the matching procedure. For the final binary matching variables, 68 of 90 were matched (data not shown). The CyberKnife and CRT groups were also comparable in age (P = .47; median age, 50–59 years), ambulatory level before metastases (all patients scored 1 in both groups), and pain before metastases (P = .67). The mean KPS was somewhat higher in the CRT group (98.2) than in the CyberKnife group (95.4), but this comparison only approached significance (P = .08).

Table 1. Patient Characteristics by Treatment Received
CharacteristicCyberKnife, No. (%)CRT, No. (%)P
  • CRT indicates conventional external beam radiotherapy; Dx, diagnosis; Mets, metastases; SE, standard error.

  • *

    The racial status for 3 patients in the CyberKnife group was missing.

  • N = 17 patients in both groups.

  • N = 14 patients in the CyberKnife group and 17 patients in the CRT group.

  • §

    N = 8 patients in the CyberKnife group and 14 patients in the CRT group.

All patients18 (100)18 (100) 
Years from Dx to Mets  .49
 0–2.98 (44)8 (44) 
 3–5.93 (17)6 (33) 
 6–157 (39)4 (22) 
Age, y  .47
  < 507 (39)8 (44) 
 50–597 (39)4 (22) 
 60–693 (17)2 (11) 
 ≥701 (6)4 (22) 
Race*  .05
 White12 (67)9 (50) 
 Black3 (17)6 (33) 
 Other0 (0)3 (17) 
 Missing3 (17)0 (0) 
Stage of disease at Dx  .83
 I4 (22)4 (22) 
 II6 (33)9 (50) 
 III3 (17)2 (11) 
 IV4 (22)2 (11) 
 Missing1 (6)1 (6) 
Activity prior to Mets  >.99
 Mean (SE)1 (0)1 (0) 
Activity after diagnosis of Mets  .27
 Mean (SE)1.3 (0.2)1.7 (0.3) 
Karnofsky performance score prior to Mets  .08
 Mean (SE)95.4 (1.4)98.2 (0.9) 
Karnofsky performance score after Mets  .29
 Mean (SE)85.0 (2.7)80.0 (3.8) 
Pain prior to Mets  .67
 Mean (SE)§0.3 (0.1)0.6 (0.5) 
Pain after diagnosis of Mets  .01
 Mean (SE)39 (0.6)62 (0.6) 

Ambulatory Level, Karnofsky Performance Score, and Pain Score

Outcome measures were assessed by treatment group from 1–24 months after treatment. The number of patients available for analysis decreased with time in both the CyberKnife and CRT groups. With respect to ambulation, patients walked without deficit (score of 1) in both groups at time of entry, so that most patients were walking at 1 month, and ambulatory scores did not differ between treatment groups (P = .81); over the first 12 months the average scores ranged from 1–1.6 in both groups, and there was no evidence of a significant difference in scores at any timepoint. At 24 months there were 5 patients or fewer in both groups and evidence of impaired walking (score of 2.5 in CyberKnife and 1.9 in the CRT group). Overall, there were no statistically significant differences in ambulatory levels between groups over the first 24 months of study.

Karnofsky Performance Scores paralleled the changes in ambulatory levels. Performance decreased over the first month, from 95.4 to 82.0 in the CyberKnife group and from 98.2 to 79.7 in the CRT group; both reductions were statistically significant (P < .05), but between groups over the first 12 months of study there was no evidence of a statistically significant difference in scores. At 24 months patients in both groups demonstrated reduced levels of performance (68.3 in the CyberKnife group and 72.5 in the CRT group).

Overall pain decreased significantly from levels before metastases to those 1 month postmetastases, and decreased somewhat thereafter, but there was no evidence of statistically significant differences between treatment groups. For both groups, pain level reflected overall pain; in the CyberKnife group this measured pain arose from nonirradiated lesions and the pain relief within the treated site was judged near-complete in all patients.


Figure 1 shows the survival curve from the start of CyberKnife or CRT treatment, which suggests slightly better survival for CyberKnife patients; however, the difference between curves is not statistically significant (P = .27). As this report is being written, 8 of the CyberKnife and 15 of the CRT patients have died.

Figure 1.

Survival from start of treatment for spinal metastases: CyberKnife versus control.

Treatment Sequelae

The percent of patients with 1 or more acute toxicities of any grade was 56% (10/18) in the CRT group compared with 39% (7/18) in the CyberKnife group. Among the 18 pairs of patients, there was agreement in classification of toxicities for 11 pairs: in 5 pairs both patients had some toxicity; in 6 pairs neither patient had toxicity; in the remaining 7 pairs a patient in 1 group had, whereas the patient in the other group did not. In this latter group of 7 pairs, 5 of the 7 with no sequelae were in the CyberKnife group. The chi-square test for paired data was not statistically significant (P > .20). One patient in the CRT group was classed as having no sequelae, although this patient developed neutropenia and sepsis 3 months after the start of CRT and died of this toxicity. In this analysis, this toxicity was not attributed to CRT.

Table 2 gives the types and severity of acute toxicities by treatment group. In the CyberKnife group the most common toxicities were nausea, fatigue, and dysphagia. All were RTOG Grade 1–2 and fatigue and dysphagia occurred in only about 10% of the patients. In the CRT group, diarrhea occurred in 4 (22%) of patients, with an average grade of 1.9. Other sequelae occurring in greater than 10% of patients were nausea and/or vomiting (4 patients, 27%) and esophagitis (11%). In the CyberKnife group, 1 patient noted transient paresthesias in 1 leg. Overall, 10 patients (56%) had at least some toxicity after CRT, compared with 6 patients (39%) receiving CyberKnife treatment, although this 17% difference was not statistically significant (P > .20) because of the small number of patients.

Table 2. Types and Severity of Complications After Treatment
No. (%)Average gradeRangeNo. (%)Average gradeRange
  1. CRT indicates conventional external beam radiotherapy.

Nausea1 (6)222 (11)1.51–2
Nausea and vomiting0 (0)2 (11)22
Fatigue2 (11)111 (6)11
Dysphagia2 (11)1.51–21 (6)1.51–2
Diarrhea0 (0)4 (22)1.91–3
Hypothyroid0 (0)1 (6)33
Esophagitis0 (0)2 (11)1.31–2
Thrombocytopenia0 (0)1 (6)11
Sore throat0 (0)1 (6)11
Dermatitis0 (0)1 (6)11
Heartburn0 (0)1 (6)11
Dry mouth0 (0)1 (6)11
Increase in blood sugar0 (0)1 (6)11
Lower extremity numbness1 (6)210 (0)


There are no standard treatment options for women with metastatic breast cancer who suffer a failure within a previously irradiated spinal site. CyberKnife radiosurgery is technically an effective option for these women, but because of the intolerance of the surrounding normal structures, such as the spinal cord, esophagus, and bowel to reirradiation, there are no ideal comparison groups to use for assessment of outcomes. We compared a series of women with recurrent spinal metastases after radiation with a matched cohort treated with standard radiation techniques up-front for spinal metastases. We believe that establishing equivalence with this group would demonstrate the efficacy of this treatment, particularly as the matched group is unmatched in an important prognostic factor—failure of prior radiation treatment—and would be expected to perform better than the group treated by the CyberKnife.

Although not well reported in the literature, there is a general concern that tumors that have failed prior radiation treatment are more resistant to subsequent courses of radiation. To our knowledge this has not been reported for breast cancer, although there are in vitro data from cell lines derived from human cancers that show increased radioresistance in recurrent tumors compared with nonrecurrent tumors.39 Second, although there are also little supporting published data, there is the possibility that women who have failed prior courses of radiation might be in later stages of their disease process, and more likely to succumb to their disease than those whose disease is controlled. It is clearly possible that the CyberKnife group represented a group with a poorer prognosis, and this would dilute any statistical benefit.

The breast cancer patients treated with CRT were matched with CyberKnife patients on the basis of time interval from initial diagnosis to metastasis,32 bone versus visceral metastases,33 prior radiotherapy to the primary region,34 prior chemotherapy, and ER/PR status. Other prognostic factors, such as performance status,34 were not matched but were equally distributed between groups. There was equivalence between these regimens with respect to palliative and survival endpoints, although KPS was improved in the short course for CyberKnife patients.

The available reports of spinal radiosurgery7, 25–31 have limited follow-up, but they demonstrate the feasibility and safety of the procedure in the treatment of various histologies. Only 1 report of dose-escalation in the spine for breast carcinoma has been published: Gerszten et al.7 treated 50 patients with breast cancer metastases with single-fraction doses of 15 to 22.5 Gy using radiosurgical techniques. Long-term (mean follow-up of 16 months) pain control was obtained in 96% of patients and long-term tumor control was obtained in each of the 8 patients for whom radiosurgery was the primary treatment modality.

The beneficial results from CyberKnife retreatment might be explained by the biologically equivalent doses delivered, possibly greater than that delivered initially by CRT. Despite the frequency of breast cancer, data concerning the dose-response for breast cancer are fragmentary. In a study by Arriagada et al.,37 in which women with localized breast cancer were treated to the breast with radiotherapy alone (that is, without lumpectomy), a dose-response was found for tumors < 4 cm in size. Doses of 40–50 Gy had a 3-year local control rate of 25%, doses of 50–60 Gy had a local control rate of 59%, doses of 70–80 Gy had a local control rate of 81%, and doses of > 80 Gy had a local control rate of 100%. It is possible to compare different fractionation schemes by using the linear-quadratic formalism, also known as the alpha/beta formalism.38 Comparisons between standard fraction schemes and hypofractionated schemes such as that used in this series can be made once the α/β value is known. The fractionation schemes in question can be converted to a suitable reference fractionation scheme, often 2 Gy fractions given 5 days a week. If we assume 300 cGy × 10 fractions, a common fraction scheme for palliative irradiation of the spine, and use α/β values of 4, 6, or 8 Gy, we have biologically equivalent doses in 2 Gy fractions of 3300, 3375, and 3500 cGy. Making a similar calculation for 900 cGy × 3 fractions, a dose received by most of the target volume when 800 cGy × 3 fractions is prescribed to the periphery using the CyberKnife, then we have biologically equivalent doses in 2 Gy fractions of 4590, 5062, and 5850 cGy for the same α/β values. If the dose-response data of Arriagada et al.37 are generalized to the metastatic setting, then the CyberKnife fractionation scheme should result in significantly better tumor control (expected at about 25%–59% at 3 years) than regimens employing 300 cGy × 10 fractions (considerably less than 25%).

As most (17/18) of the CyberKnife patients had prior radiotherapy to the involved spinal site, the general palliative equivalence of these 2 regimens demonstrated in this study is noteworthy. These patients are generally not candidates for further conventional radiation and a palliative retreatment that can be shown to be as effective as initial palliative therapy can be of substantial benefit to these patients. This also suggests further evaluation of CyberKnife radiosurgery of the spine as a primary palliative modality.


This matched-pair analysis between previously treated patients retreated with the CyberKnife and up-front conventionally irradiated patients showed equivalent benefits without a demonstrable increase in toxicity, particularly myelitis. Even equivalence in outcomes is a notable success for a previously treated group with few treatment alternatives.