A comparison of surgical resection and stereotactic radiosurgery in the treatment of solitary brain metastases

Int. J. Radiation Oncology Biol. Phys., Vol. 55, No. 5, pp. 1169 –1176, 2003 Copyright © 2003 Elsevier Science Inc. Printed in the USA. All rights reserved 0360-3016/03/$–see front matter

doi:10.1016/S0360-3016(02)04379-1

CLINICAL INVESTIGATION

Brain

A COMPARISON OF SURGICAL RESECTION AND STEREOTACTIC RADIOSURGERY IN THE TREATMENT OF SOLITARY BRAIN METASTASES BRIAN PATRICK O’NEILL, M.D.,* NANCY J. ITURRIA, PH.D.,† MICHAEL J. LINK, M.D.,‡ BRUCE E. POLLOCK, M.D.,‡ KARLA V. BALLMAN, PH.D.,† AND JUDITH R. O’FALLON, PH.D.† Departments of *Neurology and †Neurosurgery, and ‡Cancer Center Statistics, Mayo Clinic and Foundation and the Mayo Clinic Cancer Center, Rochester, MN Purpose: To determine whether neurosurgery (NS) or stereotactic radiosurgery (RS) provided better local tumor control and enhanced patient survival. Methods and Materials: Retrospective review of all solitary brain metastases (SBM) patients newly diagnosed at Mayo Clinic Rochester between 1991 and 1999. Eligible patients satisfied tumor size and SBM site criteria to qualify for both NS and RS. Results: There were no significant differences between 74 NS and 23 RS patients in terms of baseline characteristics (age, gender, systemic disease type, systemic disease status, signs/symptoms at SBM presentation) or percent of patients who received whole brain radiotherapy. Median follow-up for alive patients was 20 months (range 0 –106 months). There was no significant difference in patient survival (p ⴝ 0.15); the 1-year survival rate was 56% for the RS patients and 62% for the NS patients. Multivariate Cox regression analysis found that a significant prognostic factor for survival was a performance score of 0 or 1. There was a significant (p ⴝ 0.020) difference in local tumor control between NS and RS for solitary brain metastasis; none of the RS group had local recurrence compared to 19 (58%) of the NS group. Conclusion: The need for a Phase III study comparing these two techniques appears to be supported by the data from this study. © 2003 Elsevier Science Inc. Brain metastases, Radiosurgery, Neurosurgery, Gamma knife.

are being detected earlier in the cancer patient’s illness, the potential to improve the duration and quality of life is greatest in these patients. Unlike the patient with multiple brain metastases, local therapies should be applicable. However, it is not clear which local therapy produces superior results. It is thought that patients with SBM benefit from more aggressive care than whole brain radiation therapy (WBRT) alone. Compared with WBRT, both neurosurgery (NS) and stereotactic radiosurgery (RS) appear to provide to the patient improved local tumor control, longer survival, and better quality of life (4 –7). However there is no class 1 evidence to support the use of one technique over the other for patients with SBM not suffering from symptomatic mass effect. The purpose of this study was to compare local tumor control and survival for patients with SBM having NS or RS.

INTRODUCTION Brain metastases are an important clinical problem for cancer patients, with approximately 25% of patients with brain metastases dying from neurologic causes (1). In addition, because of their disabling impact on cognition, memory, language, mobility, and adaptive skills, brain metastases are responsible for disproportionate morbidity and mortality for this patient population. Approximately 200,000 cancer patients in the United States will develop brain metastases this year, and roughly half of these will be solitary (2). As the treatments for systemic disease improve, it is reasonable to assume that the number of cancer patients who develop brain metastases will increase, as well as the fraction of brain metastases that are solitary (3). Because solitary brain metastases (SBM)

and Deborah A. Gorman, R.N., for the Mayo Clinic Department of Neurosurgery Gamma Knife Database management. We also acknowledge the support from the Linse Bock Program in NeuroOncology, Mayo Clinic Cancer Center (MCCC), and support from the Cancer Centers Support Grant P30 CA15083 to MCCC, an NCI-designated Comprehensive Cancer Center. Last, we acknowledge the trust and support of our patients and their families. Received Aug 5, 2002, and in revised form Oct 23, 2002. Accepted for publication Oct 28, 2002.

Reprint requests to: Brian Patrick O’Neill, M.D., Department of Neurology, Mayo Clinic and Foundation, 200 SW First Street, Rochester, MN 55905. E-mail: [email protected] Supported in part by the Linse Bock Program in Neuro-Oncology, Mayo Clinic Cancer Center (MCCC) and the Cancer Centers Support Grant P30 CA15083 to MCCC, an NCI-designated Comprehensive Cancer Center. Acknowledgments—We greatly acknowledge the help of Dr. Shekhar Dagam for data retrieval; Xiomara Carrero Martinez and Angela Fought for expert data analysis support and graph design; 1169

1170

I. J. Radiation Oncology

● Biology ● Physics

Volume 55, Number 5, 2003

Table 1. Baseline characteristics

Age Mean ⫾ SD Median (IQR) Gender Female n (%) Male n (%) ECOG performance status, n (%) 0 1 2 3 Symptomatic, n (%) Yes No Side of brain, n (%) Left Right Midline Type of systemic disease, n (%) Lung cancer GU cancer GI cancer Melanoma Other cancer Status of systemic disease, n (%) No evidence of disease Stable Progression Synchronous presentation Unknown

METHODS AND MATERIALS

RS (n ⫽ 23)

NS (n ⫽ 74)

p value

61 ⫾ 13 66 (51, 71)

62 ⫾ 12 63 (55, 70)

0.80

8 (35) 15 (65)

34 (46) 40 (54)

0.47

0 (0) 15 (65) 4 (17) 4 (17)

28 (38) 30 (41) 13 (18) 3 (4)

17 (74) 6 (26)

66 (89) 8 (11)

0.09

7 (30) 12 (52) 4 (17)

37 (50) 35 (47) 2 (3)

0.029

11 (48) 4 (17) 3 (13) 1 (4) 4 (17)

40 (54) 10 (14) 7 (9) 6 (8) 11 (15)

6 (26) 8 (35) 5 (22) 4 (17) 0 (0)

33 (45) 12 (16) 11 (15) 16 (22) 2 (3)

0.92

0.24

Abbreviations: RS ⫽ stereotactic radiosurgery; NS ⫽ neurosurgery; SD ⫽ standard deviation; IQR ⫽ intraquartile range; ECOG ⫽ Eastern Cooperative Oncology Group; GU ⫽ genitourinary; GI ⫽ gastrointestinal.

All aspects of this study were reviewed and approved by the Mayo Foundation Institutional Review Board. Patients who denied access to their medical records for research purposes were excluded from this study. Patients who were potentially eligible for this study included all those with a diagnosis coded as “brain metastasis” or “brain metastases.” These patients were identified from three different databases: the Mayo Clinic Rochester (MCR) Medical Index, the department of neurosurgery’s surgical database, and that department’s radiosurgery database. The MCR Medical Index contains information of all diagnoses given to each individual patient seen at MCR. The surgical database contains information on all patients who underwent a neurosurgical procedure at MCR, and the gamma knife radiosurgery database has prospectively obtained information on all patients who underwent radiosurgery at MCR. The medical records of the patients identified here were retrospectively reviewed to identify patients who were truly eligible for this study. To be included in this study, patients had to have had a neurosurgical and neurologic evaluation at MCR, before their procedure, and the subsequent surgical procedure (either microsurgery or RS) had to be done at MCR. Finally, patients had to be candidates for both NS and RS to minimize potential confounding variables between the two groups. Patients were deemed to be candidates for either procedure if (1) they had a tumor size less than 35 mm, (2) they did not have brainstem or deep-seated lesions, and (3) they did not have ventricular obstruction. Both the RS and the NS patients typically received a standard dose of corticosteroids at the time of the procedure. Steroids were then tapered over 2– 4 weeks depending on the patients’ clinical course. Data were collected on baseline characteristics, treatment variables, treatment complications, and outcome. Differ-

Fig. 1. Overall survival.

A comparison of surgical resection and stereotactic radiosurgery

● B. P. O’NEILL et al.

1171

Fig. 2. KM Survival by PS group.

ences in categorical variables between RS patients and NS patients were compared using Wilcoxon tests (for ordered variables), chi-square tests (for nonordered variables), and Wilcoxon tests for continuous variables. The survival time was measured from the time of the procedure (RS or NS). Survival curves were constructed using Kaplan-Meier estimates (8) and were compared using the log–rank test. A multivariable Cox proportional hazards regression model (9) was used to adjust for baseline differences between the NS and RS groups and to determine the effect of treatment. Because treatment was not randomly assigned in this patient

population, potential confounding and selection biases were accounted for by developing a propensity score (10) for assignment to treatment. The rationale and methods underlying the use of a propensity score have been previously described (10). The propensity for RS treatment was determined without regard to outcome, using multivariable logistic regression analysis. A full nonparsimonious model was developed that included 25 covariates (age, categorized date of primary cancer diagnosis, types of primary cancer, lobe mainly involved, hemisphere involved, presence of symptoms, presence of signs, Eastern Cooperative Oncol-

1172

I. J. Radiation Oncology

● Biology ● Physics

Volume 55, Number 5, 2003

Fig. 3. Survival experience of RS group: significantly worse for those with left-sided lesions (upper graph); borderline significantly worse for those symptomatic at diagnosis.

ogy Group [ECOG] performance score, and status of primary disease). This model yielded a c statistic of 0.89, indicating a strong ability to differentiate between patients treated with RS and patients treated with NS. A propensity score was then calculated for each patient. This score was added as an adjusting variable to the multivariable Cox proportional hazards regression model to further adjust for potentially confounding differences between the patients in

the RS and NS groups to assess differences between the two treatments with respect to overall survival. RESULTS A total of 97 SBM patients (42 women, 55 men) met the eligibility criteria for this study. Baseline standard-dose postgadolinium magnetic resonance imaging was performed

A comparison of surgical resection and stereotactic radiosurgery

on all patients to confirm that there was only one brain tumor. Of the 97 study patients, 23 (24%) underwent RS and 74 (76%) underwent NS. The use of WBRT was similar between the two groups (82% of NS patients and 96% of RS patients, p ⫽ 0.172). Time to death or last follow-up was measured from the date of the procedure. The follow-up time, time to death or last follow-up, ranged from 0 to 106 months (median 14 months). For survivors, the time to last follow-up ranged from 0 months to 106 months (median 20 months). The time to death for the nonsurvivors ranged from 1 months to 63 months (median 12 months). The baseline characteristics of the two treatment groups are summarized in Table 1. Age distributions, gender, the distribution of systemic disease type, and the distribution of systemic disease status do not appear to differ between the two groups. The RS group had a significantly smaller fraction of patients with a good performance score (p ⫽ 0.0016) and a smaller fraction of patients with a tumor on the left side of the brain (p ⫽ 0.029). There was a larger fraction of patients who were symptomatic at the time of the metastases diagnosis in the NS group compared with the RS group (89% vs. 74%) but this difference did not achieve statistical significance (p ⫽ 0.09). The distribution of signs and symptoms at the time of the metastasis diagnosis did not appear to differ between the two groups. The Kaplan-Meier survival (Fig. 1) curves show that the two treatment groups had virtually identical survival rates during the first year after the procedure, but the NS patients survived longer during the second year. The difference in survival experience was not statistically significant (log– rank p ⫽ 0.15) though, possibly because the RS group was small (n ⫽ 23) and had a less favorable mixture of prognostic factors. Because the RS group had a substantially smaller percentage of patients with good performance scores, the Kaplan-Meier survival curves for the two treatment groups were compared separately in two patient subgroups: those patients with a performance score of 0 or 1, and those patients with a performance score of 2 or 3. The differences in survival experience between the RS and NS treatment groups were notably reduced in both of these two subgroups (Fig. 2). The survival experience in the RS group was significantly worse in the subgroup of patients with left-sided lesions (log–rank p ⫽ 0.0012) and borderline significantly worse for patients who were symptomatic at diagnosis (log–rank p ⫽ 0.07; Fig. 3). Variables found to be independently associated with survival in a multivariable Cox proportional hazards model, which adjusted for age, were ECOG performance status (0 or 1 vs. 2 or 3) and systemic disease status (no evidence of disease vs. all other) at the time of the diagnosis for brain metastases (Table 2). Notably, treatment type was not independently associated with survival (p ⫽ 0.62). After adjusting for the propensity score (which adjusted for all factors that could be potentially associated with choice of treatment), the type of treatment still was not significantly associated with survival (p ⫽ 0.88, Table 3). The causes of death did not significantly differ between

● B. P. O’NEILL et al.

1173

Table 2. Multivariable Cox proportional hazards model

Variable Age Eastern Cooperative Oncology Group performance status (0 or 1) Systemic disease status (no evidence of disease) Stereotactic radiosurgery treatment

Hazard ratio (95% confidence interval)

p value

1.02 (1.00–1.04) 0.49 (0.29–0.82)

0.038 0.006

0.53 (0.32–0.88)

0.013

1.15 (0.67–1.95)

0.62

the two groups (p ⫽ 0.22). Forty-eight percent of the patients in the RS group and 59% of the patients in the NS group died of systemic tumor alone. However, it should be noted that 6 of 21 (29%) of RS patients died of cerebral tumor compared with 10 of 54 (11%) of the NS patients (p ⫽ 0.36). There were a greater percentage of short-term complications observed in the NS group but a somewhat greater percentage of long-term complications in the RS group (Fig. 4). The differences in short-term and long-term complication rates between the two groups were not statistically significant. At the time of last follow-up, recurrence status was available on 21 of the RS patients and 64 of the NS patients. Six patients (29%) in the RS and 19 patients (30%) in the NS group had a recurrence of their brain tumor. The recurrence sites differed significantly between the two groups, with none of the RS group experiencing a local recurrence compared with 11 (58%) of the NS group experiencing a local recurrence (p ⫽ 0.020). The locations of the brain recurrences in the two groups are not statistically significant (Fig. 5, p ⫽ 0.079), but there were significantly more local recurrences in the NS group compared with the RS group (11 of 19 vs. 0 of 6, p ⫽ 0.020). DISCUSSION Typically, SBM patients have been treated with WBRT after or instead of craniotomy and removal. In a 1990 report, Patchell et al. concluded that surgical resection combined Table 3. Multivariable Cox proportional hazards model following adjustment for propensity score

Variable Age Eastern Cooperative Oncology Group performance status (0 or 1) Systemic disease status (no evidence of disease) Stereotactic radiosurgery treatment Propensity score

Hazard ratio (95% confidence interval)

p value

1.02 (1.00–1.05) 0.53 (0.31–0.90)

0.028 0.019

0.60 (0.35–1.05)

0.073

0.95 (0.50–1.80)

0.88

0.94 (0.83–1.06)

0.33

1174

I. J. Radiation Oncology

● Biology ● Physics

Volume 55, Number 5, 2003

Fig. 4. Complications.

with postoperative radiotherapy was more effective than treatment with radiotherapy alone (11). In a follow-up study, the same authors concluded that patients with cancer and single metastases to the brain who receive treatment with surgical resection and postoperative radiotherapy have fewer recurrences of cancer in the brain and are less likely to die of neurologic causes than are similar patients treated with surgical resection alone (4). However, in this study there was no significant difference between the two groups

in overall length of survival or the length of time that patients remained functionally independent. Furthermore there are good data to support an increased risk of late complications from WBRT. In 1989, DeAngelis and Posner described “radiation-induced dementia” in patients cured of brain metastases (12). In addition, the experiences with prophylactic cranial irradiation (PCI) involved field radiotherapy in glioblastoma multiforme and therapeutic wholebrain radiotherapy in primary CNS lymphoma are relevant. As

A comparison of surgical resection and stereotactic radiosurgery

● B. P. O’NEILL et al.

1175

Fig. 5. Proportion of recurrence in each group at the specified locations.

many as 30% of lung cancer survivors who received PCI had a functionally significant subcortical dementia syndrome (13). In 1980, Hochberg and Slotnick described neuropsychologic impairment in astrocytoma survivors (14). Last, significant neurocognitive decline is seen in primary central nervous system lymphoma patients treated with WBRT either alone or in combination with chemotherapy (15). Although the decline is likely a combination of the disease, comorbidity, and its treatment, the irony remains that the patients most in need of treatment may be the patients most at risk of damage from it. Although there are compelling data to support resection for SBM, roughly only half of SBM patients are surgical candidates because of comorbidity. Alternatively, many patients with brain metastases have undergone RS primarily with either the gamma knife (GK) or a linear accelerator-based system (Linac). A summary of 21 independent reports of GK or Linac radiosurgery involving more than 1700 patients and more than 2700 lesions reported by Boyd and Mehta (16) found a mean local control rate of 83% and median patient survival of 9.6 months. Factors significantly associated with improved survival were controlled extracranial disease, three or fewer lesions, and higher performance scores. A dose–response curve was suggested, and these authors recommended using tumor margin doses of 18 Gy or more whenever it was deemed safe. To the best of our knowledge, there has not been a Phase III trial comparing NS and SR for patients with SBM. Ideally, a prospective trial could be performed to better understand the relative roles that these two techniques should play in the management of this patient population. Such a study would better control for the heterogeneous nature of this patient population with respect to factors that correlate with survival but are not neurologic in nature. It would also minimize other confounding variables and treat-

ment selection bias. Nonetheless, retrospective reviews, such as our study and others, (6) should be able to provide some useful information regarding local tumor control and treatment-related morbidity. The result of our study found neither NS nor RS was superior for patients with small- to moderate-size tumors. Thus a prospective trial appears to be warranted to compare these treatment modalities. Does our study’s results justify this conclusion? The data collected reflect the nature of the MCR’s large referral practice. The Mayo Clinic Cancer Center’s strengths are in cancers that frequently metastasize to brain, such as carcinomas of the lung and breast. In addition, a multidisciplinary neuro-oncology program has been in existence since 1996. This practice model assures that each person chosen for a particular neurosurgical procedure has had a full and considered review, and that a consensus approach has been reached by the treating physicians. A unitary clinical record has been employed since 1918 with a standardized neurologic examination. This assures comprehensive and reproducible data retrieval for analysis and research. We recognize that there are several limitations to this study. First, this study rests entirely upon retrospective data. We were unable to assess postprocedure quality of life (QOL). This was a retrospective analysis and, as such, we did not feel that retroactive QOL determinations based on data extraction from the records would be valid. However, this is an important component of the prospective evaluation of any procedure. The newer and more robust QOL measures should allow them to be entered into any prospective study. Second, the cohort represents highly selected patients. Given the small population base in Rochester, the surrounding Olmsted County, and Southeastern Minnesota, a study such as this could not have been mounted. However,

1176

I. J. Radiation Oncology

● Biology ● Physics

along with a referral practice comes selection bias that may skew the results of any study. Third, a greater percentage of patients with SBM underwent RS compared with NS late in the study period compared with the first several years. This undoubtedly is related to our personal experience with the results of RS for brain metastases patients. However, the possibility of undetected differences in the two treatment groups could add confounding variables not related to their actual treatment. For example, advances in imaging over the study period could potentially distribute patients differently

Volume 55, Number 5, 2003

into the two groups. Specifically, more sensitive imaging would be better able to detect additional small tumors not appreciated on earlier studies. Although we used a propensity score to adjust for potentially confounding variables, this score does not adjust for variables that are not in the model because of absence of available data. Nevertheless, our interpretation of the data are that a Phase III study is warranted, and such a study is being planned under the auspices of the American College of Surgeons’ Oncology Group (principal investigator: Dr. Samuel A. Wells).

REFERENCES 1. Posner JB, Chenik NL. Intracranial metastases from systemic cancer. Adv Neurol 1978;19:579–592. 2. Johnson JD, Younge B. Demographics of brain metastases. Neurosurg Clin North Am 1996;7:337–344. 3. Auperin A, Arriagadia R, Pignon JP, et al. Cranial irradiation for preventing brain metastases of small cell lung cancer in patients in complete remission. N Engl J Med 1999;341:476– 484. 4. Patchell RA, Regine WF, et al. Postoperative radiotherapy in the treatment of single metastases to the brain. JAMA 1998; 280:1485–1489. 5. Gaspar L, Scott C, Rotman M, et al. Recursive partitioning analysis (RPA) of prognostic factors in three Radiation Therapy Oncology Group (RTOG) brain metastases trials. Int J Radiat Oncol Biol Phys 1997;37:745–751. 6. Muacevic A, Kreth FW, Horstmann GA, et al. Surgery and radiotherapy compared with gamma knife radiosurgery in the treatment of solitary cerebral metastases of small diameter. J Neurosurg 1999;91:35–43. 7. Loeffler JS, Kooy HM, Wen PY, et al. The treatment of recurrent brain metastases with stereotactic radiosurgery. J Clin Oncol 1990;8:576–582. 8. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457–481.

9. Cox DR. Regression models and life tables. J R Stat Soc 1972;34:187–220. 10. D’Agostino RB. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med 1998;17:2265–2281. 11. Patchell RA, Tibbs PA, Walsh JW, et al. A randomized trial of surgery in the treatment of single metastases to the brain. N Engl J Med 1990;322:494–500. 12. DeAngelis LM, Posner JB. Radiation-induced dementia in patients cured of brain metastases. Neurology 1989;39:789– 796. 13. So NK, O’Neill BP, Frytak S, et al. Delayed leukoencephalopathy in survivors with small cell lung cancer. Neurology 1987;37:1198–1201. 14. Hochberg FH, Slotnick B. Neuropsychologic impairment in astrocytoma survivors. Neurology 1980;30:172–177. 15. O’Neill BP, O’Fallon JR, et al. The consequences of treatment and disease in patients with primary central nervous system lymphoma (PCNSL): Cognitive function and performance status. Neuro-Oncol 1999;1:196–203. 16. Boyd TS, Mehta MP. Stereotactic radiosurgery for brain metastases (Review). Oncology 1999;13:1397–1409.