LINEAR ACCELERATOR STEREOTACTIC = RADIOSURGERY=20 FOR METASTATIC BRAIN TUMORS: 17 YEARS OF EXPERIENCE AT THE = UNIVERSITY OF=20 FLORIDA

[CLINICAL STUDIES]

Swinson, Bradley M. B.S.; = Friedman, William A.=20 M.D.

Department of Neurosurgery, = University of=20 Florida, Gainesville, Florida

Reprint=20 requests: William A. Friedman, M.D., P.O. Box 100265, = Department=20 of Neurosurgery, University of Florida, Gainesville, FL 32610. = Email:=20 friedman@neurosurgery.ufl.edu

Received,=20 February 2, 2007.

Accepted,=20 January 3, 2008.

Article Outline=20

Abstract=20
PATIENTS=20 AND METHODS=20
- Study=20 Design
- Statistical=20 Methods
RESULTS=20
- Patient=20 and Treatment Characteristics
- Survival=20
- Local=20 Control
- Regional=20 Control
- Complications=20
DISCUSSION=20
- Survival=20 and Local Control after SRS
- SRS=20 with or without WBRT
- Study=20 Limitations
CONCLUSIONS=20
Acknowledgments=20
REFERENCES=20
- COMMENTS=20
Citing=20 Articles

Figures/Tables=20

Abstract TOP=20

OBJECTIVE: To review one of the = largest=20 single-institution experiences treating metastatic brain disease = with=20 stereotactic radiosurgery.

METHODS:=20 We performed a retrospective analysis of 619 patients who = underwent linear=20 accelerator-based stereotactic radiosurgery for 1569 brain = metastases=20 between May 1989 and February 2006. Patient characteristics and = treatment=20 parameters were obtained prospectively. Patients were followed up = at=20 regular intervals clinically and with imaging studies to document = local=20 control, regional control, and survival. Cox proportional hazards = analysis=20 was performed using SAS version 9.1 software (SAS Institute, Cary, = NC).

RESULTS:=20 Median actuarial survival was 7.9 months. 1- and 2-year actuarial = survival=20 probabilities were 0.36 and 0.14, respectively. Radiation Therapy = Oncology=20 Group Recursive Partitioning Analysis Class I or II was associated = with=20 improved survival, but the difference between the two was = insignificant.=20 Female sex, younger age, higher Karnofsky performance status, = controlled=20 primary tumor, absence of systemic metastases, asynchronous = presentation=20 of brain metastasis, fewer brain metastases, smaller total volume = of brain=20 metastases, surgery prior to radiosurgery, and multiple = radiosurgical=20 treatments were also associated with improved survival. Melanoma=20 metastasis was associated with impaired survival. Local control = was=20 achieved in 84.3% of all lesions treated. 1- and 2-year actuarial = local=20 control probabilities were 0.82 and 0.72, respectively. Whole = brain=20 radiation therapy prior to radiosurgery was associated with = improved=20 regional control.

CONCLUSIONS:=20 Linear accelerator-based stereotactic radiosurgery is a safe and = effective=20 treatment for patients with metastatic brain tumors. Selection of = patients=20 who are likely to benefit most from radiosurgery is complex and = treatment=20 decisions should be based on the entire clinical = picture.

Conservative estimates indicate = that there=20 are approximately 170,000 new cases of metastatic brain disease = each year=20 in the United States (41),=20 and it has been suggested that there may be as many as 500,000 = cases (74).=20 This means that brain metastases are the most common intracranial = tumors,=20 outnumbering primary brain tumors more than ten to one (62).=20 Twenty to 40% of patients with systemic cancer will develop = metastatic=20 brain disease; 30 to 40% of these patients will present with a = single=20 brain metastasis and the remaining 60 to 70% will present with = multiple=20 brain metastases (2).=20 It has also been suggested that the incidence of brain metastasis = is=20 increasing, due in part to earlier diagnosis and more effective = treatment=20 of primary neoplastic disease-often with chemotherapeutic agents = that do=20 not effectively cross the blood-brain barrier-and the rising = incidence of=20 lung cancer, the most common primary source of brain metastases = (85).

The primary goal in the = management of=20 metastatic brain disease is palliation of debilitating = neurological=20 symptoms, in order to attain an optimal combination of duration = and=20 quality of life. The median survival of patients with metastatic = brain=20 disease is a dismal 1 month without intervention. Corticosteroid = therapy=20 generally produces rapid palliation of symptoms, which are usually = primarily attributable to peritumoral vasogenic edema, but extends = median=20 survival to only 2 months (65).=20 Whole-brain fractionated external-beam radiotherapy (WBRT), = generally=20 administered to a total dose of 30 Gy in 10 3-Gy fractions (10),=20 further extends median survival to 4 to 6 months (37,63,67),=20 but at least half of all patients treated with this modality = ultimately=20 die from progression of their metastatic brain disease (72).=20 Two landmark studies in the early 1990s proved that, for patients = with a=20 single resectable brain metastasis and good neurological = performance=20 status, the addition of microsurgical resection to WBRT = significantly=20 extends survival (58,61),=20 and a recent prospective randomized trial demonstrated comparable = outcome=20 with the addition of stereotactic radiosurgery (SRS) to WBRT (2).=20 Both modern surgical series and large radiosurgical series = generally=20 report median survivals of 7 to 13.5 months, with a majority of = deaths=20 attributable to systemic disease progression (1,5,7,8,11,14,19,22,28= ,29,33,38,54,55,58-62,70,71,73,75,77,86,87,89,90).=20 However, to date no prospective randomized trial has directly = compared SRS=20 and surgery.

Much debate exists concerning the = optimum=20 treatment for metastatic brain disease, and it has been suggested = that=20 aggressive treatment is probably underused in current practice (6).=20 Treatment paradigms vary from center to center, and care is = generally=20 individualized based on known prognostic factors. Numerous studies = have=20 found younger age (usually <65 yr) (1,7,20),=20 higher Karnofsky Performance Scale (KPS) score (usually = =E2=89=A570) (17,20,31,34,38,55,62,7= 0,75,77),=20 controlled primary tumor (11,17,33,36,88),=20 absence of systemic metastatic disease (1,11,17,31,36,38,49,50= ,62,70,75,77,88),=20 and asynchronous presentation of brain metastases (21,75)=20 to be significantly associated with improved survival. Many have = also=20 found Radiation Therapy Oncology Group Recursive Partitioning = Analysis=20 (RPA) Class (27),=20 a composite of age, KPS score, primary tumor status, and status of = systemic metastatic disease, to be highly correlated with survival = (7,12,33,54-56).=20 As larger experiences are accumulated and analyzed, the = significance of=20 such prognostic indicators is backed by greater statistical power, = allowing future treatment decisions to be made with greater = confidence,=20 even as we await more abundant Class I data. We report, to the = best of our=20 knowledge, the third-largest single-institution experience = treating=20 metastatic brain disease with SRS (28,38,74),=20 and the largest such experience using a linear accelerator-based = SRS=20 system.

PATIENTS AND METHODS TOP=20

Study Design TOP=20

Between May 1989 and February = 2006, 619=20 patients underwent linear accelerator-based SRS at the University = of=20 Florida, Gainesville, for metastatic brain tumors. We have = conducted a=20 retrospective review of all patient characteristics, treatment = parameters,=20 and follow-up data, in order to evaluate the impact of patient and = treatment variables available at the time of treatment on = survival, local=20 control, and regional control following treatment. No patients = were=20 excluded from this analysis.

Diagnosis of metastatic brain = disease was=20 based on a history of an appropriate primary tumor and = characteristic=20 appearance of the brain lesion(s) on magnetic resonance imaging = (MRI).=20 Patient characteristics were entered prospectively into a = computerized=20 database at the time of treatment. These included age, sex, KPS, = site and=20 histology of the primary tumor, whether the primary tumor was well = controlled, whether the patient had systemic metastatic disease, = whether=20 the presentation of brain metastasis occurred within 3 months of = diagnosis=20 of the primary tumor (synchronous presentation of brain = metastasis) or=20 more than three months after diagnosis of the primary tumor = (asynchronous=20 presentation of brain metastasis), number of brain metastases, = location of=20 brain metastases, history of WBRT, and history of surgical = resection of a=20 metastatic brain lesion. Primary tumor control was operationally = defined=20 as no body imaging demonstrating progression of primary disease at = the=20 time of evaluation for and delivery of SRS. Lesions determined=20 radiographically to be located in motor, sensory, visual, or = language=20 areas of the cerebrum or in the brainstem were considered lesions = to=20 eloquent brain. The use of WBRT and/or surgical resection in the = initial=20 management of metastatic brain disease was solely at the = discretion of the=20 referring physician.

All patients were treated using = the=20 University of Florida linear accelerator-based SRS system. The = system and=20 its use have been described in detail in multiple previous = publications=20 (23-25).=20 Treatment parameters, including radiation dose, total dose volume = (equal=20 to total tumor volume), and number of isocenters used, were = entered=20 prospectively into a computerized database at the time of = treatment.

Follow-up of each patient = consisted of=20 frequent clinical evaluation and serial MRI scans obtained every 3 = months,=20 either at the University of Florida or at a location closer to the = patient's home. All scans were reviewed by the senior author (WAF) = as soon=20 as they were available and were compared with the stereotactic = treatment=20 scans. His findings were entered into a computerized database at = the time=20 of review. Local control was evaluated on a per-lesion basis. If a = follow-up scan demonstrated any perceivable enlargement of a = previously=20 treated lesion, local control of that lesion was considered to be = lost on=20 the date of that scan. Additionally, if a previously treated = lesion was=20 resected for any reason, local control of that lesion was = considered to be=20 lost on the date of surgery, regardless of prior radiographic = evidence of=20 local control or the results of surgical pathology. For the = purpose of=20 this analysis, duration of local control is calculated as the = interval=20 between the date of treatment and the date on which local control = was=20 proven to be lost. If a follow-up scan demonstrated metastatic = brain=20 disease not evident on the treatment scan, it was counted as a = loss of a=20 regional control. No data on duration of regional control was = collected.=20 Recurrent and/or new brain metastases were managed with palliative = care,=20 WBRT, surgery, or repeat SRS, at the discretion of the treating = physician=20 and depending on the wishes of the patient and his or her = family.

In order to obtain the most = accurate,=20 up-to-date data possible on the vital status of each patient, a = search of=20 the Social Security Death Index (SSDI) was conducted using two = publicly=20 available online search engines (80,81)=20 and the search algorithm described by Hauser and Ho (30).=20 This algorithm, which involves manually searching for one patient = at a=20 time, first by social security number alone, then by last name, = first=20 name, birth year, birth month, and birth date, and finally by last = name,=20 first name, birth year, and birth month, has been shown to be 86%=20 sensitive and 100% specific (30).=20 If a patient who was not listed as deceased in the SSDI was = discovered to=20 be overdue for a follow-up clinical evaluation or MRI scan, phone = calls=20 were placed to the patient and/or referring physician in order to=20 determine the vital status of that patient.

Statistical Methods TOP=20

The Cox proportional hazards = model was used=20 to explore the relationship between sex, age, KPS score, primary = tumor=20 type (non-small cell lung cancer reference), status of the primary = tumor=20 (controlled versus uncontrolled), presence or absence of systemic=20 metastases, timing of presentation of brain metastasis = (synchronous versus=20 asynchronous), number of brain metastases, total volume of brain=20 metastases, location of brain metastases (eloquent versus = non-eloquent),=20 history of WBRT, history of surgery, SRS dose, and number of SRS=20 treatments, and survival and local control. Added-in-last with = fixed order=20 was used to select the significant variables; variables were added = in the=20 order listed here. Final model analysis of variables that were = significant=20 on model selection analysis was performed using the Type I = (sequential)=20 method, and Type I P values, Type I = hazard=20 ratios, and 95% confidence intervals are reported. These analyses = were=20 conducted using the PROC PHREG and PROC LIFETEST procedures in SAS = (Version 9.1; SAS Institute, Cary, NC).

Logistic regression was used to = explore the=20 relationship between these same variables and regional control.=20 Added-in-last with fixed order was used to select the significant=20 variables; variables were added in the order listed here. Final = model=20 analysis of variables that were significant on model selection = analysis=20 was performed using the Type I (sequential) method, and Type I P values, Type I odds ratios, and 95% = confidence=20 intervals are reported. These analyses were conducted using the = PROC=20 LOGISTIC procedure in SAS (Version 9.1).

Because RPA class is a composite = variable=20 composed of age, KPS score, presence or absence of systemic = metastases,=20 and status of the primary tumor (controlled versus uncontrolled), = each=20 analysis was conducted twice: once including these four variables = but=20 excluding RPA class, and once excluding these four variables and = including=20 RPA class. The results of both analyses are reported. Parameters = were=20 deemed to be statistically significant at a P=20 value of 0.05 or less.

RESULTS TOP=20

Patient and Treatment = Characteristics TOP=20

Between May 1989 and February = 2006, a total=20 of 619 patients underwent 745 treatments for 1569 metastatic brain = tumors.=20 Median age at the time of first treatment was 60 years (mean, 59 = yr;=20 range, 4-86 yr). Median KPS score was 80 (mean, 81.5; range, = 40-100) and=20 579 patients (93.5%) had a KPS score of 70 or greater. The = remainder of=20 the pertinent patient data is summarized in Table=20 1. Three hundred forty-eight patients (56.2%) were initially = treated=20 for single brain metastasis and 271 patients (43.8%) were = initially=20 treated for multiple metastases (868 lesions total). The locations = of all=20 348 single metastases are summarized in Table=20 2. Ninety patients (14.5%) underwent additional radiosurgery: = 10 for=20 lesions recurrent after radiosurgical treatment, 65 for new = lesions, and=20 15 for both recurrent and new lesions. Twenty-nine metastases in = 25=20 patients were treated with radiosurgery twice; no lesion was = treated with=20 radiosurgery more than twice. Median dose to the target periphery = was 1750=20 cGy (mean, 1736 cGy; range, 1000-2250 cGy). Almost all lesions = were=20 treated to the 70 or 80% isodose line. In general, larger lesions = were=20 treated with lower doses, and doses used early in our experience = were=20 lower than those presently used. Currently, we recommend a dose of = 2000=20 cGy to the periphery of lesions smaller than 2.5 cm in diameter. = Median=20 total dose volume (equal to total tumor volume) was 4.6 cm3 (mean, 8.0 cm³; range,=20 <0.1-49.4 cm³). The median number = of=20 isocenters used was two (mean, 3.3; range, 1-22). The number of = metastases=20 treated during each SRS procedure is summarized in Table=20 3.

TABLE 1. Patient = characteristicsa =

TABLE 2. Single brain=20 metastases

TABLE 3. Radiosurgical=20 treatments

Survival TOP=20

Survival data were available for = 597 of 619=20 patients (96.4%) treated during the study interval. At the time of = analysis, 46 patients (7.4%) were alive, with a median follow-up = of 12.8=20 months (mean, 29.5 mo; range, 5.6-117.6 mo) and 551 patients = (89.0%) had=20 died. No data on cause of death was collected. Twenty-six patients = (4.2%)=20 died within 30 days of their first treatment. Twenty-two patients = (3.6%)=20 were lost to follow-up a median of 8.4 months after treatment = (mean, 25.9=20 mo; range, 0-124.0 mo), and although 12 of them were considered = very=20 likely to be alive at the time of manuscript preparation, based on = available pre-treatment and follow-up data, all were excluded from = the=20 statistical analysis. The median actuarial survival was 7.9 months = (95%=20 confidence interval, 7.0-8.8 mo) from the time of first SRS = treatment. The=20 1- and 2-year actuarial survival estimates were 0.36 (95% = confidence=20 interval, 0.34-0.38) and 0.14 (95% confidence interval, = 0.12-0.16),=20 respectively (Fig.=20 1). RPA class was found to be significantly associated with = survival,=20 as discussed below. Median actuarial survival was 9.4 (range, = 8.0-12.7),=20 7.8 (range, 6.7-8.7), and 3.9 (range, 2.0-6.0) months for RPA = Classes I,=20 II, and III, respectively. The 1- and 2-year actuarial survival = estimates=20 were 0.42 =C2=B1 0.05 and 0.21 =C2=B1 0.04, 0.35 =C2=B1 0.02 and = 0.13 =C2=B1 0.02, and 0.25 =C2=B1=20 0.07 and 0.03 =C2=B1 0.03, respectively (Fig.=20 2).

FIGURE 1. Kaplan-Meier = plot=20 demonstrating actuarial = survival.

FIGURE 2. Kaplan-Meier = plot=20 demonstrating actuarial survival, stratified by Recursive=20 Partitioning Analysis (RPA) = class. Survival=20 did not differ significantly between RPA Classes I and II, = but RPA=20 Class III was associated with impaired=20 survival.

Model selection hazard analysis = identified=20 female sex, younger age, higher KPS score, tumor type, controlled = primary=20 tumor, absence of systemic metastases, better RPA class, = asynchronous=20 presentation of brain metastasis, fewer metastases, smaller total = tumor=20 volume, prior surgery, and more SRS treatments to be significantly = associated with improved survival when adjusting for their = respective=20 previous variables in the model. Eloquent tumor location, prior = WBRT, and=20 SRS dose were not significantly associated with survival = probability.

Type I final model hazard = analysis identified=20 the following: female sex was associated with improved survival, = with male=20 sex exhibiting a 33.3% higher hazard rate (P =3D=20 0.001). Younger age was associated with improved survival, with = each=20 increasing year exhibiting a 0.9% higher hazard rate (P =3D 0.016). Higher KPS score was = associated with=20 improved survival, with each additional 10 points exhibiting a = 1.3% lower=20 hazard rate (P =3D 0.004). Using = non-small cell=20 lung cancer as the reference category, melanoma metastasis was = associated=20 with impaired survival, exhibiting a 64.5% higher hazard rate in = the=20 analysis excluding RPA class (P =3D = 0.001) and a=20 51.5% higher hazard rate in the analysis including RPA class (P =3D 0.004). There was a trend toward = improved=20 survival with small-cell lung cancer metastasis, which exhibited a = 27.2%=20 lower hazard rate in the analysis excluding RPA class (P =3D 0.072) and a 28.6% lower hazard rate = in the=20 analysis including RPA class (P =3D = 0.054), as well=20 as with metastasis from an unknown primary, which exhibited a = 36.5% lower=20 hazard rate in the analysis excluding RPA class (P =3D 0.053) and a 33.9% lower hazard rate = in the=20 analysis including RPA class (P =3D = 0.079). Primary=20 control was associated with improved survival, exhibiting a 19.9% = lower=20 hazard rate than uncontrolled primary tumor (P =3D=20 0.024). Absence of systemic metastases was associated with = improved=20 survival, with systemic metastases exhibiting a 24.7% higher = hazard rate=20 (P =3D 0.026). RPA Class III was = associated with=20 impaired survival, exhibiting a 128.2% higher hazard rate than RPA = Class I=20 (P =3D 0.000). Notably, although model = selection=20 analysis identified a significant survival advantage for RPA Class = I=20 relative to RPA Class II, this significance was lost on Type I = final model=20 analysis, in which RPA Class II exhibited only a 4.9% higher = hazard rate=20 than RPA Class I (P =3D 0.625). = Asynchronous=20 presentation of brain metastasis was associated with improved = survival,=20 with synchronous presentation exhibiting a 24.9% higher hazard = rate in the=20 analysis including RPA class (P =3D = 0.044).=20 Synchronicity of presentation was not evaluated in the final model = analysis excluding RPA class because it was not significant in the = model=20 selection analysis excluding RPA class, although it did approach=20 significance in that model (P =3D = 0.068). Fewer=20 metastases was associated with improved survival, with each = additional=20 lesion exhibiting a 6.2% higher hazard rate in the analysis = excluding RPA=20 class (P =3D 0.014) and a 7.7% higher = hazard rate=20 in the analysis including RPA class (P = =3D 0.002).=20 Smaller total tumor volume was associated with improved survival, = with=20 each additional cm³ exhibiting a 1.2% = higher=20 hazard rate in the analysis excluding RPA class (P =3D 0.041). Total tumor volume was not = evaluated in=20 the final model analysis including RPA class because it was not=20 significant in the model selection analysis excluding RPA class, = although=20 it did approach significance in that model (P =3D=20 0.059). Prior surgery was associated with improved survival, = exhibiting a=20 28.0% lower hazard rate in the analysis excluding RPA class (P =3D 0.001) and a 26.9% lower hazard rate = in the=20 analysis including RPA class (P =3D = 0.002).=20 Finally, multiple SRS treatments was associated with improved = survival,=20 with each additional treatment exhibiting a 37.8% lower hazard = rate in the=20 analysis excluding RPA class (P =3D = 0.000) and a=20 37.2% lower hazard rate in the analysis including RPA class (P =3D 0.000) (Table=20 4).

TABLE 4. Hazard analysis = of factors=20 contributing to survival^a=20

Local Control TOP=20

Local control data were available = for 878=20 (56.0%) of 1569 brain metastases in 366 (59.1%) of 619 patients = treated=20 during the study interval. At the time of analysis, 740 metastases = were=20 well controlled (84.3%) and control of the remaining 138 = metastases=20 (15.7%) had been lost. The median actuarial local control was 7.67 = years=20 (95% confidence interval, 7.67 yr-greater than current length of=20 follow-up) from the time of SRS treatment. The 1- and 2-year = actuarial=20 local control estimates were 0.82 (95% confidence interval, = 0.80-0.84) and=20 0.72 (95% confidence interval, 0.69-0.75), respectively (Fig.=20 3).

FIGURE 3. Kaplan-Meier = plot=20 demonstrating actuarial local = control.

Model selection hazard analysis = identified=20 female sex, uncontrolled primary tumor, asynchronous presentation = of brain=20 metastasis, more metastases, and more SRS treatments to be = significantly=20 associated with longer duration of local control when adjusting = for their=20 respective previous variables in the model. Age, KPS score, tumor = type,=20 presence or absence of systemic metastases, RPA class, prior WBRT, = and SRS=20 dose were not significantly associated with local control = probability.=20 Tumor volume, eloquent location, and prior surgery were not = evaluated in=20 this analysis because these variables were not, for the purposes = of this=20 study, readily available on a per-lesion basis in cases of = multiple brain=20 metastases. However, a separate model selection hazard analysis = performed=20 on data from a subset of 167 single metastases, including all of = the=20 previously mentioned variables, found tumor volume (P =3D 0.591), eloquent location (P =3D 0.296), and prior surgery (P =3D 0.668) to be insignificant.

Type I final model hazard = analysis identified=20 the following: female sex was associated with longer duration of = local=20 control, with male sex exhibiting a 74.5% higher hazard rate (P =3D 0.001). Uncontrolled primary tumor = was associated=20 with longer duration of local control, with controlled primary = exhibiting=20 a 74.6% higher hazard rate (P =3D = 0.004). Multiple=20 metastases was associated with longer duration of local control, = with each=20 additional lesion exhibiting a 31.7% lower hazard rate in the = analysis=20 excluding RPA class (P =3D 0.000) and a = 32.1% lower=20 hazard rate in the analysis including RPA class (P =3D 0.000). Finally, multiple SRS = treatments was=20 associated with longer duration of local control, with each = additional=20 treatment exhibiting a 17.4% lower hazard rate in the analysis = excluding=20 RPA class (P =3D 0.005) and a 17.9% = lower hazard=20 rate in the analysis including RPA class (P =3D=20 0.002) (Table=20 5).

TABLE 5. Hazard analysis = of factors=20 contributing to local control^a=20

Regional Control TOP=20

Regional control data were = available for 366=20 (59.1%) of 619 patients treated during the study interval. Two = hundred=20 nineteen (59.8%) of these 366 patients received WBRT prior to=20 radiosurgery, and 147 (40.2%) did not. At the time of analysis, = regional=20 control had been maintained in 243 patients (66.4%) and lost in = the=20 remaining 123 patients (33.6%).

Model selection logistic analysis = identified=20 asynchronous presentation of brain metastasis, fewer brain = metastases,=20 eloquent tumor location, and prior WBRT to be significantly = associated=20 with improved regional control when adjusting for their respective = previous variables in the model. Prior surgery, higher SRS dose, = and=20 multiple SRS treatments were associated with impaired regional = control.=20 Sex, age, KPS score, tumor type, status of primary tumor = (controlled=20 versus uncontrolled), presence or absence of systemic metastases, = RPA=20 class, and total tumor volume were not significantly associated = with local=20 control probability.

Type I final model logistic = analysis=20 identified the following: fewer metastases was associated with = improved=20 regional control, with each additional lesion exhibiting a 13.8% = lower=20 odds ratio (P =3D 0.031). Eloquent = tumor location=20 was associated with improved regional control, exhibiting a 75.9% = higher=20 odds ratio than non-eloquent tumor location (P =3D=20 0.014). Prior WBRT was associated with improved regional control,=20 exhibiting a 68.5% higher odds ratio in the analysis including RPA = class=20 (P =3D 0.025). Prior WBRT was not = evaluated in the=20 final model analysis excluding RPA class because it was not = significant in=20 the model selection analysis excluding RPA class, although it did = approach=20 significance in that model (P =3D = 0.056). Prior=20 surgery was associated with impaired regional control, exhibiting = a 66.0%=20 lower odds ratio in the analysis excluding RPA class (P =3D 0.000) and a 66.6% lower odds ratio = in the=20 analysis including RPA class (P =3D = 0.000). Higher=20 SRS dose was associated with impaired regional control, with each = 1 Gy=20 increase in dose exhibiting a 18.6% lower odds ratio in the = analysis=20 excluding RPA class (P =3D 0.000) and a = 18.2% lower=20 odds ratio in the analysis including RPA class (P=20 =3D 0.000). Finally, multiple SRS treatments was associated with = impaired=20 regional control, with each additional treatment exhibiting a = 96.8% lower=20 odds ratio (P =3D 0.000). This is = because most=20 retreatments were to treat new metastases outside the original=20 radiosurgical field (Table=20 6).

TABLE 6. Logistic analysis = of factors=20 contributing to regional control^a=20

Complications TOP=20

Fourteen (2.3%) of 619 patients = developed=20 radiation necrosis a median of 8.7 months after treatment (mean, = 8.8 mo;=20 range, 3.1-17.3 mo). No patient developed more than one necrotic = focus,=20 and no lesion that ultimately became necrotic was treated with=20 radiosurgery more than once. Twelve patients underwent resection = of the=20 necrotic focus and were definitively diagnosed based on surgical=20 pathology; one of these patients subsequently underwent a second = operation=20 for the same necrotic focus. Two patients were diagnosed based on=20 characteristic symptoms and MRI findings and responded well to=20 corticosteroid therapy.

Forty-four (7.1%) of 619 patients = underwent=20 surgical resection for symptomatic tumor recurrence a median of = 6.3 months=20 after treatment (mean, 7.8 mo; range, 0.9-38.0 mo). One patient = underwent=20 a single operation for resection of two recurrent lesions, one = patient=20 underwent two operations for resection of two separate recurrent = lesions,=20 and one patient underwent three operations for the repeated = resection of=20 the same recurrent lesion. Ten resected recurrent tumors in as = many=20 patients were found to have features of radiation necrosis. One of = these=20 patients later underwent a second resection of the same recurrent = lesion,=20 at which time surgical pathology revealed only radiation necrosis. = This=20 patient was also included among the 14 patients with radiation = necrosis,=20 reported above. Three resected tumors in two patients-one = non-small cell=20 lung metastasis and two melanoma metastases-were hemorrhagic.

Two additional patients each = underwent=20 surgical resection of a lesion previously treated with = radiosurgery, 20.4=20 and 20.8 months after treatment, respectively. However, because = these=20 operations were performed at outside medical centers, the surgical = pathology reports were not available for review.

No data on acute or minor = complications=20 associated with radiosurgical treatment were collected.

DISCUSSION TOP=20

SRS was first described by = Leksell (44)=20 in 1951 as a method of using multiple, convergent radiation beams = to=20 deliver a high dose of radiation to a discrete intracranial target = with=20 the precision necessary to safely and effectively perform=20 minimally-invasive functional neurosurgery. Use of SRS in a = clinical=20 setting began in 1967 (43),=20 but the first report of SRS for metastatic brain disease was not = published=20 until 20 years later (46,84).=20 Since that time, tens of thousands of patients with metastatic = brain=20 tumors have been treated with both linear accelerator-based and = gamma=20 knife SRS, and the two platforms have proven equally safe and = effective=20 (2,79).

SRS is well suited for the = treatment of brain=20 metastases. It is minimally invasive and does not require general=20 anesthesia, making it a viable option for patients who cannot = tolerate=20 surgery, and is generally performed as an outpatient procedure. = Metastatic=20 lesions are typically spherical and clearly defined on = neuroimaging due to=20 contrast uptake, facilitating treatment planning. They are also = small and=20 histologically well defined, justifying the rapid fall-off of = radiation=20 dose at the periphery of the target volume generated by SRS (47).=20 SRS is readily applicable to lesions located nearly anywhere in = the brain,=20 including eloquent locations and deep brain locations that would = be=20 difficult or impossible to access surgically with acceptable = morbidity,=20 and can be used to treat multiple lesions in a single treatment = session=20 (7,26,28,32,33,54,70,71= ).=20 It has proven to be an effective treatment even for melanoma, = sarcoma, and=20 renal cell carcinoma metastases-lesions traditionally considered = to be=20 radioresistant (12,15,28,42,49,50,51,5= 7).=20 SRS may also be used in combination with WBRT and/or surgery, and = may be=20 repeated when necessary to treat recurrent disease (16,28,56,70,73).=20 Additionally, SRS has been shown to be more cost-effective than = surgery=20 (48,66).

Like all treatment modalities, = SRS has=20 disadvantages as well. It is incapable of rapidly resolving = symptoms due=20 to tumor mass effect, and surgical resection is generally a more=20 appropriate treatment for patients with symptomatic mass effect, = provided=20 that they are surgical candidates. SRS also has limited = applicability for=20 large metastatic lesions (=E2=89=A53 cm in diameter) due to = increasing radiation=20 exposure to normal brain with increasing target volume (73)=20 and for lesions located within close proximity of structures with = low=20 radiation tolerance, such as the optic nerves and chiasm (83).

Currently, the optimal role of = SRS in the=20 management of metastatic brain disease is not well defined. = Results of the=20 first prospective randomized trials of WBRT with or without SRS = have only=20 recently been reported (2,40),=20 and although they have shown that SRS offers a significant = survival=20 benefit for patients with a single brain metastasis, they have not = corroborated a large body of retrospective literature suggesting = that SRS=20 similarly improves survival for patients with multiple brain = metastases=20 (7,33,54,70,71).=20 Although large SRS series consistently report survival durations=20 comparable to those reported in surgical series, no randomized = prospective=20 comparison of these two treatment modalities has been conducted = (1,5,7,8,11,14,19,22,28= ,29,33,38,54,55,58-62,70,71,76,77,86,87,89,90).

Survival and Local Control = after SRS TOP=20

We report a median actuarial = survival of 7.9=20 months following SRS and 1- and 2-year actuarial survival = probabilities of=20 36 and 14%, respectively. These values are consistent with those = reported=20 in other large radiosurgical series, in which median survival = typically=20 ranges from about 7 to 13.5 months (1,5,7,11,14,19,22,28,2= 9,33,38,54,55,62,70,71,75,76,77,86,87,90),=20 and 1- and 2-year survival probabilities typically range from 31 = to 53%=20 and 13 to 30%, respectively (5,14,54,55,62,76,90),=20 but fall at the low end of these ranges. However, many of these = other=20 series were limited to solitary brain metastasis (5,22,77,87),=20 excluded patients with large tumor volumes (14,28,76),=20 or excluded patients with unfavorable characteristics (1,5,28,29).=20 The wider range of patient and tumor characteristics encompassed = in the=20 present study may account for the lower than expected median = survival. The=20 use of the SSDI to determine survival status may also contribute, = because=20 many patients who would otherwise have been considered lost to = follow-up=20 and excluded from the survival analysis were identified as = deceased in the=20 SSDI and were considered deceased for the purposes of this = analysis due to=20 the 100% specificity of the SSDI (30).

We also report a median local = control time of=20 7.67 years, a crude local control rate of 84.3%, and 1- and 2-year = actuarial local control probabilities of 82 and 72%, respectively. = These=20 values are consistent with those reported in other large = radiosurgical=20 series, in which crude local control rates typically range from 82 = to 96%=20 (5,9,19,22,28,29,33,35,= 54,55,69,75,77,87),=20 and 1- and 2-year local control probabilities typically range from = 64 to=20 91% and 46 to 77%, respectively (1,5,7,14,29,54,55,62,6= 8,71,82).

Andrews et al. (2)=20 recently reported the Phase III results of Radiation Therapy = Oncology=20 Group protocol 9508, the first prospective, randomized trial of = WBRT with=20 or without SRS for the treatment of metastatic brain disease to = reach full=20 accrual. Three hundred thirty-three patients with one to three = newly=20 diagnosed brain metastases, KPS score of 70 or greater, and no = history of=20 WBRT were randomly allocated either WBRT alone (167 patients) or = WBRT=20 followed by an SRS boost (164 patients). Analysis revealed a = significant=20 survival advantage for patients with a single brain metastasis = treated=20 with WBRT plus SRS rather than WBRT alone (median survival, 6.5 = versus 4.9=20 mo), but failed to reveal a similar advantage for patients with = two to=20 three brain metastases. However, all patients in the WBRT plus SRS = group=20 were significantly more likely to have a stable or improved KPS = score 6=20 months post-treatment, and were less likely to be dependent on=20 corticosteroids, than those treated with WBRT alone. Multivariate = analysis=20 revealed RPA Class I and non-small cell lung primary to be = significantly=20 associated with improved survival. Analysis also revealed = significantly=20 better 1-year local control of lesions treated with WBRT plus SRS, = compared to those treated with WBRT alone (82 versus 71%). The = authors=20 concluded that the addition of SRS to WBRT should be standard = treatment=20 for patients with a single brain metastasis and considered for = patients=20 with two to three brain metastases (2).=20 Kondziolka et al. (40)=20 had previously conducted a similar randomized prospective study, = in which=20 patients with two to four brain metastases and KPS score of 70 or = greater=20 were randomly allocated either WBRT alone or WBRT plus SRS, = however, this=20 study was stopped at an interim evaluation (60% accrual: 14 = patients in=20 the WBRT alone group and 13 in the WBRT plus SRS group) due to = detection=20 of significantly better local control following WBRT plus SRS = compared to=20 WBRT alone. Median time to local failure was 6 months following = WBRT alone=20 versus 36 months following WBRT plus SRS; 1 year after treatment, = no=20 patient in the WBRT group had maintained local control, compared = with 92%=20 of patients in the WBRT plus SRS group. Although interim analysis = revealed=20 a trend toward improved survival following WBRT plus SRS compared = to WBRT=20 alone (median survival durations of 11 mo and 7.5 mo, = respectively), no=20 conclusions could be made due to the small number of patients in = the study=20 (43).

Many groups have conducted = retrospective=20 analyses in order to determine patient characteristics and = treatment=20 variables associated with improved local control and survival = following=20 SRS for metastatic brain disease. In this study, many patient and=20 treatment variables-in fact nearly every patient and treatment = variable=20 assessed-were significantly associated with survival after SRS. = Many other=20 large retrospective analyses, involving hundreds of patients = and/or=20 metastatic brain tumors, most of which are based on gamma knife = series,=20 have found higher KPS score (usually =E2=89=A570) (17,20,31,34,38,55,62,7= 0,75,77),=20 controlled primary tumor (11,17,33,36,88),=20 absence of systemic metastatic disease (1,11,17,31,36,38,49,50= ,62,70,75,77,88),=20 better RPA Class (7,12,33,54-56),=20 and asynchronous presentation of brain metastases (21,75)=20 to be significantly associated with better survival (Table=20 7). Our results are consistent with these findings. However, = even the=20 largest reviews sometimes produce discordant results, and the = prognostic=20 value of age and number of metastases, in particular, remains = unclear.=20 Although numerous studies, including this one, have found younger = age=20 (usually <65) to be significantly associated with improved = survival (1,7,20),=20 Noel et al. (55)=20 recently reviewed 117 patients aged 65 years or older treated with = linear=20 accelerator-based SRS for 227 brain metastases and found a median = survival=20 of 8 months and 1- and 2-year survival rates of 31 and 13%, = respectively.=20 These results are comparable to those of numerous studies = including=20 patients of all ages. Similarly, although numerous studies, = including this=20 one, have found fewer brain metastases to be significantly = associated with=20 improved survival (1-3),=20 Bhatnagar et al. (7)=20 recently reviewed 205 patients treated with gamma knife SRS for = four or=20 more brain metastases each and found a median survival of 8 = months. These=20 results are comparable to median survivals reported for series of = patients=20 with solitary brain metastasis or oligometastatic disease. Nam et = al. (54)=20 found no significant survival difference in a retrospective = comparison of=20 SRS for four or more brain metastases versus one to three = metastases.=20 Interestingly, several retrospective reviews have determined that = smaller=20 total tumor volume is significantly associated with improved = survival,=20 independent of the number of metastases (7,33,62).=20 Smaller total tumor volume is often also significantly associated = with=20 improved local control (1,7,15,33,62,75,76,88<= /A>),=20 although these are not independent variables, because smaller = lesions can=20 be safely treated with higher doses (73).=20 Interestingly, neither smaller tumor volume nor higher SRS dose = was=20 associated with improved local control in the present study.

TABLE 7. Large = retrospective reviews=20 of stereotactic radiosurgery for metastatic brain = diseasea =

SRS with or without WBRT TOP=20

In the present study, WBRT prior = to SRS was=20 associated with improved regional control, but not improved local = control=20 or survival. Patients treated with SRS for metastatic brain = disease often=20 also receive WBRT. The rationale behind this treatment strategy is = the=20 assumption that WBRT will effectively treat micrometastatic foci = outside=20 the radiosurgical field, hopefully preventing them from = progressing to=20 become detectable, symptomatic brain lesions that may threaten = duration=20 and quality of life. However, there is evidence that the addition = of WBRT=20 to SRS is associated with an increased incidence of mild to = moderate=20 treatment-related morbidity and there is concern that WBRT may = also cause=20 serious neurological deficit in some long-term survivors. = Kondziolka et=20 al. (39)=20 reported the results of 104 surveys completed by patients treated = with SRS=20 with or without WBRT and/or their families. They found that = patients=20 treated with SRS plus WBRT were significantly more likely to = report=20 fatigue, concentration problems, short- and long-term memory = problems, and=20 disorders of mood, compared with patients treated with SRS alone = (39).=20 DeAngelis et al. (18)=20 reported 12 patients cured of brain metastases who developed = progressive=20 dementia, ataxia, and urinary incontinence a median of 14 months = (range,=20 5-36 mo) after treatment with WBRT (total dose, 25-39 Gy delivered = in 3-6=20 Gy fractions) with or without surgery. They found a 1.9 to 5.1% = incidence=20 of WBRT-induced dementia in their study population and estimated = that the=20 true incidence is probably higher (18).=20 As primary treatment of metastatic brain disease continues to = become more=20 and more effective, the same patients who will benefit most from = freedom=20 from new brain metastases will be at the greatest risk of = developing=20 long-term treatment-related morbidities. Accordingly, many studies = have=20 sought to better define the effect of WBRT on survival, local = control, and=20 regional control following SRS.

Aoyama et al. (3)=20 recently reported results of the first prospective, = multi-institutional,=20 randomized trial of SRS with or without WBRT for the treatment of=20 metastatic brain disease. One hundred thirty-two patients with one = to four=20 brain metastases 3 cm or more in diameter and KPS scores of 70 or = greater=20 were randomly allocated either SRS plus WBRT (65 patients) or SRS = alone=20 (67 patients). Adjuvant WBRT did not affect survival; patients who = received both SRS and WBRT had a median survival of 7.5 months and = a=20 1-year actuarial survival rate of 38.5%, compared to 8.0 months = and 28.4%=20 for those treated with SRS alone. However, patients who did not = initially=20 receive WBRT underwent significantly more salvage procedures. = Multivariate=20 analysis revealed age younger than 65 years, controlled primary = tumor,=20 stable systemic disease, and KPS score of 90 or greater to be=20 significantly associated with improved survival. Analysis revealed = significantly better local, regional, and total brain control = following=20 treatment with SRS plus WBRT rather than SRS alone. Patients = treated with=20 SRS plus WBRT had 1-year actuarial local control, regional = control, and=20 total brain control rates of 88.7, 58.5, and 53.2%, respectively, = compared=20 with 72.5, 36.3, and 23.6%, respectively, for patients treated = with SRS=20 alone. Multivariate analysis also revealed stable systemic disease = and KPS=20 score of 80 or greater to be significantly associated with = improved=20 regional control, and a single brain metastasis approached = significance=20 (P =3D 0.06). Analysis revealed no = difference in=20 post-treatment neurological performance; patients treated with SRS = plus=20 WBRT had a 1-year actuarial neurological preservation rate of = 72.1%,=20 compared with 70.3% for patients treated with SRS alone (3).=20 These findings are consistent with the only prospective, = randomized trial=20 to date of the role of postoperative WBRT in the management of = metastatic=20 brain disease; in their trial, Patchell et al. (60)=20 randomized 95 patients with a single resectable brain metastasis = and KPS=20 score of 70 or greater to surgery plus WBRT or surgery alone and=20 determined that patients who received postoperative WBRT did not = have a=20 survival advantage but did have significantly better local, = regional, and=20 total brain control.

Two similar randomized, = prospective trials of=20 SRS with or without WBRT for the treatment of metastatic brain = disease are=20 currently underway. North Central Cancer Treatment Group protocol = N0574=20 will randomize 528 patients with one to three newly diagnosed = brain=20 metastases to treatment with SRS plus WBRT or SRS alone. Primary = outcome=20 will be duration of survival; secondary outcomes will be time to = CNS=20 failure, quality of life, duration of functional independence, = long-term=20 neurocognitive status, and post-treatment toxicity (13).=20 This trial is a continuation of American College of Surgeons = Oncology=20 Group protocol Z0300, which was inactive following suspension of=20 enrollment in December 2004 (4).=20 European Organization for Research and Treatment of Cancer = protocol 22952=20 to 26001 will accumulate 340 patients with one to three newly = diagnosed=20 brain metastases. One hundred seventy of these patients will have=20 undergone complete surgical resection and will then be randomized = to=20 postoperative WBRT or no postoperative WBRT. The other 170 = patients will=20 be randomized to SRS plus WBRT or SRS alone. Primary outcome will = be=20 duration of survival with a World Health Organization performance = status=20 of 2 or less, secondary outcomes will be duration of survival, = duration of=20 progression-free survival, time to neurological progression, acute = toxicity, late toxicity, and quality of life (53).

Due to the current paucity of = Class I data,=20 several groups have conducted matched nonrandomized prospective = and=20 retrospective studies of SRS with or without WBRT for metastatic = brain=20 disease (Table=20 8). They have consistently determined that the addition of = WBRT to SRS=20 does not affect survival, but may affect local and regional = control (45,64,78,79).=20 Results of this nature have led some authors to conclude that it = may be=20 best to initially employ SRS as a monotherapy and reserve WBRT for = possible salvage treatment (19,35,38,39,52).=20 Many centers, including the University of Florida, often use = repeat SRS in=20 lieu of WBRT for salvage treatment (16).

TABLE 8. Stereotactic = radiosurgery=20 with or without whole brain radiation therapy: Class II and = III=20 data^a =

Study Limitations TOP=20

This analysis is limited in that = it is based=20 on the experience of a single institution. As with any = retrospective=20 analysis, selection bias may be present. It is further limited by = the lack=20 of data on post-SRS quality of life and cause of death. Such data = are very=20 difficult to obtain reliably for large retrospective analyses such = as this=20 one, but must be an important component of any prospective = analysis of=20 SRS, and in particular, any prospective trial comparing SRS to = surgery.=20 Additionally, only 56% of all treated lesions could be included in = the=20 local control analysis and 59.1% of patients in the regional = control=20 analysis, due to lack of availability of follow-up imaging as a = result of=20 the large distances many patients traveled to reach the University = of=20 Florida for treatment.

CONCLUSIONS TOP=20

Linear accelerator-based SRS is a = safe and=20 effective treatment for patients with metastatic brain tumors. = Female sex,=20 younger age, higher KPS score, controlled primary tumor, absence = of=20 systemic metastases, RPA Class I or II, asynchronous presentation = of brain=20 metastasis, fewer brain metastases, smaller total volume of brain=20 metastases, surgery prior to radiosurgery, and multiple = radiosurgical=20 treatments were associated with improved survival. Melanoma = histology was=20 associated with impaired survival. This underscores the complexity = of=20 selecting patients most likely to benefit from radiosurgical = management of=20 metastatic brain disease.

In this single-institution, = nonrandomized,=20 retrospective study, the use of WBRT in combination with SRS = improved=20 regional control, but did not improve local control or = survival.

Acknowledgments TOP=20

Bradley M. Swinson, B.S., was = supported in=20 part by National Institutes of Health grant T35 HL07489. The = authors are=20 very grateful to Qin A. Li, Ph.D., and Keith E. Muller, Ph.D., = Department=20 of Epidemiology and Health Policy Research, University of Florida, = Gainesville, for their skillful statistical analysis.

REFERENCES TOP=20

1. Alexander E 3rd, Moriarty TM, = Davis RB, Wen=20 PY, Fine HA, Black PM, Kooy HM, Loeffler JS: Stereotactic = radiosurgery for=20 the definitive, noninvasive treatment of brain metastases. J Natl Cancer Inst 87:34-40, 1995.=20

[Medl= ine=20 Link] [CrossRef]=20 [Context=20 Link]

2. Andrews DW, Scott CB, Sperduto = PW, Flanders=20 AE, Gaspar LE, Schell MC, Werner-Wasik M, Demas W, Ryu J, Bahary = JP,=20 Souhami L, Rotman M, Mehta MP, Curran WJ Jr: Whole brain radiation = therapy=20 with or without stereotactic radiosurgery boost for patients with = one to=20 three brain metastases: Phase III results of the RTOG 9508 = randomised=20 trial. Lancet 363:1665-1672, = 2004.=20

[Med= line=20 Link] [CrossRef]=20 [Context=20 Link]

3. Aoyama H, Shirato H, Tago M, = Nakagawa K,=20 Toyoda T, Hatano K, Kenjyo M, Oya N, Hirota S, Shioura H, Kunieda = E,=20 Inomata T, Hayakawa K, Katoh N, Kobashi G: Stereotactic = radiosurgery plus=20 whole-brain radiation therapy vs stereotactic radiosurgery alone = for=20 treatment of brain metastases: A randomized controlled trial. = JAMA 295:2483-2491, 2006.=20

[Medline=20 Link] [CrossRef]=20 [Context=20 Link]

4. Asher A: American College of = Surgeons=20 Oncology Group study synopsis Z0300. http:/= /www.acosog.org/studies/synopses/Z0300_Synopsis.pdf=20 . Accessed 8/18/2006.=20