comparison of prognostic indices in patients who undergo

CLINICAL ARTICLEJ Neurosurg 128:14–22, 2018

Stereotactic radiosurgery (SRS) is a minimally invasive treatment option for patients with brain metastases, which occur in 10%–40% of patients

with melanoma.1,16, 17, 19,20 The management options for melanoma brain metastases include resection in cases of symptomatic surgically accessible tumors, whole-brain radiotherapy (WBRT), chemotherapy, use of molecularly targeted antitumor agents,11,21 and other novel treatment modalities.3,7,13 In recent years, WBRT has been relegated

to patients with miliary or meningeal melanoma brain metastases, because melanoma is thought to be relatively radioresistant.4,6

Several prognostic grading systems for the survival of patients with brain metastases have been proposed. These prognostic indices are commonly used as a guide for treat-ment decision making and for trial eligibility. The recur-sive partitioning analysis (RPA) classification5 is one of the major grading systems for assessing the prognosis of

ABBREVIATIONS BSBM = Basic Score for Brain Metastases; DS-GPA = Diagnosis-Specific Graded Prognostic Assessment; HR = hazard ratio; KPS = Karnofsky Perfor-mance Scale; MST = median survival time; OS = overall survival; RPA = recursive partitioning analysis; SIR = Score Index for Radiosurgery; SRS = stereotactic radiosur-gery; WBRT = whole-brain radiation therapy.SUBMITTED April 19, 2016. ACCEPTED September 15, 2016.INCLUDE WHEN CITING Published online January 20, 2017; DOI: 10.3171/2016.9.JNS161011.

Comparison of prognostic indices in patients who undergo melanoma brain metastasis radiosurgeryHideyuki Kano, MD, PhD,1,3 Alejandro Morales-Restrepo, BA,5 Aditya Iyer, MD,6 Gregory M. Weiner, MD,1 Seyed H. Mousavi, MD,1 John M. Kirkwood, MD,4 Ahmad A. Tarhini, MD, PhD,4 John C. Flickinger, MD,1–3 and L. Dade Lunsford, MD1,3

Departments of 1Neurological Surgery and 2Radiation Oncology, 3Center for Image-Guided Neurosurgery, and 4Division of Hematology/Oncology, Department of Medicine, 5University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania; and 6Department of Neurological Surgery, Stanford University, Stanford, California

OBJECTIVE The goal of this study was to use 4 prognostic indices to compare survival times of patients who under-went Gamma Knife stereotactic radiosurgery (SRS) to treat melanoma brain metastases.METHODS The authors analyzed 422 consecutive patients (1440 brain metastases) who underwent Gamma Knife SRS. The median total brain tumor volume was 4.7 cm3 (range 0.3–69.3 cm3), and the median number of metastases was 2 (range 1–32). One hundred thirty-two patients underwent whole-brain radiation therapy. Survival times were compared using recursive partitioning analysis (RPA), the Score Index for Radiosurgery (SIR), the Basic Score for Brain Metastases (BSBM), and the Diagnosis-Specific Graded Prognostic Assessment (DS-GPA).RESULTS The overall survival times after SRS were compared. With the RPA index, survival times were 2.6 months (Class III, n = 27), 5.5 months (Class II, n = 348), and 13.0 months (Class I, n = 47). With the DS-GPA index, survival times were 2.8 months (Scores 0–1, n = 67), 4.2 months (Scores 1.5–2.0, n = 143), 6.6 months (Scores 2.5–3.0, n = 111), and 9.4 months (Scores 3.5–4.0, n = 101). With the SIR, survival times were 3.2 months (Scores 0–3, n = 56), 5.8 months (Scores 4–7, n = 319), and 12.7 months (Scores 8–10, n = 47). With the BSBM index, survival times were 2.6 months (BSBM0, n = 47), 5.4 months (BSBM1, n = 282), 11.0 months (BSBM2, n = 86), and 8.8 months (BSBM3, n = 7). The DS-GPA index was the most balanced by case numbers in each class and provided the overall best prognostic index for overall survival.CONCLUSIONS The DS-GPA index proved most balanced and predictive of survival for patients with melanoma who underwent SRS as part of management for brain metastases. Patients whose DS-GPA score was ≥ 2.5 had predictably improved survival times after SRS.https://thejns.org/doi/abs/10.3171/2016.9.JNS161011KEY WORDS melanoma; brain metastases; stereotactic radiosurgery; Gamma Knife; recursive partitioning analysis; Score Index for Radiosurgery; Basic Score for Brain Metastases; Graded Prognostic Assessment; oncology

J Neurosurg Volume 128 • January 201814 ©AANS 2018, except where prohibited by US copyright law

Unauthenticated | Downloaded 12/13/21 05:32 PM UTC

Stereotactic radiosurgery for melanoma brain metastases

J Neurosurg Volume 128 • January 2018 15

patients with brain metastases. Other prognostic indices, such as the Score Index for Radiosurgery (SIR)21 and the Basic Score for Brain Metastases (BSBM),12 were subse-quently developed. The Graded Prognostic Assessment (GPA) was updated recently to the Diagnosis-Specific GPA (DS-GPA) index.18 In this study, we compared the predictive accuracy of these 4 major brain metastasis prognostic indices in patients with metastatic brain mela-noma.

MethodsData Collection

The data were collected retrospectively in a review of patient medical records. When not available from our medical records, patient survival data were obtained from the Internet-based Social Security Index. This study was approved by University of Pittsburgh Institutional Review Board.

Patient PopulationBetween August 1987 and December 2012, 444 con-

secutive patients underwent Gamma Knife SRS for mela-noma brain metastases at the University of Pittsburgh Medical Center. Twenty-two patients were excluded from the analysis because we could not obtain their complete data, which left a final series of 422 patients (1440 brain metastases). Additional therapies included systemic che-motherapy in 203 (48%) patients, immunotherapy (e.g., interferon-a and interleukin 2) in 207 (49%) patients, local extracranial radiation therapy in 56 (13%) patients, and vaccine therapy in 39 (9%) patients. Only 1 patient with BRAF mutation underwent vemurafenib molecularly targeted antitumor agent therapy, and this patient died 3.7 months after SRS as a result of the progression of numer-ous new brain metastases. Demographics and clinical data of the patient population are shown in Table 1.

For the management of brain metastases, Gamma Knife SRS was used as the primary option in 252 (60%) patients. One hundred thirty-two (31%) patients underwent initial WBRT before SRS, typically 30 Gy (range 21–60 Gy) in 10–14 fractions. Sixty-five (15%) patients under-went craniotomy before SRS, 4 (1%) patients underwent needle cyst aspiration, and 14 (3%) patients underwent ste-reotactic biopsy.

Grading SystemsRPA5 consists of the following 3 classes: I (patients <

65 years old, Karnofsky Performance Scale [KPS] score ≥ 70, controlled primary tumor, and no extracranial me-tastases), II (all patients not at Class I or III), and III (KPS score < 70). The SIR21 was obtained by summing the points from the following 5 variables: age (0, ≥ 60 years; 1, 51–59 years; 2, ≤ 50 years); KPS score at the time of SRS (0, ≤ 50; 1, 60–70; 2, 80–100); systemic disease sta-tus (0, progressive; 1, stable or partial remission; 2, com-plete remission or no evidence of disease); volume of the largest lesion (0, > 13 cm3; 1, 5–13 cm3; 2, < 5 cm3); and number of lesions (0, ≥ 3; 1, 2; 2, 1). Each SIR score was subgrouped as 0–3, 4–7, or 8–10. The BSBM12 was ob-tained by summing the points of the following 3 variables:

TABLE 1. Demographics and clinical data

Characteristic/Demographic

All Patients

Patients Who Underwent Only SRS

No. % No. %

Sex Male 284 67 165 65 Female 138 33 87 35 Age ≥65 yrs 145 34 167 66 <65 yrs 277 66 85 34No. of metastases 1 142 34 90 36 2–3 131 31 73 29 4–6 81 19 51 20 ≥7 68 16 38 15 Initial cerebral presentation Staging MRI 223 53 169 67 Seizures 42 10 23 9 Tumor hemorrhage 40 9 12 5 Mass effect (w/o bleed) 117 28 48 19 Location of metastases Lobar supratentorial 1201 83 760 84 Deep supratentorial 73 5 34 4 Cerebellum 121 8 78 9 Brainstem 29 2 17 2 Cranium 16 1 15 2 Previous CNS therapy WBRT 132 31 0 0 Craniotomy 65 15 0 0 Cyst aspiration 4 1 0 0 Stereotactic biopsy 14 3 0 0Previous systemic therapy Chemotherapy 203 48 124 49 Immunotherapy 207 49 131 52 Extracranial radiation 56 13 37 15 Extent of systemic disease CNS only 24 6 11 4 Primary site only 38 9 23 9 Primary + 1 lymph node chain 45 11 29 12 Primary + >1 lymph node

chain or visceral85 20 47 19

Disseminated (>2 visceral sites)

230 55 142 56

Systemic disease status Active 313 74 187 74 Controlled 109 26 65 26 Main neurological symptom None 227 54 166 66 Headache only 38 9 19 8 Seizures 18 4 9 4 Focal deficits 119 28 50 20

CONTINUED ON PAGE 16 »


H. Kano et al.

J Neurosurg Volume 128 • January 201816

KPS score (0, 50–70; 1, 80–100); control of primary tumor (0, no; 1, yes); and extracranial metastases (0, yes; 1, no). The DS-GPA18 was obtained by summing the points of the following 2 variables: KPS score (0, < 70; 1, 70–80; 2, 90–100) and number of brain metastases (0, > 3; 1, 2–3; 2, 1). We evaluated the prognostic grading scales based on the distribution of patient populations in each class of the grading scale and differences that appeared when pairs of adjacent classes in the same scale were compared. The best prognostic grading scales demonstrated a well-balanced patient population in each class and large differ-ences in outcomes between adjacent pairs of classes.

Stereotactic Radiosurgery ProceduresA dose plan that conformed closely to tumor volume

was designed using dose-planning software (Kula or Gamma Plan) supplied by the Gamma Knife manufac-turer (Elekta AB). The median dose to margin was 18 Gy (range 10–22 Gy), and the median maximal dose was 33.3 Gy (range 20–50 Gy). The doses used were based on tu-mor location, tumor volume, and history of WBRT. The median target volume for the individual tumors was 1.4 cm3 (range 0.03–37.2 cm3), and the median target volume for the largest tumor was 3.3 cm3. The median total tumor volume treated per procedure was 4.7 cm3 (range 0.03–69.3 cm3). SRS was performed using a Leksell Gamma Knife model U, B, or C or the Perfexion model.

Statistical AnalysisThe SPSS 22 software (IBM Corp.) was used for sta-

tistical analyses. Kaplan-Meier analyses were performed to assess overall survival (OS). The log-rank test for cat-egorical data and the Cox proportional hazards model for continuous data were performed as univariate analyses. Multivariate analyses were performed to assess the prog-nostic value of different variables using the Cox propor-tional hazard.

ResultsPatient Survival

At the time of assessment, 383 patients were de-ceased and 39 were alive. Among the deceased patients, 114 (30%) died as a result of brain disease progression, 10 (3%) patients died as a result of intracranial hemor-rhage from brain metastases, 93 (24%) died as a result of

systemic disease, and 156 (41%) died from an unknown cause. The median survival time (MST) from diagnosis of the primary tumor was 47.7 months (95% CI 41.25–52.89 months). The MST from the diagnosis of brain metastases was 8.3 months (95% CI 7.34–9.26 months). The MST af-ter SRS was 5.7 months (95% CI 4.97–6.37 months). The OS rates after SRS were 71.3% at 3 months, 47.7% at 6 months, 25.8% at 12 months, 11.8% at 24 months, 5.0% at 36 months, and 1.8% at 5 years. Results of the univariate and multivariate analyses are shown in Table 2. Patient survival was not significantly associated with the year in which patients underwent their SRS (p = 0.119 [1988–2007 vs 2008–2012]; p = 0.853 [1988–2009 vs 2010–2012]).

In the entire group, 132 (31%) patients underwent WBRT before SRS, 65 (15%) had undergone craniotomy, 4 (1%) had undergone cyst aspiration, and 14 (3%) had undergone diagnostic stereotactic biopsy. In the group of patients who underwent SRS as a primary CNS treatment (252), 225 pa-tients died and 27 survived. The MST from diagnosis of the primary tumor was 47.1 months (95% CI 38.07–56.07 months), and the MST from the diagnosis of brain metasta-ses was 7.9 months (95% CI 6.78–9.09 months). The MST after SRS was 6.4 months (95% CI 5.52–7.34 months). The OS rates after SRS were 74.7% at 3 months, 52.2% at 6 months, 27.9% at 12 months, 13.0% at 24 months, 6.1% at 36 months, and 2.3% at 5 years. Results of the univariate and multivariate analyses are shown in Table 2.

Prognostic Grading SystemsIn the entire series, the OS curves after SRS accord-

ing to the 4 major grading systems are shown in Fig. 1. As shown in Table 3, the RPA, SIR, and BSBM grading systems had large discrepancies in their patient numbers. A total of 82% of the patients were classified at RPA Class II, 76% at SIR Scores 4–7, and 67% at BSBM1. However, with DS-GPA grading, the patient population was better distributed into each score group (Scores 0–1, 16%; Scores 1.5–2, 34%; Scores 2.5–3, 26%; Scores 3.5–4, 24%).

The MSTs after SRS according to the RPA grading scales were 13.0 months (Class I), 5.5 months (Class II), and 2.6 months (Class III) (Table 3). The RPA grading re-vealed statistically significant differences when pairs of ad-jacent classes were compared (p = 0.001, hazard ratio [HR] 1.8 [RPA Class I vs II]; p = 0.004, HR 1.8 [Class II vs III]). The OS rates 6 months after SRS were 81% in RPA Class I, 45% in RPA Class II, and 22% in RPA Class III (Table 4).

The MSTs after SRS according to the SIR system were 3.2 months (Scores 1–3), 5.8 months (Scores 4–7), and 12.7 months (Scores 8–10). The SIR revealed statistically significant differences when pairs of adjacent SIR groups were compared (p = 0.004, HR 1.6 [Scores 8–10 vs 4–7; p = 0.002, HR 1.6 [Scores 4–7 vs 0–3]). The OS rates 6 months after SRS were 68% (Scores 8–10), 48% (Scores 4–7), and 29% (Scores 0–3) (Table 4).

The MSTs after SRS according to the BSBM classifica-tion were 2.6 months (BSBM0), 5.4 months (BSBM1), 11.0 months (BSBM2), and 2.6 months (BSBM3). The BSBM grading revealed statistically significant differences when pairs of adjacent BSBM grading groups were compared, except for BSBM3 vs BSBM2 (p = 0.994 [BSBM3 vs BSBM2]; p = 0.0001, HR 1.7 [BSBM2 vs BSBM1]; p =

TABLE 1. Demographics and clinical data

Characteristic/Demographic

All Patients

Patients Who Underwent Only SRS

No. % No. % Cognitive deficits 20 5 8 3

KPS score 90–100 267 63 179 71 70–80 128 30 61 24 <70 27 6 12 5

» CONTINUED FROM PAGE 15




0.0001, HR 2.2 [BSBM1 vs BSBM0]). The OS rates 6 months after SRS were 57% (BSBM3), 72% (BSBM2), 45% (BSBM1), and 27% (BSBM0) (Table 4).

The MSTs after SRS according to the DS-GPA were 2.9 months (Scores 0–1), 4.2 months (Scores 1.5–2), 6.6 months (Scores 2.5–3), and 9.4 months (Scores 3.5–4). The DS-GPA grading revealed statistically significant dif-ferences when pairs of adjacent classes were compared (p = 0.048, HR 1.3 [Scores 3.5–4 vs 2.5–3]; p = 0.010, HR 1.4 [Scores 2.5–3 vs 1.5–2]; p = 0.041, HR 1.4 [Scores 1.5–2 vs 0–1]). The OS rates 6 months after SRS were 66% (Scores 3.5–4), 54% (Scores 2.5–3), 40% (Scores 1.5–2), and 27% (Scores 0–1) (Table 4).

Stereotactic Radiosurgery as the Primary Strategy for Treating Brain Metastasis

In the group of patients who underwent SRS as the primary management for brain metastases (n = 242), the post-SRS OS curves are shown in Fig. 2 according to the 4 major grading systems. As shown in Table 3, the RPA, SIR, and BSBM grading systems had large discrepancies in patient numbers among their groups. A total of 83% of the patients were classified at RPA Class II, 78% at SIR Scores 4–7, and 69% at BSBM1. However, with DS-GPA

grading, the patient population was better distributed into each score group (Scores 0–1, 12%; Scores 1.5–2, 34%; Scores 2.5–3, 27%; Scores 3.5–4, 27%).

The MSTs after SRS according to RPA were 12.4 months (Class I), 5.7 months (Class II), and 2.1 months (Class III) (Table 3). The RPA grading revealed statisti-cally significant differences when RPA Classes I and II were compared (p = 0.005), whereas RPA Class II was not statistically different from RPA Class III (p = 0.296). RPA Class I was significantly associated with a longer survival time than was RPA Class III (p = 0.013). The OS rates 6 months after SRS were 79% (Class I), 49% (Class II), and 33% (Class III) (Table 4).

The MSTs after SRS according to the SIR were 3.1 months (Scores 1–3), 6.4 months (Scores 4–7), and 15.4 months (Scores 8–10). The SIR scores revealed statistical-ly significant differences when pairs of adjacent SIR score groups were compared (p = 0.005 [Scores 8–10 vs 4–7; p = 0.024 [Scores 4–7 vs 0–3]). The OS rates 6 months after SRS were 71% (Scores 8–10), 55% (Scores 4–7), and 26% (Scores 0–3) (Table 4).

The MSTs after SRS according to the BSBM were 2.0 months (BSBM0), 6.1 months (BSBM1), 12.4 months (BSBM2), and 3.7 months (BSBM3). BSBM grading re-vealed statistically significant differences when pairs

TABLE 2. Prognostic values for patient survival

VariableHigh-Risk

GroupUnivariate Analysis Multivariate Analysis

p Value HR (95% CI) p Value HR (95% CI)

Entire group Age Older 0.148 1.01 (0.99–1.01) 0.191 NA Sex Male 0.450 0.92 (0.75–1.14) NA NA KPS score Lower <0.0001 0.99 (0.98–0.99) 0.001 0.986 (0.978–0.994) No. of brain tumors Larger no. <0.0001 1.07 (1.05–1.10) <0.0001 1.067 (1.044–1.091) Largest tumor Larger 0.012 1.02 (1.01–1.03) 0.241 NA Total tumor vol Larger <0.0001 1.02 (1.01–1.03) 0.006 1.016 (1.005–1.028) Previous chemotherapy No 0.001 1.42 (1.16–1.74) 0.125 NA Previous immunotherapy No 0.508 0.93 (0.76–1.14) NA NA Previous WBRT Yes <0.0001 1.57 (1.263–1.944) 0.002 1.411 (1.129–1.764) Interval from primary cancer diagnosis to brain metastasis Shorter 0.158 1.001 (1.000–1.002) 0.468 NA Neurological symptoms Yes 0.064 1.22 (0.99–1.51) NA NA Extracranial status Active <0.0001 1.96 (1.55–2.48) <0.0001 2.200 (1.725–2.806)SRS-alone group Age Older 0.047 1.01 (1.00–1.02) 0.191 NA Sex Male 0.781 0.96 (0.73–1.26) NA NA KPS score Lower 0.028 0.99 (0.98–0.99) 0.023 0.987 (0.975–0.998) No. of brain tumors Larger no. <0.0001 1.07 (1.04–1.10) <0.0001 1.067 (1.039–1.095) Largest tumor vol Larger 0.024 1.03 (1.00–1.05) 0.057 NA Total tumor vol Larger <0.0001 1.04 (1.02–1.06) 0.005 1.027 (1.008–1.046) Previous chemotherapy No <0.0005 1.62 (1.24–2.12) 0.032 1.355 (1.026–1.798) Previous immunotherapy No 0.481 0.91 (0.70–1.18) NA NA Interval from primary cancer diagnosis to brain metastasis Shorter 0.344 1.001 (0.99–1.002) NA NA Neurological symptoms Yes 0.437 0.437 (0.68–1.18) NA NA Extracranial status Active <0.0001 1.95 (1.44–2.66) <0.0001 2.158 (1.559–2.987)

NA = not applicable.


H. Kano et al.


of adjacent BSBM groups were compared (p = 0.016 [BSBM3 vs BSBM2]; p = 0.001 [BSBM2 vs BSBM1]; p = 0.0001 [BSBM1 vs BSBM0]). The OS rates 6 months after SRS were 33% (BSBM3), 74% (BSBM2), 50% (BSBM1), and 11% (BSBM0) (Table 4).

The MSTs after SRS according to the DS-GPA were 2.8 months (Scores 0–1), 5.0 months (Scores 1.5–2), 6.6 months (Scores 2.5–3), and 9.7 months (Scores 3.5–4). DS-GPA grading revealed statistically significant differences when Scores 3.5–4 vs 2.5–3 were compared (p = 0.043), whereas comparisons of Scores 2.5–3 vs 1.5–2 and Scores 1.5–2 vs 0–1 were not statistically different. DS-GPA Scores 3.5–4 were significantly associated with longer sur-vival times than were DS-GPA Scores 1.5–2 (p < 0.0001,

HR 2.0 [95% CI 1.48–2.59]) and Scores 0–1 (p < 0.0001, HR 3.0 [95% CI 2.12–4.20]). The OS rates 6 months after SRS were 65% (Scores 3.5–4), 54% (Scores 2.5–3), 47% (Scores 1.5–2), and 31% (Scores 0–1) (Table 4).

DiscussionRole of Prognostic Indices

Brain metastases are a heterogeneous entity, and sur-vival time is related to the type of primary cancer, system-ic control, type of treatment, and responses to treatment. Factors that affect prognosis are relevant to therapeutic decision making. Survival predictions can be difficult when they are applied to individual patients, so the quest

FIG. 1. Kaplan-Meier curves for OS of the patients for each index according to individual class in patients with melanoma brain metastases in the entire series (n = 422). MST values are in months.




for the best system is important.9 The prognostic indices for brain metastases can identify patients whose MST after SRS can be estimated with greater accuracy. The RPA classification was the earliest and most commonly validated predictive grading system. The RPA classifica-tion was based on patients who underwent WBRT, but it was reported recently to be applicable also to patients who underwent SRS.8,10,15,23,24 In most reports, the largest num-bers of patients are found to be at RPA Class II. In our study, 82% of the patients with melanoma were classified at RPA Class II. The RPA classes were highly correlated with patient survival (p = 0.001 [Class I vs II]; p = 0.004 [Class II vs III]) (Table 3). The SIR requires a more com-plex calculation because it consists of 5 prognostic fac-tors. Moreover, the SIR requires a volume of the largest lesion at the time of SRS. Sperduto et al.18 suggested that the role of prognostic indices is to better predict outcome to guide decision making before and during treatment. Volume measurement is usually performed at the time of SRS, which makes its prognostic scoring difficult before treatment. The SIR also suffers from an unbalanced pa-

tient proportion within each prognostic class. In our study, 76% of the patients were classified as having an SIR score of 4–7. In fact, all 3 SIR score groups (Scores 8–10, 4–7, and 0–3) were highly correlated with patient survival (p = 0.004 [Scores 8–10 vs 4–7]; p = 0.002 [Scores 4–7 vs 0–3]) (Table 3).

The BSBM consists of the following 3 systemic prog-nostic factors: KPS score, control of primary tumor, and presence of extracranial metastases. Most patients are found to have a BSBM score of 1. In our study, 67% of the patients were classified at BSBM1, and only 2% of the patients were classified at BSBM3. When compared with BSBM2, BSBM3 was not associated with patient survival (p = 0.994), because too few patients were in the BSBM3 group (Table 3). Sperduto et al.18 mentioned that prognostic factors for patients with brain metastases varied according to diagnosis, and for each diagnosis, a robust separation into different GPA scores was discerned, which implies considerable heterogeneity in outcome. Therefore, they up-dated their initial GPA to the DS-GPA. Although the RPA was based on patients who underwent WBRT, the DS-GPA

TABLE 3. Statistical results of quantitative survival

Grading System Class/Score No. % MST (mos) 95% CI Comparison HR 95% CI p Value

Entire series RPA I 47 11 13.0 9.90–16.1

II 348 82 5.5 4.81–6.19 vs Class I 1.8 1.26–2.43 0.001III 27 6 2.6 1.89–3.25 vs Class II 1.8 1.20–2.65 0.004

SIR 8–10 47 11 12.7 6.96–18.44–7 319 76 5.8 5.14–6.46 vs Scores 8–10 1.6 1.16–2.23 0.0040–3 56 13 3.2 1.65–4.82 vs Scores 4–7 1.6 1.20–2.14 0.002

BSBM 3 7 2 8.8 0.0–6.432 86 20 11.0 7.22–14.8 vs Score 3 1.0 0.40–2.47 0.9941 282 67 5.4 4.70–6.16 vs Score 2 1.7 1.32–2.21 0.00010 47 11 2.6 1.76–3.38 vs Score 1 2.2 1.63–3.07 0.0001

DS-GPA 3.5–4 101 24 9.4 7.14–11.62.5–3 111 26 6.6 5.13–8.13 vs Scores 3.5–4 1.3 1.01–1.80 0.0481.5–2 143 34 4.2 3.20–5.26 vs Scores 2.5–3 1.4 1.09–1.84 0.010

0–1 67 16 2.9 1.86–3.89 vs Scores 1.5–2 1.4 1.01–1.84 0.041SRS-alone series RPA I 30 12 12.4 7.94–16.86

II 210 83 5.7 4.74–6.72 vs Class I 1.8 1.20–2.82 0.005III 12 5 2.1 0.33–.3.93 vs Class II 1.4 0.75–2.55 0.296

SIR 8–10 32 13 15.4 1.17–29.694–7 197 78 6.4 5.46–7.34 vs Scores 8–10 1.8 1.12–2.67 0.0050–3 23 9 3.1 1.14–5.21 vs Scores 4–7 1.7 1.07–2.61 0.024

BSBM 3 3 1 3.7 0–8.92 55 22 12.4 7.40–17.40 vs Score 3 0.2 0.07–0.76 0.0161 175 69 6.1 5.01–7.13 vs Score 2 1.8 1.29–2.52 0.0010 19 8 2.0 1.46–2.48 vs Score 1 3.2 1.96–5.20 0.0001

DS-GPA 3.5–4 68 27 9.7 5.61–13.852.5–3 69 27 6.6 2.85–9.41 vs Scores 3.5–4 1.5 1.01–2.09 0.0431.5–2 86 34 5.0 2.96–7.10 vs Scores 2.5–3 1.3 0.90–1.75 0.184

0–1 29 12 2.8 2.40–3.20 vs Scores 1.5–2 1.5 1.00–2.38 0.053


H. Kano et al.


was derived from patient populations treated with WBRT, craniotomy, SRS, or a combination of multiple treatment modalities. The DS-GPA contains the following 4 prog-nostic score groups: DS-GPA Scores 0–1, 1.5–2, 2.5–3, and 3.5–4. In this study, the distributions of patients in various DS-GPA groups were reasonably balanced within each prognostic score group (Scores 3.5–4, 24%; Scores 2.5–3, 26%; Scores 1.5–2, 34%; Scores 0–1, 16%) (Table 3). The DS-GPA was significantly correlated with patient survival between each DS-GPA score group. Overall, we found that the DS-GPA was the best grading system for the prediction of OS of patients with melanoma, all of whom received SRS as primary or adjuvant management.

Stereotactic Radiosurgery as a Primary Brain Metastasis Treatment

In many previous reports of brain metastasis grading systems, many patients underwent WBRT before SRS.10,14,24

Previous WBRT was a poor prognostic predictor for pa-tients with brain metastases.10 Adding WBRT to SRS did not seem to improve patient survival, but it might have im-proved both local tumor control and distant tumor control at 6 months.2 However, in a previous study,10 managing patients with WBRT followed by salvage or boost SRS was associated with decreased OS. This observation pro-vided additional evidence that SRS without WBRT is a reasonable initial treatment option for most patients with melanoma and brain metastases. In our current study, 31% of the patients underwent previous WBRT, 15% had un-dergone previous craniotomy, and 1% had undergone cyst aspiration as their brain metastasis treatment, which left the study with 252 patients who underwent SRS as their primary brain metastasis treatment. We evaluated 4 prog-nostic grading indices in patients with SRS as their pri-mary brain metastasis treatment (Table 3). In multivari-ate analysis, for both the entire group and the SRS-alone group, factors associated with shorter OS after SRS in-cluded lower KPS score, larger number of brain metas-tases, larger total tumor volume, and active extracranial metastases (Table 2). Of those who underwent SRS alone, the largest proportion (83%) of patients was at RPA Class II. RPA Class I was significantly associated with longer OS than RPA Class II (p = 0.005), whereas RPA Class II was not associated with OS when compared with RPA Class III (p = 0.296), which might be a result of the smaller number (n = 12) of patients at RPA Class III (Table 3). The SIR also suffered from unbalanced patient proportions within the prognostic classes (Scores 4–7, 78%). However, all 3 SIR score groups (Scores 8–10, 4–7, and 0–3) were highly correlated with patient survival (p = 0.005 [Scores 8–10 vs 4–7]; p = 0.024 [Scores 4–7 vs 0–3]) (Table 3). Al-though the BSBM also suffered from unbalanced patient proportions within the prognostic classes (BSBM1, 69%), all BSBM classes were significantly correlated with OS (p = 0.016 [BSBM3 vs BSBM2]; p = 0.001 [BSBM2 vs BSBM1]; p = 0.0001 [BSBM1 vs BSBM0]). In the SRS-alone group, the DS-GPA had a well-balanced proportion of patients within the prognostic score groups (Scores 3.5–4, 27%; Scores 2.5–3, 27%; Scores 1.5–2, 34%; Scores 0–1, 12%). DS-GPA Scores 3.5–4 were significantly as-sociated with longer OS than were DS-GPA Scores 2.5–3 (p = 0.043). Other DS-GPA score groups were not corre-lated with OS, which might be because of the small num-bers of patients in each DS-GPA score group. GPA Scores 3.5–4 were significantly associated with longer OS than were GPA Scores 1.5–2 (p < 0.0001) and 0–1 (p < 0.0001). We believe that the DS-GPA was the best grading system for the prediction of OS because of the better balance in patient numbers in each score group. Patients whose DS-GPA score was ≥ 2.5 had significantly improved survival after SRS. However, results of recent studies that used mo-lecularly targeted antitumor agents, such as vemurafenib, dabrafenib, and trametinib, suggested that these agents can result in remarkable improvement in selected patients with melanoma.11,22 The impact of these agents on existing intracranial disease is unknown at present, but a reduc-tion in the risk of new intracranial melanoma spread might have a major effect on survival. Therefore, our finding that the DS-GPA score currently best predicts the response of

TABLE 4. OS times according to each grading index

Grading System

Class/Score No. %

OS (%)6

Mos12

Mos 24

Mos36

Mos

Entire series RPA I 47 11 80.5 56.6 16.6 10.0

II 348 82 45.3 22.7 11.6 4.8III 27 6 22.2 11.1 7.4 0

SIR 8–10 47 11 67.6 50.0 21.6 9.64–7 319 76 48.2 25.0 11.5 5.00–3 56 13 28.6 9.5 5.7 0

BSBM 3 7 2 57.1 42.9 0 02 86 20 71.9 49.4 19.7 10.91 282 67 44.9 21.2 10.8 3.50 47 11 19.1 6.4 2.1 0

DS-GPA 3.5–4 101 24 65.8 42.3 23.0 10.52.5–3 111 26 53.5 27.8 14.1 9.51.5–2 143 34 40.3 21.6 5.3 2.7

0–1 67 16 26.6 6.3 4.7 0SRS-alone

series RPA I 30 12 79.4 51.8 24.2 12.1

II 210 83 49.4 25.1 11.7 5.4III 12 5 33.3 16.7 8.3 0

SIR 8–10 32 13 71.2 55.0 27.5 13.84–7 197 78 52.2 25.2 11.5 5.20–3 23 9 26.1 13.0 4.3 0

BSBM 3 3 1 33.3 0 0 02 55 22 74.2 50.1 25.1 14.71 175 69 50.2 23.7 10.7 4.30 19 8 10.5 5.3 0 0

DS-GPA 3.5–4 68 27 65.4 43.9 24.6 13.82.5–3 69 27 54.2 26.6 13.1 4.51.5–2 86 34 47.3 23.6 7.1 2.8

0–1 29 12 31.0 6.9 3.4 0




brain SRS in patients with melanoma might evolve in the future.

Study LimitationsWe acknowledge that this retrospective study had in-

herent limitations and that the results might relate in part to selection bias. In an attempt to offset the variations in referral patterns and significant treatment paradigms, such as the use of SRS for increasing numbers of metastases, these results were analyzed and a 20-year time period was used. We found that patient survival was not significantly associated with the year in which the patients underwent their SRS. In the future, molecularly targeted antitumor

agents might further improve survival in patients with melanoma and intracranial spread. Over the next 10 years, we anticipate performing a follow-up analysis of the out-comes of melanoma after it spreads to the brain in patients who underwent molecular therapy as a part of the standard of care.

ConclusionsThe 4 major grading systems studied here yielded sta-

tistically significant differences in patients with melanoma brain metastases. However, in this series of patients with melanoma selected for SRS, the RPA, SIR, and BSBM had

FIG. 2. Kaplan-Meier curves for OS of the patients for each index according to individual class in patients who underwent SRS as primary treatment for their melanoma brain metastases (n = 252). MST values are in months.


H. Kano et al.


significant variations in the numbers of patients distrib-uted into graded subgroups. Among these 4 major grad-ing systems, the DS-GPA index proved most balanced and predictive of survival. Patients who had a DS-GPA score of ≥ 2.5 had significantly improved survival after SRS.

AcknowledgmentsWe thank Professor Douglas Kondziolka, MD (NYU Langone

Medical Center), for his significant contribution to patient manage-ment at the University of Pittsburgh.

References 1. Bafaloukos D, Gogas H: The treatment of brain metastases in

melanoma patients. Cancer Treat Rev 30:515–520, 2004 2. Brown PD, Brown CA, Pollock BE, Gorman DA, Foote RL:

Stereotactic radiosurgery for patients with “radioresistant” brain metastases. Neurosurgery 62 (Suppl 2):790–801, 2008

3. Christopoulou A, Retsas S, Kingsley D, Paddick I, Lindquist C: Integration of Gamma Knife surgery in the management of cerebral metastases from melanoma. Melanoma Res 16:51–57, 2006

4. Doss LL, Memula N: The radioresponsiveness of melanoma. Int J Radiat Oncol Biol Phys 8:1131–1134, 1982

5. Gaspar L, Scott C, Rotman M, Asbell S, Phillips T, Wasser-man T, et al: Recursive partitioning analysis (RPA) of prog-nostic factors in three Radiation Therapy Oncology Group (RTOG) brain metastases trials. Int J Radiat Oncol Biol Phys 37:745–751, 1997

6. Geara FB, Ang KK: Radiation therapy for malignant mela-noma. Surg Clin North Am 76:1383–1398, 1996

7. Hara W, Tran P, Li G, Su Z, Puataweepong P, Adler JR Jr, et al: Cyberknife for brain metastases of malignant melanoma and renal cell carcinoma. Neurosurgery 64 (2 Suppl):A26–A32, 2009

8. Kondziolka D, Kano H, Harrison GL, Yang HC, Liew DN, Niranjan A, et al: Stereotactic radiosurgery as primary and salvage treatment for brain metastases from breast cancer. Clinical article. J Neurosurg 114:792–800, 2011

9. Kondziolka D, Parry PV, Lunsford LD, Kano H, Flickinger JC, Rakfal S, et al: The accuracy of predicting survival in in-dividual patients with cancer. J Neurosurg 120:24–30, 2014

10. Liew DN, Kano H, Kondziolka D, Mathieu D, Niranjan A, Flickinger JC, et al: Outcome predictors of Gamma Knife surgery for melanoma brain metastases. Clinical article. J Neurosurg 114:769–779, 2011

11. Long GV, Trefzer U, Davies MA, Kefford RF, Ascierto PA, Chapman PB, et al: Dabrafenib in patients with Val600Glu or Val600Lys BRAF-mutant melanoma metastatic to the brain (BREAK-MB): a multicentre, open-label, phase 2 trial. Lan-cet Oncol 13:1087–1095, 2012

12. Lorenzoni J, Devriendt D, Massager N, David P, Ruíz S, Vanderlinden B, et al: Radiosurgery for treatment of brain metastases: estimation of patient eligibility using three strati-fication systems. Int J Radiat Oncol Biol Phys 60:218–224, 2004

13. Mathieu D, Kondziolka D, Cooper PB, Flickinger JC, Niran-jan A, Agarwala S, et al: Gamma Knife radiosurgery in the management of malignant melanoma brain metastases. Neu-rosurgery 60:471–482, 2007

14. Serizawa T, Higuchi Y, Nagano O, Hirai T, Ono J, Saeki N, et al: Testing different brain metastasis grading systems in ste-reotactic radiosurgery: Radiation Therapy Oncology Group’s RPA, SIR, BSBM, GPA, and modified RPA. J Neurosurg 117 Suppl:31–37, 2012

15. Serizawa T, Higuchi Y, Nagano O, Matsuda S, Ono J, Saeki N, et al: A new grading system focusing on neurological outcomes for brain metastases treated with stereotactic ra-diosurgery: the modified Basic Score for Brain Metastases. J Neurosurg 121 Suppl:35–43, 2014

16. Sloan AE, Nock CJ, Einstein DB: Diagnosis and treatment of melanoma brain metastasis: a literature review. Cancer Contr 16:248–255, 2009

17. Sperduto PW, Berkey B, Gaspar LE, Mehta M, Curran W: A new prognostic index and comparison to three other indices for patients with brain metastases: an analysis of 1,960 pa-tients in the RTOG database. Int J Radiat Oncol Biol Phys 70:510–514, 2008

18. Sperduto PW, Kased N, Roberge D, Xu Z, Shanley R, Luo X, et al: Summary report on the graded prognostic assessment: an accurate and facile diagnosis-specific tool to estimate survival for patients with brain metastases. J Clin Oncol 30:419–425, 2012

19. Tarhini AA, Agarwala SS: Management of brain metastases in patients with melanoma. Curr Opin Oncol 16:161–166, 2004

20. Tsao H, Atkins MB, Sober AJ: Management of cutaneous melanoma. N Engl J Med 351:998–1012, 2004

21. Weltman E, Salvajoli JV, Brandt RA, de Morais Hanriot R, Prisco FE, Cruz JC, et al: Radiosurgery for brain metastases: a score index for predicting prognosis. Int J Radiat Oncol Biol Phys 46:1155–1161, 2000

22. Wolf A, Zia S, Verma R, Pavlick A, Wilson M, Golfinos JG, et al: Impact on overall survival of the combination of BRAF inhibitors and stereotactic radiosurgery in patients with mela-noma brain metastases. J Neurooncol 127:607–615, 2016

23. Yamamoto M, Kawabe T, Higuchi Y, Sato Y, Barfod BE, Kasuya H, et al: Validity of three recently proposed prognos-tic grading indexes for breast cancer patients with radiosurgi-cally treated brain metastases. Int J Radiat Oncol Biol Phys 84:1110–1115, 2012

24. Yamamoto M, Sato Y, Serizawa T, Kawabe T, Higuchi Y, Nagano O, et al: Subclassification of recursive partitioning analysis Class II patients with brain metastases treated ra-diosurgically. Int J Radiat Oncol Biol Phys 83:1399–1405, 2012

DisclosuresDr. Lunsford is a consultant for and stockholder in Elekta AB and a consultant for the Focused Ultrasound Foundation; and Dr. Kirkwood is a consultant for BMS, Roche, Amgen, Green Peptide, and Genentech. This study is supported by an Elekta research grant.

Author ContributionsConception and design: Kano. Acquisition of data: Kano, Morales-Restrepo, Iyer, Weiner, Mousavi. Drafting the article: Kano. Critically revising the article: Kano, Kirkwood, Tarhini, Flickinger, Lunsford. Reviewed submitted version of manu-script: all authors. Approved the final version of the manuscript on behalf of all authors: Kano. Statistical analysis: Kano. Study supervision: Kano.

CorrespondenceHideyuki Kano, Department of Neurological Surgery, University of Pittsburgh, Ste. B-400, UPMC Presbyterian, 200 Lothrop St., Pittsburgh, PA 15213. email: [email protected].


comparison of prognostic indices in patients who undergo

Documents