- Email: [email protected]
New Breast Cancer Recursive Partitioning Analysis Prognostic Index in Patients With Newly Diagnosed Brain Metastases
Int. J. Radiation Oncology Biol. Phys., Vol. 82, No. 5, pp. 2065–2071, 2012 Copyright Ó 2012 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/$ - see front matter
doi:10.1016/j.ijrobp.2010.10.077
CLINICAL INVESTIGATION
Breast Cancer
NEW BREAST CANCER RECURSIVE PARTITIONING ANALYSIS PROGNOSTIC INDEX IN PATIENTS WITH NEWLY DIAGNOSED BRAIN METASTASES y NSKA, M.D., PH.D.,* AND MAGDALENA MURAWSKA, M.SC. ANNA NIWI
*Department of Breast Cancer and Reconstructive Surgery, The Maria Sk1odowska-Curie Memorial Cancer Center and Institute of Oncology, Warsaw, Poland; and yDepartment of Biostatistics, Erasmus University Medical Center, Rotterdam, The Netherlands Purpose: The aim of the study was to present a new breast cancer recursive partitioning analysis (RPA) prognostic index for patients with newly diagnosed brain metastases as a guide in clinical decision making. Methods and Materials: A prospectively collected group of 441 consecutive patients with breast cancer and brain metastases treated between the years 2003 and 2009 was assessed. Prognostic factors significant for univariate analysis were included into RPA. Results: Three prognostic classes of a new breast cancer RPA prognostic index were selected. The median survival of patients within prognostic Classes I, II, and III was 29, 9, and 2.4 months, respectively (p < 0.0001). Class I included patients with one or two brain metastases, without extracranial disease or with controlled extracranial disease, and with Karnofsky performance status (KPS) of 100. Class III included patients with multiple brain metastases with KPS of #60. Class II included all other cases. Conclusions: The breast cancer RPA prognostic index is an easy and valuable tool for use in clinical practice. It can select patients who require aggressive treatment and those in whom whole-brain radiotherapy or symptomatic therapy is the most reasonable option. An individual approach is required for patients from prognostic Class II. Ó 2012 Elsevier Inc. Brain metastases, Breast cancer, Karnofsky performance status, Prognostic index, Recursive partitioning analysis.
Recently, several prognostic indexes had been proposed in the literature (1–9). They include (1) the Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis (RPA) (1, 2) with three classes: Class I (patients <65 years, Karnofsky performance status [KPS] $70, controlled primary tumor, and no extracranial metastases), Class III (KPS <70), and Class II (all patients not in Class I or III); (2) the Score Index for Radiosurgery (3), which is the sum of scores (0–2) for each of five prognostic factors (age, KPS, status of systemic disease, number of lesions in the brain, and largest volume of the lesion); (3) the Basic Score for Brain Metastases (4), which is the sum of scores (0–1) for three prognostic factors (KPS, control of primary tumor, and extracranial metastases) and (4) the Graded Prognostic Assessment (GPA) (5) which is the sum of scores (0, 0.5, and 1) for four factors: age, KPS, extracranial metastases, and number of metastases in the brain (one, two to three, and more). The GPA has been updated and now is the sum of four factors: KPS, HER2 receptor, ER/PR receptors, and number of brain metastases (6, 9). Recently, six diagnosis-specific GPA prognostic indexes for different primary tumor types were proposed (7).
INTRODUCTION Brain metastasis in breast cancer patients is a growing problem because of the high incidence of central nervous system involvement and a lack of clear therapeutic procedures. Not all patients with brain metastases have the same prognosis, and the use of the same treatment methods for all patients is no longer appropriate. Nowadays, the treatment of brain metastasis from breast cancer includes neurosurgery, whole-brain radiation therapy (WBRT), stereotactic radiotherapy, and systemic therapy (chemotherapy, endocrine therapy, and targeted therapy). The choice and sequence of the appropriate method of treatment for a given patient should depend on many factors that are related to patient characteristics (performance status, age) and cancer characteristics such as biologic subtype of breast cancer, number of brain metastases, extracranial metastases, and location of extracranial metastases. Because patients with newly diagnosed brain metastases from breast cancer are a heterogeneous group, there is a need for introducing a simple breast cancer–specific prognostic index that can help clinicians choose a proper treatment.
Conflict of interest: none. Received July 12, 2010, and in revised form Oct 14, 2010. Accepted for publication Oct 21, 2010.
Reprint requests to: Anna Niwinska, M.D., Ph.D., Department of Breast Cancer and Reconstructive Surgery, The Maria Sk1odowskaCurie Memorial Cancer Center and Institute of Oncology, ul. Roentgena 5, 02-781 Warsaw, Poland. Tel: (+ 48) 22-644-00-24; Fax: (+48) 22-644-00-24; E-mail: [email protected] 2065
2066
I. J. Radiation Oncology d Biology d Physics
Table1. Characteristics of 441 patients with brain metastases Characteristic
No. of patients
Volume 82, Number 5, 2012
Table1. Characteristics of 441 patients with brain metastases (Continued ) % Characteristic
Age at initial diagnosis (y) Median Range Initial TNM stage I II III IV Histologic type Ductal carcinoma Lobular carcinoma Medullar, apocrinal, papillar, mucinous, planoepithelial, neuroendocrine carcinomas Cancer cells or invasive cancer after chemotherapy Estrogen/progesterone receptor status Positive Negative Missing HER2 status Positive Negative Missing Biologic subtype* Triple-negative Luminal A Luminal B HER2 Missing Localization of metastases Brain as the first or only site Liver Lung Bone Locoregional recurrence Number of brain metastases 1 2 3 Multiple RPA RTOG prognostic class I II III Type of treatment of brain metastases Neurosurgery 1 metastasis operated 2 metastases operated Radiotherapy 40 Gy/20 fractions 30 Gy/10 fractions or 20 Gy/5 fractions No radiotherapy Systemic therapy after brain metastases Yes No Type of systemic therapyy Hormonal therapy Chemotherapy Targeted therapyz Type of chemotherapyy Vinorelbine Capecitabine
50 21–76 44 181 148 68
10 41 34 15
352 30 21
80 7 5
38
8
195 238 8
44 54 2
232 196 13
53 44 3
111 81 101 127 21
25 18 23 29 5
90 146 224 198 137
20 33 51 45 31
109 31 10 291
25 7 2 66
43 253 145
10 57 33
77/441 68 9 432/441 46 386
17 88 12 98 10 88
9
2
307 134
70 30
73/307 255/307 113/307
24 83 37
89/255 35 63/255 25 (Continued )
Anthracycline Platinum Taxane Fluorouracil Cyclophosphamide Etoposide Other Type of endocrine therapyy Aromatase inhibitors Tamoxifen Goserelin Megestrol acetate Fulvestrant Type of targeted therapyy Trastuzumab Lapatinib
No. of patients
%
42/255 35/255 27/255 20/255 18/255 15/255 4/255
16 14 11 8 7 6 2
38/73 19/73 8/73 8/73 7/73
52 26 11 11 10
102/113 11/113
90 10
Abbreviations: RPA = recursive partitioning analysis; RTOG = Radiation Therapy Oncology Group. * Triple-negative, ER/PR negative HER2-negative; luminal A, ER/PR-positive HER2-negative; luminal B, ER/PR positive HER2-positive; HER2, ER/PR-negative HER2-positive. y In some patients many types of systemic treatment were ordered. z 37% of all patients and 50% of patients with HER2-positive breast cancer.
The aim of the present study was to analyze the outcome in 441 consecutive patients with breast cancer and brain metastases based on factors related to patient, cancer, and treatment. The main goal of the study was to develop a new breast cancer RPA prognostic index for patients with newly diagnosed brain metastases as a guide in clinical decision making. It was created and based only on factors related to patient characteristics and cancer. METHODS AND MATERIALS Data from 441 prospectively collected consecutive patients with breast cancer and brain metastases, treated in the Breast Cancer Department at the Cancer Center and Institute of Oncology in Warsaw, Poland, from January 1, 2003, to December 31, 2009, formed the basis for this study. Brain metastases were detected by magnetic resonance imaging (97%) or computed tomography (3%). All patients were prospectively evaluated after the detection of brain metastases. In each case, treatment options were approved by a team of medical oncologists, radiation oncologist, and neurosurgeon and were performed after the patients signed written consent. WBRT was performed in 98% of patients using a 4- to 6-MV photon beam by two lateral opposed standard fields covering all intracranial contents. The most common regimen of WBRT was 30Gy in 10 fractions. Nine patients (2%) were not irradiated because of poor performance status at the diagnosis of brain metastases. WBRT was performed in most cases by the same radiation oncologist (A.N.). Seventy-seven patients (17%) with one or two brain metastases were operated on before receiving WBRT. Systemic treatment after WBRT was ordered in 307 (70%) patients. Most patients (83%) received chemotherapy. Schedules with vinorelbine and capecitabine were the most frequent types of chemotherapy
New index in breast cancer brain metastases d A. NIWINSKA AND M. MURAWSKA
Table 2. Factors influencing survival from brain metastases: univariate analysis Covariate Patient-related factors KPS Age (y) Cancer-related factors Biologic subtype* Extracranial disease Extracranial disease Locoregional recurrence Lung metastases Liver metastases Bone metastases Number of brain metastases Treatment-related factors Neurosurgery WBRT Systemic treatment
Comparison
p value
$70 vs. <70 <40 vs. 40–65 40–65 vs. >65 #65 vs. >65
<0.0001 NS 0.009 0.002
Triple-negative vs. HER2 Triple-negative vs. luminal B Triple-negative vs. luminal A Present vs. absent Absent or controlled vs. uncontrolled Present vs. absent
<0.001 <0.0001 <0.0001 0.016 <0.0001 0.005
Present vs. absent Present vs. absent Present vs. absent 1 vs. 2
NS 0.002 NS NS
1 vs. 3 1 vs. multiple
NS 0.0001
Yes vs. no 20 Gy vs. 30 Gy 20 Gy vs. 40 Gy Yes vs. no
0.0001 0.0001 0.0001 0.0001
Abbreviations: KPS = Karnofsky performance status; NS = not significant; WBRT = whole-brain radiotherapy. * Triple-negative, ER/PR negative HER2-negative; luminal A, ER/PR-positive HER2-negative; luminal B, ER/PR positive HER2positive; HER2, ER/PR-negative HER2-positive.
used. Hormonal therapy was used in 24% of patients. Aromatase inhibitors were the most frequently ordered. Targeted therapy (trastuzumab or lapatinib) was ordered in 113 patients, which included 50% of all HER2-positive population in the study. Out of 441 patients, 420 patients were divided into four biologic subtypes based on the expression of ER, PgR, and HER2 receptors through immunohistochemistry and fluorescence in situ hybridization. The patients were divided into four biologic subtypes as follows: (1) triple-negative (ER-negative, PgR-negative, HER2-negative), (2) HER2 (HER2-positive, ER-negative, PgR-negative), (3) luminal B (HER2-positive, ER-positive and/or PgR-positive), and (4) luminal A (ER-positive and/or PgR-positive HER2-negative). Triple-negative and luminal A subsets were HER2-negative. The HER2 and luminal B subsets were HER2-positive. Clinical characteristics and type of systemic treatment after WBRT are presented in Table 1.
Statistical analysis To determine the factors influencing survival from brain metastases, univariate analysis was performed. Patient-related covariates taken into the analysis were age (<40 vs. 40–65 and vs. >65 years) and Karnofsky performance status ($70 vs. <70). Cancer-related factors were biologic subtype of breast cancer (triple-negative vs. luminal A, triple-negative vs. HER2, triple-negative vs. luminal B), locoregional recurrence (present vs. absent), extracranial metastases (present vs. absent), extracranial disease (absent/controlled
2067
vs. uncontrolled), lung metastases (present vs. absent), liver metastases (present vs. absent), bone metastases (present vs. absent), and number of brain metastases (1 vs. 2–3 and 1 vs. multiple). Treatment-related factors included neurosurgery (yes vs. no), WBRT (20 Gy vs. 30 Gy and 20 Gy vs. 40 Gy), and systemic therapy (yes vs. no). Factors that were significant in univariate analysis were included in the recursive partitioning tree. Recursive partitioning analysis, a method of building a decision tree to model predictors as described by Gordon and Olshen (10), was applied to establish prognostic groups. Each variable was examined for the best split within that variable, and all variables were examined for the best split for a given population. A variable was defined as either nominal or ordered. The first node of RPA included all patients because all variables were examined in this node. The Kaplan-Meier method was used by RPA to estimate time to event. The node was split if the log-rank statistic was significant for any variable beyond the 0.05 probability level. The significance level was adjusted for the number of multiple comparisons. Two RPA trees were built: The first included covariates related to patient, cancer, and treatment; the second assessed only factors related to patient and cancer. The first RPA analysis was performed to distinguish several groups of breast cancer patients with brain metastases with different survivals depending on type of treatment, and the second was performed to develop a new RPA breast cancer prognostic index based only on factors related to patient and cancer. In first tree, a total of 11 and in the second a total of eight variables were tested in the first run (first node) in the recursive partitioning algorithm. Age, KPS, and number of brain metastases were analyzed as continuous variables. Pairwise comparisons were performed to analyze statistically significant differences between prognostic Classes I, II, and III (log-rank, Mantel-Cox test). Additionally, survival from brain metastases based on the RPA RTOG prognostic index proposed by Gaspar et al. (1) and survival depending on KPS as the only variable was presented.
RESULTS The median survival from brain metastases in the entire group was 7 months (95% confidence interval 6.036–8.28). The results of univariate analysis are presented in Table 2. Eight statistically significant factors were related to patient and cancer and three others were statistically significantly related to treatment. Table 3 presents 10 terminal nodes of the recursive partitioning tree using factors related to patient, cancer, and treatment, with median survival ranging from 2 to 29 months. Five statistically significant variables were selected: KPS, biologic subtype of breast cancer, liver metastases, neurosurgery, and systemic treatment after WBRT. A recursive partitioning tree of the new breast cancer RPA prognostic index is presented in Fig. 1. Six terminal nodes of the recursive partitioning tree had median survival ranging from 2.4 to 29 months. Three prognostic classes of the new breast cancer RPA prognostic index were selected. The median survival of patients within prognostic Classes I, II, and III was 29 months, 9 months, and 2.4 months, respectively (p < 0.0001). Class I included patients with one or two brain metastases, without extracranial disease or with controlled extracranial disease, and with KPS of 100. Class III included patients with multiple brain metastases with KPS of #60. Class II included all other cases. Pairwise
2068
I. J. Radiation Oncology d Biology d Physics
Volume 82, Number 5, 2012
Table 3. Median survival of patients by factors related to patients, cancer, and treatment: 10 nodes of RPA tree (426 patients)
Node 1 2 3 4 5 6 7 8 9 10
Variables With systemic treatment after WBRT, operated, with controlled extracranial disease With systemic treatment after WBRT, operated, with uncontrolled extracranial disease With systemic treatment after WBRT, without operation, HER2-positive, KPS 100, without liver mets With systemic treatment after WBRT, without operation, HER2-positive, KPS<=90, without liver mets With systemic treatment after WBRT, without operation, HER2-positive, with liver mets With systemic treatment after WBRT, without operation, TNBC or Luminal A, KPS >60 With systemic treatment after WBRT, without operation, TNBC or luminal A, KPS #60 Without systemic treatment after WBRT, KPS >80 Without systemic treatment after WBRT, KPS 60–80 Without systemic treatment after WBRT, KPS #50
No. of patients
Median survival (mo)
25
29
25
20
30
22
51
9
67
9
67
7
29
3
23
7
55
3
54
2
Abbreviations: RPA = recursive partitioning analysis; KPS = Karnofsky performance status; mets = metastases; WBRT = whole-brain radiation therapy; HER2-positive = all patients with HER2 receptor positive (HER2 and luminal B); TNBC = triplenegative breast cancer (ER-negative, PR-negative, HER2-negative); luminal A (ER/PR-positive, HER2-negative).
comparisons revealed statistically significant differences in survival between the three prognostic classes (I vs. II, p < 0.0001; I vs. III, p < 0.0001; II vs. I, p < 0.0001; II vs.III, p < 0.0001; III vs. I, p < 0.0001; III vs II, p < 0.0001). The median survivals of three prognostic classes are presented in Table 4 and Fig. 2. According to the RPA RTOG proposed by Gaspar et al. (1), the survival of patients in prognostic Classes I, II, and III was 16 months, 10 months, and 2.7 months, respectively (p < 0.0001). The results are presented in Table 5. Table 6 presents survivals from brain metastases depending on KPS status. A direct proportional relationship between KPS and survival was revealed. DISCUSSION Why a new prognostic index? The stratification of patients with brain metastases into prognostic classes could be useful for everyday clinical practice because it helps in the choice of proper treatment. Several prognostic indexes that have been proposed since 1997 are related mainly to patients with lung cancer. The RPA
Fig. 1. Recursive tree from a database of 441 patients with breast cancer brain metastases after the inclusion of eight prognostic factors related to patient and cancer. KPS = Karnofsky performance status.
RTOG prognostic index proposed by Gaspar et al. (1) was based on 1,200 patients from three RTOG trials, but only 137 (12%) of them were patients with breast cancer. A comparison of four prognostic indices, performed by Sperduto et al. (5) on 1,960 patients from five RTOG clinical trials (11–15), included only 222 (11%) patients with breast cancer; however, a recent study by Sperduto et al. concerned 642 patients with breast cancer brain metastases (7). In the present study we assessed the usefulness of new prognostic indices in a group of 441 consecutive breast cancer patients. The rationale for introducing a breast cancer–specific prognostic index is that breast cancer differs from other primaries in terms of biology, survival, and response to systemic treatment. To build a new breast cancer prognostic index, we analyzed eight factors related to patient and disease, including a new one: number of brain lesions and biologic subtype of breast cancer. In our breast cancer RPA prognostic index, 6% of patients were in prognostic Class I, with median survival of 29 months. In such patients, combined modality treatment with neurosurgery/WBRT or radiosurgery/WBRT and systemic therapy should be a reasonable treatment option. By contrast, one fourth of our patients were in prognostic Class III and had a poor prognosis, with median survival of 2.4 months. Based on our previous experience (16), most such patients do not benefit from WBRT, and symptomatic therapy (corticosteroids) is equivalent to WBRT. These patients could be spared the time, costs, and side effects of aggressive treatment. Patients belonging to these two extreme prognostic classes can be easily selected and properly treated. The prognostic Class II includes more than half of breast cancer patients with great heterogeneity, and those patients require an individual therapeutic approach. Breast cancer–specific prognostic factors In our new breast cancer–specific RPA prognostic index, KPS, number of brain metastases, and status of extracranial
New index in breast cancer brain metastases d A. NIWINSKA AND M. MURAWSKA
2069
Table 4. New breast cancer recursive partitioning analysis prognostic index after inclusion of eight factors related to patient and cancer Class I II III
Characteristic
No. of patients (%)
1–2 brain metastases and extracranial disease absent or controlled and KPS 100 All others Multiple brain metastases and KPS #60
25 (6%)
Median survival (mo) 29
296 (67%) 120 (27%)
95% CI
Significance
(11.544; 45.912)
9 2.4
(8.028; 10.752) (2.052; 2.676)
p < 0.0001
Abbreviations: CI = confidence index; KPS = Karnofsky performance status.
disease were the most important prognosticators. By contrast to the RPA RTOG prognostic index (1) and other indexes (5, 6), in the present study, age was not a statistically significant factor influencing survival of patients with brain metastases from breast cancer.
Performance status Our data along with published data revealed that KPS is a key determinant of prognosis in patients with breast cancer and brain metastases (1, 2, 6, 7, 16, 17). Sperduto et al. (7) demonstrated the importance of KPS in patients with lung, breast, renal, and gastrointestinal cancers as well as in cases of melanoma. In the present study we confirmed that KPS is the most important prognostic factor in breast cancer patients with brain metastases (Tables 2–5). The survival of patients with Class III of the new breast cancer RPA prognostic index (2.4 months), in which KPS was a significant variable, was as poor as the survival of patients in Class III of the RPA RTOG prognostic index by Gaspar et al. (2.3 months) (1) and in Class III lung cancer patients from the Polish study by Ke˛pka et al. (2.5 months) (18). In the present study, patients with single brain metastasis and poor performance status had an unsatisfactory prognosis. It was comparable, or even worse, to the prognosis of
patients with multiple brain metastases and good performance status. Number of brain metastases The important limitation of some previously proposed prognostic indexes was the avoidance of the number of brain metastases (1, 2). The number of brain lesions has been shown to be an important factor influencing survival in a randomized trial (11). The authors found a survival advantage for patients with one metastasis, but not for those with two or three metastases. In our RPA tree, the number of brain metastases was treated as a continuous variable. The results suggest that patients with one or two brain lesions with KPS of 100 and without extracranial disease or with controlled extracranial disease have the best prognosis. These patients should be given very aggressive treatment. Intensive local treatment can control the disease in the brain for a long time, and systemic treatment ordered after local treatment can control occult extracranial metastases. This treatment policy can improve survival significantly. Controlled/uncontrolled extracranial disease In the RPA RTOG prognostic index proposed by Gaspar et al. (1), the status of the primary tumor and extracranial metastases was one of the most important prognostic factors influencing the survival of patients with brain metastases. However, in some publications that appeared later, the factor called uncontrolled disease has been assessed as subjective and not convincing (5). In the present study, restaging of every patient was performed after the detection of brain metastases. The status of the primary tumor (local failure present/absent and controlled/uncontrolled) and the status of the extracranial metastases (present/absent and controlled/uncontrolled) was assessed in every case. For the present study, patients with controlled extracranial disease were those with a complete Table 5. Median survival of 441 patients in RPA RTOG prognostic classes by Gaspar (1) RPA RTOG prognostic class I II III
Fig. 2. Breast cancer recursive partitioning analysis prognostic index: survival from brain metastases (meta) within three prognostic cases.
No. of patients (%)
Median survival (mo)
Significance
43 (10) 253 (57) 145 (33)
16 10 2.7
p < 0.0001
Abbreviations: RPA = recursive partitioning analysis; RTOG = Radiation Therapy Oncology Group.
I. J. Radiation Oncology d Biology d Physics
2070
Table 6. Median survival of 441 patients depending on Karnofsky performance status (KPS) KPS
No. of patients
Median survival (mo)
Significance
100 90 70–80 #60
120 55 121 145
16 12 7 2.7
p < 0.0001
or partial response to the treatment of locoregional relapse or distant metastases. They did not require active treatment of extracranial disease at the time of detection of brain lesions. They usually had single/several bone or lung metastases or local recurrence, which had been successfully treated before the detection of brain metastases. In fact, the term ‘‘controlled extracranial disease’’ means that the probability of survival is at least 6 months, with a low risk of death because of locoregional failure or extracranial metastases. In the present study, the variable ‘‘controlled/uncontrolled extracranial disease’’ was more valuable than ‘‘extracranial metastases (yes vs. no)’’ or ‘‘locoregional failure (yes vs. no)’’; the median survival of patients with single brain metastases and controlled extracranial disease was comparable to that of patients with solitary brain metastasis (one brain lesion without extracranial disease). Biologic subtype The biologic subsets of breast cancer differ according to natural prognosis and response to systemic treatment (17, 19). Triple-negative and HER2-positive breast cancer have a special predisposition to metastasize to the brain, and the natural prognosis of these subtypes is poor. However, after systemic treatment that follows whole-brain radiotherapy, the survival of these two types has changed significantly. In the posttrastuzumab era, the survival of HER2-positive breast cancer patients is the longest of all biologic subtypes (17). It exceeds historical estimates and argues against a nihilistic approach in this subgroup of patients. By contrast, the triple-negative subtype has the poorest prognosis because of the natural rapid progress of the disease and the lack of effective drugs against systemic and central nervous system diseases. In the present study, the biologic subtype of breast cancer was significant only when 11 variables, together with treatment-related ones, were analyzed.
Volume 82, Number 5, 2012
Among all biologic subsets, the most favorable survival was assessed in patients with HER2-positive subtype, without liver metastases. In our new breast cancer RPA prognostic index, biologic subtype was not a significant variable.
Strength and limitations of the study Strength. In most clinical trials, only patients with good performance status are analyzed. By contrast, our group consisted of nonselected patients, together with those with poor performance status. This caused the results of the present study to be comparable with the results observed in everyday clinical practice. The differences in survival between prognostic Classes I, II, and III are significant and distinguishable, and this enables us to make quick decisions regarding treatment policy. The results confirm the value of the number of brain metastases as a prognostic factor that helps in treatment decision making. We also analyzed, for the first time to our knowledge, the value of biologic subtype of breast cancer. Limitations. It is very challenging to develop a new prognostic index based on a group of patients who have already been treated, because the outcome in those patients depends not only on factors related to patient and cancer, but also on treatment interventions. From an ethical point of view, it is not feasible to draw up a prognostic index on patients who have not been treated. Although we included only factors related to patients and cancer in the new index, we obtained survivals of previously treated patients, and this can be an important methodologic bias influencing our results.
CONCLUSIONS The new breast cancer RPA prognostic index can act as a guide for clinical decision making for the heterogeneous population of patients with brain metastases. It can distinguish between the group of patients with good prognosis (median survival 29 months) requiring aggressive combined treatment policy and the group of patients with very poor prognosis (median survival 2.4 months) for whom WBRT is questionable and symptomatic treatment is a reasonable option.
REFERENCES 1. Gaspar L, Scott CH, 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. 2. Gaspar L, Scott C, Murray K, et al. Validation of the RTOG recursive partitioning analysis (RPA) classification for brain metastases. Int J Radiat Oncol Biol Phys 2000;47: 1001–1006. 3. Weltman E, Salvajoli JV, Brandt RA, et al. Radiosurgery for brain metastases: A score index for predicting prognosis. Int J Radiat Oncol Biol Phys 2000;46:1155–1161.
4. Lorenzoni J, Devriendt D, Massager N, et al. Radiosurgery for treatment of brain metastases: Estimation of patient eligibility using three stratification systems. Int J Radiat Oncol Biol Phys 2004;60:218–224. 5. Sperduto PW, Berkey B, Gaspar LE, et al. A new prognostic index and comparison to three other indices for patients with brain metastases: An analysis of 1960 patients in the RTOG database. Int J Radiat Oncol Biol Phys 2008;70:510–514. 6. Sperduto PW, Xu P, Sneed P, et al. A graded prognostic assessment (GPA) for women with breast cancer and brain metastases [Abstract]. ASCO Proceedings J Clin Oncol 2010;28:12.
New index in breast cancer brain metastases d A. NIWINSKA AND M. MURAWSKA
7. Sperduto PW, Chao ST, Sneed PK, et al. Diagnosis-specific prognostic factors, indexes, and treatment outcomes for patients with newly diagnosed brain metastases: A multiinstitutional analysis of 4,259 patients. Int J Radiat Oncol Biol Phys 2010;77:655–661. 8. Le Scodan R, Massard Ch, Mouret-Fourme E, et al. Brain metastases from breast carcinoma: Validation of the Radiation Therapy Oncology Group recursive partitioning analysis classification and proposition of a new prognostic score. Int J Radiat Oncol Biol Phys 2007; 69:839–845. 9. Sperduto CM, Watanabe Y, Mullan J, et al. A validation study of a new prognostic index for patients with brain metastases: The Graded Prognostic Assessment. J Neurosurg 2008;109: 87–89. 10. Gordon L, Olshen RA. Tree-structured survival analysis. Cancer Treat Rep 1985;69:1065–1069. 11. Andrews DW, Scott CB, Sperduto PW, et al. Whole brain radiation therapy with or without stereotactic radiosurgery boost for patients with one to three brain metastases: Phase III results of the RTOG 9508 randomized trial. Lancet 2004;363: 1665–1672. 12. Komarnicky LT, Philips TL, Martz K, et al. A randomized phase III protocol for the evaluation of misonidasole combined with radiation in the treatment of patients with brain metastases (RTOG 79-16). Int J Radiat Oncol Biol Phys 1991;20:53–58.
2071
13. Sause WT, Scott C, Kirsch R, et al. Phase I/II trial of accelerated fractionation in brain metastases, RTOG 85-28. Int J Radiat Oncol Biol Phys 1993;26:653–657. 14. Phillips TL, Scott CB, Leibel S, et al. Results of a randomized comparison of radiotherapy and bromodeoxyuridine to radiotherapy alone for brain metastases: Report of RTOG trial 89-05. Int J Radiat Oncol Biol Phys 1995;33:339–348. 15. Murray KJ, Scott C, Greenberg HM, et al. A randomized phase III study of accelerated hyperfractionation versus standard in patients with unresected brain metastasis: A report of RTOG 9104. Int J Radiat Oncol Biol Phys 1997;39:571–574. 16. Komosi nska K, Ke˛pka L, Niwi nska A, et al. Prospective evaluation of the palliative effect of whole-brain radiotherapy in patients with brain metastases and poor performance status. Acta Oncol 2010;49:382–388. 17. Niwi nska A, Murawska M, Pogoda K. Breast cancer subtypes and response to systemic treatment after whole-brain radiotherapy in patients with brain metastases [Abstract]. ASCO Proceedings J Clin Oncol 2010;28:121. 18. Ke˛pka L, Cieslak E, Bujko K, et al. Results of the whole-brain radiotherapy for patients with brain metastases from lung cancer: The RTOG RPA intra-classes analysis. Acta Oncol 2005; 44:389–398. 19. Niwi nska A, Murawska M, Pogoda K. Breast cancer brain metastasis: Differences in survival depending on biologic subtype, RPA RTOG prognostic class and systemic treatment after whole brain radiotherapy (WBRT). Ann Oncol 2010;21:942–948.
doi:10.1016/j.ijrobp.2010.10.077
CLINICAL INVESTIGATION
Breast Cancer
NEW BREAST CANCER RECURSIVE PARTITIONING ANALYSIS PROGNOSTIC INDEX IN PATIENTS WITH NEWLY DIAGNOSED BRAIN METASTASES y NSKA, M.D., PH.D.,* AND MAGDALENA MURAWSKA, M.SC. ANNA NIWI
*Department of Breast Cancer and Reconstructive Surgery, The Maria Sk1odowska-Curie Memorial Cancer Center and Institute of Oncology, Warsaw, Poland; and yDepartment of Biostatistics, Erasmus University Medical Center, Rotterdam, The Netherlands Purpose: The aim of the study was to present a new breast cancer recursive partitioning analysis (RPA) prognostic index for patients with newly diagnosed brain metastases as a guide in clinical decision making. Methods and Materials: A prospectively collected group of 441 consecutive patients with breast cancer and brain metastases treated between the years 2003 and 2009 was assessed. Prognostic factors significant for univariate analysis were included into RPA. Results: Three prognostic classes of a new breast cancer RPA prognostic index were selected. The median survival of patients within prognostic Classes I, II, and III was 29, 9, and 2.4 months, respectively (p < 0.0001). Class I included patients with one or two brain metastases, without extracranial disease or with controlled extracranial disease, and with Karnofsky performance status (KPS) of 100. Class III included patients with multiple brain metastases with KPS of #60. Class II included all other cases. Conclusions: The breast cancer RPA prognostic index is an easy and valuable tool for use in clinical practice. It can select patients who require aggressive treatment and those in whom whole-brain radiotherapy or symptomatic therapy is the most reasonable option. An individual approach is required for patients from prognostic Class II. Ó 2012 Elsevier Inc. Brain metastases, Breast cancer, Karnofsky performance status, Prognostic index, Recursive partitioning analysis.
Recently, several prognostic indexes had been proposed in the literature (1–9). They include (1) the Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis (RPA) (1, 2) with three classes: Class I (patients <65 years, Karnofsky performance status [KPS] $70, controlled primary tumor, and no extracranial metastases), Class III (KPS <70), and Class II (all patients not in Class I or III); (2) the Score Index for Radiosurgery (3), which is the sum of scores (0–2) for each of five prognostic factors (age, KPS, status of systemic disease, number of lesions in the brain, and largest volume of the lesion); (3) the Basic Score for Brain Metastases (4), which is the sum of scores (0–1) for three prognostic factors (KPS, control of primary tumor, and extracranial metastases) and (4) the Graded Prognostic Assessment (GPA) (5) which is the sum of scores (0, 0.5, and 1) for four factors: age, KPS, extracranial metastases, and number of metastases in the brain (one, two to three, and more). The GPA has been updated and now is the sum of four factors: KPS, HER2 receptor, ER/PR receptors, and number of brain metastases (6, 9). Recently, six diagnosis-specific GPA prognostic indexes for different primary tumor types were proposed (7).
INTRODUCTION Brain metastasis in breast cancer patients is a growing problem because of the high incidence of central nervous system involvement and a lack of clear therapeutic procedures. Not all patients with brain metastases have the same prognosis, and the use of the same treatment methods for all patients is no longer appropriate. Nowadays, the treatment of brain metastasis from breast cancer includes neurosurgery, whole-brain radiation therapy (WBRT), stereotactic radiotherapy, and systemic therapy (chemotherapy, endocrine therapy, and targeted therapy). The choice and sequence of the appropriate method of treatment for a given patient should depend on many factors that are related to patient characteristics (performance status, age) and cancer characteristics such as biologic subtype of breast cancer, number of brain metastases, extracranial metastases, and location of extracranial metastases. Because patients with newly diagnosed brain metastases from breast cancer are a heterogeneous group, there is a need for introducing a simple breast cancer–specific prognostic index that can help clinicians choose a proper treatment.
Conflict of interest: none. Received July 12, 2010, and in revised form Oct 14, 2010. Accepted for publication Oct 21, 2010.
Reprint requests to: Anna Niwinska, M.D., Ph.D., Department of Breast Cancer and Reconstructive Surgery, The Maria Sk1odowskaCurie Memorial Cancer Center and Institute of Oncology, ul. Roentgena 5, 02-781 Warsaw, Poland. Tel: (+ 48) 22-644-00-24; Fax: (+48) 22-644-00-24; E-mail: [email protected] 2065
2066
I. J. Radiation Oncology d Biology d Physics
Table1. Characteristics of 441 patients with brain metastases Characteristic
No. of patients
Volume 82, Number 5, 2012
Table1. Characteristics of 441 patients with brain metastases (Continued ) % Characteristic
Age at initial diagnosis (y) Median Range Initial TNM stage I II III IV Histologic type Ductal carcinoma Lobular carcinoma Medullar, apocrinal, papillar, mucinous, planoepithelial, neuroendocrine carcinomas Cancer cells or invasive cancer after chemotherapy Estrogen/progesterone receptor status Positive Negative Missing HER2 status Positive Negative Missing Biologic subtype* Triple-negative Luminal A Luminal B HER2 Missing Localization of metastases Brain as the first or only site Liver Lung Bone Locoregional recurrence Number of brain metastases 1 2 3 Multiple RPA RTOG prognostic class I II III Type of treatment of brain metastases Neurosurgery 1 metastasis operated 2 metastases operated Radiotherapy 40 Gy/20 fractions 30 Gy/10 fractions or 20 Gy/5 fractions No radiotherapy Systemic therapy after brain metastases Yes No Type of systemic therapyy Hormonal therapy Chemotherapy Targeted therapyz Type of chemotherapyy Vinorelbine Capecitabine
50 21–76 44 181 148 68
10 41 34 15
352 30 21
80 7 5
38
8
195 238 8
44 54 2
232 196 13
53 44 3
111 81 101 127 21
25 18 23 29 5
90 146 224 198 137
20 33 51 45 31
109 31 10 291
25 7 2 66
43 253 145
10 57 33
77/441 68 9 432/441 46 386
17 88 12 98 10 88
9
2
307 134
70 30
73/307 255/307 113/307
24 83 37
89/255 35 63/255 25 (Continued )
Anthracycline Platinum Taxane Fluorouracil Cyclophosphamide Etoposide Other Type of endocrine therapyy Aromatase inhibitors Tamoxifen Goserelin Megestrol acetate Fulvestrant Type of targeted therapyy Trastuzumab Lapatinib
No. of patients
%
42/255 35/255 27/255 20/255 18/255 15/255 4/255
16 14 11 8 7 6 2
38/73 19/73 8/73 8/73 7/73
52 26 11 11 10
102/113 11/113
90 10
Abbreviations: RPA = recursive partitioning analysis; RTOG = Radiation Therapy Oncology Group. * Triple-negative, ER/PR negative HER2-negative; luminal A, ER/PR-positive HER2-negative; luminal B, ER/PR positive HER2-positive; HER2, ER/PR-negative HER2-positive. y In some patients many types of systemic treatment were ordered. z 37% of all patients and 50% of patients with HER2-positive breast cancer.
The aim of the present study was to analyze the outcome in 441 consecutive patients with breast cancer and brain metastases based on factors related to patient, cancer, and treatment. The main goal of the study was to develop a new breast cancer RPA prognostic index for patients with newly diagnosed brain metastases as a guide in clinical decision making. It was created and based only on factors related to patient characteristics and cancer. METHODS AND MATERIALS Data from 441 prospectively collected consecutive patients with breast cancer and brain metastases, treated in the Breast Cancer Department at the Cancer Center and Institute of Oncology in Warsaw, Poland, from January 1, 2003, to December 31, 2009, formed the basis for this study. Brain metastases were detected by magnetic resonance imaging (97%) or computed tomography (3%). All patients were prospectively evaluated after the detection of brain metastases. In each case, treatment options were approved by a team of medical oncologists, radiation oncologist, and neurosurgeon and were performed after the patients signed written consent. WBRT was performed in 98% of patients using a 4- to 6-MV photon beam by two lateral opposed standard fields covering all intracranial contents. The most common regimen of WBRT was 30Gy in 10 fractions. Nine patients (2%) were not irradiated because of poor performance status at the diagnosis of brain metastases. WBRT was performed in most cases by the same radiation oncologist (A.N.). Seventy-seven patients (17%) with one or two brain metastases were operated on before receiving WBRT. Systemic treatment after WBRT was ordered in 307 (70%) patients. Most patients (83%) received chemotherapy. Schedules with vinorelbine and capecitabine were the most frequent types of chemotherapy
New index in breast cancer brain metastases d A. NIWINSKA AND M. MURAWSKA
Table 2. Factors influencing survival from brain metastases: univariate analysis Covariate Patient-related factors KPS Age (y) Cancer-related factors Biologic subtype* Extracranial disease Extracranial disease Locoregional recurrence Lung metastases Liver metastases Bone metastases Number of brain metastases Treatment-related factors Neurosurgery WBRT Systemic treatment
Comparison
p value
$70 vs. <70 <40 vs. 40–65 40–65 vs. >65 #65 vs. >65
<0.0001 NS 0.009 0.002
Triple-negative vs. HER2 Triple-negative vs. luminal B Triple-negative vs. luminal A Present vs. absent Absent or controlled vs. uncontrolled Present vs. absent
<0.001 <0.0001 <0.0001 0.016 <0.0001 0.005
Present vs. absent Present vs. absent Present vs. absent 1 vs. 2
NS 0.002 NS NS
1 vs. 3 1 vs. multiple
NS 0.0001
Yes vs. no 20 Gy vs. 30 Gy 20 Gy vs. 40 Gy Yes vs. no
0.0001 0.0001 0.0001 0.0001
Abbreviations: KPS = Karnofsky performance status; NS = not significant; WBRT = whole-brain radiotherapy. * Triple-negative, ER/PR negative HER2-negative; luminal A, ER/PR-positive HER2-negative; luminal B, ER/PR positive HER2positive; HER2, ER/PR-negative HER2-positive.
used. Hormonal therapy was used in 24% of patients. Aromatase inhibitors were the most frequently ordered. Targeted therapy (trastuzumab or lapatinib) was ordered in 113 patients, which included 50% of all HER2-positive population in the study. Out of 441 patients, 420 patients were divided into four biologic subtypes based on the expression of ER, PgR, and HER2 receptors through immunohistochemistry and fluorescence in situ hybridization. The patients were divided into four biologic subtypes as follows: (1) triple-negative (ER-negative, PgR-negative, HER2-negative), (2) HER2 (HER2-positive, ER-negative, PgR-negative), (3) luminal B (HER2-positive, ER-positive and/or PgR-positive), and (4) luminal A (ER-positive and/or PgR-positive HER2-negative). Triple-negative and luminal A subsets were HER2-negative. The HER2 and luminal B subsets were HER2-positive. Clinical characteristics and type of systemic treatment after WBRT are presented in Table 1.
Statistical analysis To determine the factors influencing survival from brain metastases, univariate analysis was performed. Patient-related covariates taken into the analysis were age (<40 vs. 40–65 and vs. >65 years) and Karnofsky performance status ($70 vs. <70). Cancer-related factors were biologic subtype of breast cancer (triple-negative vs. luminal A, triple-negative vs. HER2, triple-negative vs. luminal B), locoregional recurrence (present vs. absent), extracranial metastases (present vs. absent), extracranial disease (absent/controlled
2067
vs. uncontrolled), lung metastases (present vs. absent), liver metastases (present vs. absent), bone metastases (present vs. absent), and number of brain metastases (1 vs. 2–3 and 1 vs. multiple). Treatment-related factors included neurosurgery (yes vs. no), WBRT (20 Gy vs. 30 Gy and 20 Gy vs. 40 Gy), and systemic therapy (yes vs. no). Factors that were significant in univariate analysis were included in the recursive partitioning tree. Recursive partitioning analysis, a method of building a decision tree to model predictors as described by Gordon and Olshen (10), was applied to establish prognostic groups. Each variable was examined for the best split within that variable, and all variables were examined for the best split for a given population. A variable was defined as either nominal or ordered. The first node of RPA included all patients because all variables were examined in this node. The Kaplan-Meier method was used by RPA to estimate time to event. The node was split if the log-rank statistic was significant for any variable beyond the 0.05 probability level. The significance level was adjusted for the number of multiple comparisons. Two RPA trees were built: The first included covariates related to patient, cancer, and treatment; the second assessed only factors related to patient and cancer. The first RPA analysis was performed to distinguish several groups of breast cancer patients with brain metastases with different survivals depending on type of treatment, and the second was performed to develop a new RPA breast cancer prognostic index based only on factors related to patient and cancer. In first tree, a total of 11 and in the second a total of eight variables were tested in the first run (first node) in the recursive partitioning algorithm. Age, KPS, and number of brain metastases were analyzed as continuous variables. Pairwise comparisons were performed to analyze statistically significant differences between prognostic Classes I, II, and III (log-rank, Mantel-Cox test). Additionally, survival from brain metastases based on the RPA RTOG prognostic index proposed by Gaspar et al. (1) and survival depending on KPS as the only variable was presented.
RESULTS The median survival from brain metastases in the entire group was 7 months (95% confidence interval 6.036–8.28). The results of univariate analysis are presented in Table 2. Eight statistically significant factors were related to patient and cancer and three others were statistically significantly related to treatment. Table 3 presents 10 terminal nodes of the recursive partitioning tree using factors related to patient, cancer, and treatment, with median survival ranging from 2 to 29 months. Five statistically significant variables were selected: KPS, biologic subtype of breast cancer, liver metastases, neurosurgery, and systemic treatment after WBRT. A recursive partitioning tree of the new breast cancer RPA prognostic index is presented in Fig. 1. Six terminal nodes of the recursive partitioning tree had median survival ranging from 2.4 to 29 months. Three prognostic classes of the new breast cancer RPA prognostic index were selected. The median survival of patients within prognostic Classes I, II, and III was 29 months, 9 months, and 2.4 months, respectively (p < 0.0001). Class I included patients with one or two brain metastases, without extracranial disease or with controlled extracranial disease, and with KPS of 100. Class III included patients with multiple brain metastases with KPS of #60. Class II included all other cases. Pairwise
2068
I. J. Radiation Oncology d Biology d Physics
Volume 82, Number 5, 2012
Table 3. Median survival of patients by factors related to patients, cancer, and treatment: 10 nodes of RPA tree (426 patients)
Node 1 2 3 4 5 6 7 8 9 10
Variables With systemic treatment after WBRT, operated, with controlled extracranial disease With systemic treatment after WBRT, operated, with uncontrolled extracranial disease With systemic treatment after WBRT, without operation, HER2-positive, KPS 100, without liver mets With systemic treatment after WBRT, without operation, HER2-positive, KPS<=90, without liver mets With systemic treatment after WBRT, without operation, HER2-positive, with liver mets With systemic treatment after WBRT, without operation, TNBC or Luminal A, KPS >60 With systemic treatment after WBRT, without operation, TNBC or luminal A, KPS #60 Without systemic treatment after WBRT, KPS >80 Without systemic treatment after WBRT, KPS 60–80 Without systemic treatment after WBRT, KPS #50
No. of patients
Median survival (mo)
25
29
25
20
30
22
51
9
67
9
67
7
29
3
23
7
55
3
54
2
Abbreviations: RPA = recursive partitioning analysis; KPS = Karnofsky performance status; mets = metastases; WBRT = whole-brain radiation therapy; HER2-positive = all patients with HER2 receptor positive (HER2 and luminal B); TNBC = triplenegative breast cancer (ER-negative, PR-negative, HER2-negative); luminal A (ER/PR-positive, HER2-negative).
comparisons revealed statistically significant differences in survival between the three prognostic classes (I vs. II, p < 0.0001; I vs. III, p < 0.0001; II vs. I, p < 0.0001; II vs.III, p < 0.0001; III vs. I, p < 0.0001; III vs II, p < 0.0001). The median survivals of three prognostic classes are presented in Table 4 and Fig. 2. According to the RPA RTOG proposed by Gaspar et al. (1), the survival of patients in prognostic Classes I, II, and III was 16 months, 10 months, and 2.7 months, respectively (p < 0.0001). The results are presented in Table 5. Table 6 presents survivals from brain metastases depending on KPS status. A direct proportional relationship between KPS and survival was revealed. DISCUSSION Why a new prognostic index? The stratification of patients with brain metastases into prognostic classes could be useful for everyday clinical practice because it helps in the choice of proper treatment. Several prognostic indexes that have been proposed since 1997 are related mainly to patients with lung cancer. The RPA
Fig. 1. Recursive tree from a database of 441 patients with breast cancer brain metastases after the inclusion of eight prognostic factors related to patient and cancer. KPS = Karnofsky performance status.
RTOG prognostic index proposed by Gaspar et al. (1) was based on 1,200 patients from three RTOG trials, but only 137 (12%) of them were patients with breast cancer. A comparison of four prognostic indices, performed by Sperduto et al. (5) on 1,960 patients from five RTOG clinical trials (11–15), included only 222 (11%) patients with breast cancer; however, a recent study by Sperduto et al. concerned 642 patients with breast cancer brain metastases (7). In the present study we assessed the usefulness of new prognostic indices in a group of 441 consecutive breast cancer patients. The rationale for introducing a breast cancer–specific prognostic index is that breast cancer differs from other primaries in terms of biology, survival, and response to systemic treatment. To build a new breast cancer prognostic index, we analyzed eight factors related to patient and disease, including a new one: number of brain lesions and biologic subtype of breast cancer. In our breast cancer RPA prognostic index, 6% of patients were in prognostic Class I, with median survival of 29 months. In such patients, combined modality treatment with neurosurgery/WBRT or radiosurgery/WBRT and systemic therapy should be a reasonable treatment option. By contrast, one fourth of our patients were in prognostic Class III and had a poor prognosis, with median survival of 2.4 months. Based on our previous experience (16), most such patients do not benefit from WBRT, and symptomatic therapy (corticosteroids) is equivalent to WBRT. These patients could be spared the time, costs, and side effects of aggressive treatment. Patients belonging to these two extreme prognostic classes can be easily selected and properly treated. The prognostic Class II includes more than half of breast cancer patients with great heterogeneity, and those patients require an individual therapeutic approach. Breast cancer–specific prognostic factors In our new breast cancer–specific RPA prognostic index, KPS, number of brain metastases, and status of extracranial
New index in breast cancer brain metastases d A. NIWINSKA AND M. MURAWSKA
2069
Table 4. New breast cancer recursive partitioning analysis prognostic index after inclusion of eight factors related to patient and cancer Class I II III
Characteristic
No. of patients (%)
1–2 brain metastases and extracranial disease absent or controlled and KPS 100 All others Multiple brain metastases and KPS #60
25 (6%)
Median survival (mo) 29
296 (67%) 120 (27%)
95% CI
Significance
(11.544; 45.912)
9 2.4
(8.028; 10.752) (2.052; 2.676)
p < 0.0001
Abbreviations: CI = confidence index; KPS = Karnofsky performance status.
disease were the most important prognosticators. By contrast to the RPA RTOG prognostic index (1) and other indexes (5, 6), in the present study, age was not a statistically significant factor influencing survival of patients with brain metastases from breast cancer.
Performance status Our data along with published data revealed that KPS is a key determinant of prognosis in patients with breast cancer and brain metastases (1, 2, 6, 7, 16, 17). Sperduto et al. (7) demonstrated the importance of KPS in patients with lung, breast, renal, and gastrointestinal cancers as well as in cases of melanoma. In the present study we confirmed that KPS is the most important prognostic factor in breast cancer patients with brain metastases (Tables 2–5). The survival of patients with Class III of the new breast cancer RPA prognostic index (2.4 months), in which KPS was a significant variable, was as poor as the survival of patients in Class III of the RPA RTOG prognostic index by Gaspar et al. (2.3 months) (1) and in Class III lung cancer patients from the Polish study by Ke˛pka et al. (2.5 months) (18). In the present study, patients with single brain metastasis and poor performance status had an unsatisfactory prognosis. It was comparable, or even worse, to the prognosis of
patients with multiple brain metastases and good performance status. Number of brain metastases The important limitation of some previously proposed prognostic indexes was the avoidance of the number of brain metastases (1, 2). The number of brain lesions has been shown to be an important factor influencing survival in a randomized trial (11). The authors found a survival advantage for patients with one metastasis, but not for those with two or three metastases. In our RPA tree, the number of brain metastases was treated as a continuous variable. The results suggest that patients with one or two brain lesions with KPS of 100 and without extracranial disease or with controlled extracranial disease have the best prognosis. These patients should be given very aggressive treatment. Intensive local treatment can control the disease in the brain for a long time, and systemic treatment ordered after local treatment can control occult extracranial metastases. This treatment policy can improve survival significantly. Controlled/uncontrolled extracranial disease In the RPA RTOG prognostic index proposed by Gaspar et al. (1), the status of the primary tumor and extracranial metastases was one of the most important prognostic factors influencing the survival of patients with brain metastases. However, in some publications that appeared later, the factor called uncontrolled disease has been assessed as subjective and not convincing (5). In the present study, restaging of every patient was performed after the detection of brain metastases. The status of the primary tumor (local failure present/absent and controlled/uncontrolled) and the status of the extracranial metastases (present/absent and controlled/uncontrolled) was assessed in every case. For the present study, patients with controlled extracranial disease were those with a complete Table 5. Median survival of 441 patients in RPA RTOG prognostic classes by Gaspar (1) RPA RTOG prognostic class I II III
Fig. 2. Breast cancer recursive partitioning analysis prognostic index: survival from brain metastases (meta) within three prognostic cases.
No. of patients (%)
Median survival (mo)
Significance
43 (10) 253 (57) 145 (33)
16 10 2.7
p < 0.0001
Abbreviations: RPA = recursive partitioning analysis; RTOG = Radiation Therapy Oncology Group.
I. J. Radiation Oncology d Biology d Physics
2070
Table 6. Median survival of 441 patients depending on Karnofsky performance status (KPS) KPS
No. of patients
Median survival (mo)
Significance
100 90 70–80 #60
120 55 121 145
16 12 7 2.7
p < 0.0001
or partial response to the treatment of locoregional relapse or distant metastases. They did not require active treatment of extracranial disease at the time of detection of brain lesions. They usually had single/several bone or lung metastases or local recurrence, which had been successfully treated before the detection of brain metastases. In fact, the term ‘‘controlled extracranial disease’’ means that the probability of survival is at least 6 months, with a low risk of death because of locoregional failure or extracranial metastases. In the present study, the variable ‘‘controlled/uncontrolled extracranial disease’’ was more valuable than ‘‘extracranial metastases (yes vs. no)’’ or ‘‘locoregional failure (yes vs. no)’’; the median survival of patients with single brain metastases and controlled extracranial disease was comparable to that of patients with solitary brain metastasis (one brain lesion without extracranial disease). Biologic subtype The biologic subsets of breast cancer differ according to natural prognosis and response to systemic treatment (17, 19). Triple-negative and HER2-positive breast cancer have a special predisposition to metastasize to the brain, and the natural prognosis of these subtypes is poor. However, after systemic treatment that follows whole-brain radiotherapy, the survival of these two types has changed significantly. In the posttrastuzumab era, the survival of HER2-positive breast cancer patients is the longest of all biologic subtypes (17). It exceeds historical estimates and argues against a nihilistic approach in this subgroup of patients. By contrast, the triple-negative subtype has the poorest prognosis because of the natural rapid progress of the disease and the lack of effective drugs against systemic and central nervous system diseases. In the present study, the biologic subtype of breast cancer was significant only when 11 variables, together with treatment-related ones, were analyzed.
Volume 82, Number 5, 2012
Among all biologic subsets, the most favorable survival was assessed in patients with HER2-positive subtype, without liver metastases. In our new breast cancer RPA prognostic index, biologic subtype was not a significant variable.
Strength and limitations of the study Strength. In most clinical trials, only patients with good performance status are analyzed. By contrast, our group consisted of nonselected patients, together with those with poor performance status. This caused the results of the present study to be comparable with the results observed in everyday clinical practice. The differences in survival between prognostic Classes I, II, and III are significant and distinguishable, and this enables us to make quick decisions regarding treatment policy. The results confirm the value of the number of brain metastases as a prognostic factor that helps in treatment decision making. We also analyzed, for the first time to our knowledge, the value of biologic subtype of breast cancer. Limitations. It is very challenging to develop a new prognostic index based on a group of patients who have already been treated, because the outcome in those patients depends not only on factors related to patient and cancer, but also on treatment interventions. From an ethical point of view, it is not feasible to draw up a prognostic index on patients who have not been treated. Although we included only factors related to patients and cancer in the new index, we obtained survivals of previously treated patients, and this can be an important methodologic bias influencing our results.
CONCLUSIONS The new breast cancer RPA prognostic index can act as a guide for clinical decision making for the heterogeneous population of patients with brain metastases. It can distinguish between the group of patients with good prognosis (median survival 29 months) requiring aggressive combined treatment policy and the group of patients with very poor prognosis (median survival 2.4 months) for whom WBRT is questionable and symptomatic treatment is a reasonable option.
REFERENCES 1. Gaspar L, Scott CH, 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. 2. Gaspar L, Scott C, Murray K, et al. Validation of the RTOG recursive partitioning analysis (RPA) classification for brain metastases. Int J Radiat Oncol Biol Phys 2000;47: 1001–1006. 3. Weltman E, Salvajoli JV, Brandt RA, et al. Radiosurgery for brain metastases: A score index for predicting prognosis. Int J Radiat Oncol Biol Phys 2000;46:1155–1161.
4. Lorenzoni J, Devriendt D, Massager N, et al. Radiosurgery for treatment of brain metastases: Estimation of patient eligibility using three stratification systems. Int J Radiat Oncol Biol Phys 2004;60:218–224. 5. Sperduto PW, Berkey B, Gaspar LE, et al. A new prognostic index and comparison to three other indices for patients with brain metastases: An analysis of 1960 patients in the RTOG database. Int J Radiat Oncol Biol Phys 2008;70:510–514. 6. Sperduto PW, Xu P, Sneed P, et al. A graded prognostic assessment (GPA) for women with breast cancer and brain metastases [Abstract]. ASCO Proceedings J Clin Oncol 2010;28:12.
New index in breast cancer brain metastases d A. NIWINSKA AND M. MURAWSKA
7. Sperduto PW, Chao ST, Sneed PK, et al. Diagnosis-specific prognostic factors, indexes, and treatment outcomes for patients with newly diagnosed brain metastases: A multiinstitutional analysis of 4,259 patients. Int J Radiat Oncol Biol Phys 2010;77:655–661. 8. Le Scodan R, Massard Ch, Mouret-Fourme E, et al. Brain metastases from breast carcinoma: Validation of the Radiation Therapy Oncology Group recursive partitioning analysis classification and proposition of a new prognostic score. Int J Radiat Oncol Biol Phys 2007; 69:839–845. 9. Sperduto CM, Watanabe Y, Mullan J, et al. A validation study of a new prognostic index for patients with brain metastases: The Graded Prognostic Assessment. J Neurosurg 2008;109: 87–89. 10. Gordon L, Olshen RA. Tree-structured survival analysis. Cancer Treat Rep 1985;69:1065–1069. 11. Andrews DW, Scott CB, Sperduto PW, et al. Whole brain radiation therapy with or without stereotactic radiosurgery boost for patients with one to three brain metastases: Phase III results of the RTOG 9508 randomized trial. Lancet 2004;363: 1665–1672. 12. Komarnicky LT, Philips TL, Martz K, et al. A randomized phase III protocol for the evaluation of misonidasole combined with radiation in the treatment of patients with brain metastases (RTOG 79-16). Int J Radiat Oncol Biol Phys 1991;20:53–58.
2071
13. Sause WT, Scott C, Kirsch R, et al. Phase I/II trial of accelerated fractionation in brain metastases, RTOG 85-28. Int J Radiat Oncol Biol Phys 1993;26:653–657. 14. Phillips TL, Scott CB, Leibel S, et al. Results of a randomized comparison of radiotherapy and bromodeoxyuridine to radiotherapy alone for brain metastases: Report of RTOG trial 89-05. Int J Radiat Oncol Biol Phys 1995;33:339–348. 15. Murray KJ, Scott C, Greenberg HM, et al. A randomized phase III study of accelerated hyperfractionation versus standard in patients with unresected brain metastasis: A report of RTOG 9104. Int J Radiat Oncol Biol Phys 1997;39:571–574. 16. Komosi nska K, Ke˛pka L, Niwi nska A, et al. Prospective evaluation of the palliative effect of whole-brain radiotherapy in patients with brain metastases and poor performance status. Acta Oncol 2010;49:382–388. 17. Niwi nska A, Murawska M, Pogoda K. Breast cancer subtypes and response to systemic treatment after whole-brain radiotherapy in patients with brain metastases [Abstract]. ASCO Proceedings J Clin Oncol 2010;28:121. 18. Ke˛pka L, Cieslak E, Bujko K, et al. Results of the whole-brain radiotherapy for patients with brain metastases from lung cancer: The RTOG RPA intra-classes analysis. Acta Oncol 2005; 44:389–398. 19. Niwi nska A, Murawska M, Pogoda K. Breast cancer brain metastasis: Differences in survival depending on biologic subtype, RPA RTOG prognostic class and systemic treatment after whole brain radiotherapy (WBRT). Ann Oncol 2010;21:942–948.