Neuro-Oncology Advance Access published December 4, 2013

Neuro-Oncology

Neuro-Oncology 2013; 0, 1 – 8, doi:10.1093/neuonc/not200

Assessment of prognostic scores in brain metastases from breast cancer Emeline Tabouret, Philippe Metellus, Anthony Gonc¸alves, Benjamin Esterni, Emmanuelle Charaffe-Jauffret, Patrice Viens, and Agne´s Tallet

Corresponding author: Emeline Tabouret, MD, Institut Paoli Calmette, 232 Bd Sainte Marguerite, 13009 Marseille, France ([email protected]).

Background. Breast cancer (BC) is the second most common cause of brain metastases (BM). Optimal management of BM from BC is still debated. In an attempt to provide appropriate treatment and to assist with optimal patient selection, several specific prognostic classifications for BM from BC have been established. We evaluated the prognostic value and validity of the 6 proposed scoring systems in an independent population of BC patients with BM. Methods. We retrospectively reviewed all consecutive BC patients referred to our institution for newly diagnosed BM between October 1995 and July 2011 (n ¼ 149). Each of the 6 scores proposed for BM from BC (Sperduto, Niwinska, Park, Nieder, Le Scodan, and Claude) was applied to this population. The discriminative ability of each score was assessed using the Brier score and the C-index. Individual prognostic values of clinical and histological factors were analyzed using uni- and multivariate analyses. Results. Median overall survival was 15.1 months (95% CI,11.5– 18.7). Sperduto-GPA (P , .001), Nieder (P , .001), Park (P , .001), Claude (P , .001), Niwinska (P , .001), and Le Scodan (P ¼ .034) scores all showed significant prognostic value. The Nieder score showed the best discriminative ability (C-index, 0.672; Brier score error reduction, 16.1%). Conclusion. The majority of prognostic scores were relevant for patients with BM from BC in our independent population, and the Nieder score seems to present the best predictive value but showed a relatively low positive predictive value. Thus, these results remain insufficient and challenge the routine use of these scoring systems. Keywords: brain metastases, breast cancer, prognostic scores.

Breast cancer (BC) is the most prevalent neoplasm in adult women with more than 209 000 new diagnoses in the United States and 50 000 new cases per year in France.1 Improvements in systemic therapy have increased the overall survival of BC patients, including metastatic patients.2 In the context of controlled systemic disease, the prevalence of brain metastases (BM) from BC is increasing.3 BC is the second leading cause of BM after lung cancer and accounts for 17% –20% of all cases.4,5 Median survival after BM onset depends on many factors related to patient characteristics, disease status, and treatment modalities.6 – 8 BM treatment options currently include whole-brain radiotherapy (WBRT), surgery, stereotactic radiosurgery (SRS), or a combination of these methods. Recently, a growing body of evidence indicates that systemic therapies may have antitumor activity at the central nervous system level.9 However, the optimal management of BM is still debated.

In an attempt to provide appropriate treatment and to assist with optimal patient selection, several prognostic classifications for BM have been established,10 – 15 which were mainly based on patients with non-small cell lung cancer (52.4% –77% of cases). More recently, based on the assumption that the prognosis of patients with newly diagnosed BM would differ depending on the primary site,16 many authors have described independent prognostic factors in the particular setting of BMBC, and 6 prognostic classifications for this patient subgroup have been proposed.10 – 15 However, these scoring systems were not validated with an independent population, and they have not been compared with each other in the same population. The aim of our study was to evaluate the prognostic value and validity of these 6 scoring system in an independent BC population with newly diagnosed BM who were referred to our institution for treatment.

Received 10 July 2013; accepted 2 October 2013 # The Author(s) 2013. Published by Oxford University Press on behalf of the Society for Neuro-Oncology. All rights reserved. For permissions, please e-mail: [email protected].

1 of 8

Downloaded from http://neuro-oncology.oxfordjournals.org/ at Aston University on January 26, 2014

APHM, Timone Hospital, Departement of Neuro-Oncology, Marseille, France (E.T.); L’institut Paoli-Calmettes, Department of Medical Oncology, Marseille, France (E.T., A.G.); APHM, Timone Hospital, Department of Neurosurgery, Marseille, France (P.M.); L’institut Paoli-Calmettes, Department of Biostatistics, Marseille, France (B.E.); UMR911, CRO2, Aix-Marseille Universite´, Marseille, France (P.M.); L’institut Paoli-Calmettes, Department of Anatomic Pathology, Marseille, France (E.C.-J.); L’institut Paoli-Calmettes, Department of Radiotherapy, Marseille, France (A.T.)

Tabouret et al: Prognostic score for brain metastases

Results

We conducted a retrospective monocentric analysis using the medical records of patients newly diagnosed with BM from BC between October 1995 and July 2011. Patients with leptomeningeal disease synchronous with or before BM diagnosis were excluded. This study was approved by our local ethics committee and complies with the principles of the Declaration of Helsinki. All patients had histologically proven breast carcinoma. BM was diagnosed or confirmed through brain magnetic resonance imaging (MRI) with or without pathological confirmation. Treatment modalities included WBRT, surgical resection, SRS, and systemic treatment (such as chemotherapy, endocrine therapy, and targeted therapy). BM treatment consisted of these modalities alone or in combination. We extracted the following data from the patients’ charts when available: age (,60 years old or .60 years old), Karnofsky performance status (KPS) at the time of BM diagnosis, number of BM, hormonal receptor status (estrogen receptor [ER] and progesterone receptor [PR]), human epidermal growth factor receptor 2 (HER2) expression profile in the primary tumor, tumor subtype, presence of extracranial metastases (ECM), BM as the first recurrence site (not exclusive of concomitant systemic site), BM treatment modalities, interval between initial diagnosis of BC and diagnosis of BM (,40 months or .40 months), interval between first diagnosis of metastasis and diagnosis of BM (,11.5 months or .11.5 months), and lymphocyte count. Lymphopenia was defined as ,0.7 G/l on blood count at BM diagnosis. Controlled systemic disease was defined as no progressive disease on 2 subsequent evaluations that were 3 months apart. Immunohistochemistry (IHC) was carried out for the evaluation of ER, PR, and HER2 expression. Fluorescence in situ hybridization (FISH) analysis for HER2 amplification was carried out for an IHC score of 2+. Tumor subtype may be approximated as follows: basal (triple negative12 or HER2/ER/PR negative), luminal A (HER2 negative, ER/PR positive), luminal B (HER2/ER/PR positive), and HER2 (HER2 positive, ER/PR negative). Absence of extracerebral disease was defined as complete remission of initial BC and absence of metastatic sites other than BM. Each of the 6 prognostic scores proposed for BM from BC was applied to this population, provided that all prognostic factors required for its assignation were available (Supplementary data, Table 1), leading to a different sample size for each model. Overall survival (OS) was defined as the time from BM diagnosis to death from any cause, censored at the date of last contact. The date of BM diagnosis was defined as the day when BM was confirmed by imaging or by pathological examination of resected specimens. OS was estimated by the Kaplan–Meier product limit method. The log-rank test was used to compare survival rates by univariate analysis. Prognostic factors with P , .15 in univariate analysis were explored by multivariate analysis. Cox proportional hazards models were used for multivariate analyses and to estimate hazard ratios in regression models. Multivariate analyses were restricted to the subset of patients (140 of 152) with all the aforementioned prognostic factors available. Reported P-values were 2-sided, and P , .05 was considered statistically significant. The scores’ predictive values were estimated by the C-index and the Brier score. The C-index (Harrell’s concordance index) was used to estimate the discriminative ability and the concordance with the final multivariate model. Brier scores measured the mean square error of each score class, while Brier relative error reduction measured the improvement, by the scoring system evaluated, of the predictions relative to a model that gave all women the same outcome probability. In order to evaluate scoring system accuracy, positive predictive values (PPV) of the worst and best classes were analyzed. PPV was defined as the proportion of patients in the worst prognostic group with an OS less than 3.5 months (first quartile of OS) and the proportion of patients in the best prognostic group with an OS more than median OS (14.0 months). Statistical analyses were performed with SPSS version 17 (SPSS Inc., Chicago) and R version 2.15.2.17

Population Characteristics

2 of 8

During the study period, 152 consecutive patients with newly diagnosed BM from BC were referred to our institution for treatment. The median times from initial diagnosis to BM occurrence was significantly longer (P , .001) for patients with luminal A subtype (65.8 months; 95% CI, 51.4– 80.1) than for luminal B subtype (41.3 months; 95% CI, 30.5– 52.0), HER2 subtype (29.2 months; 95% CI, 13.2 –45.2), and TN subtype (30.5 months; 95% CI, 22.7– 38.4). BM was the first metastatic site for 28% of patients (37% of patients with TN status and 22% of HER2-positive patients). At BM diagnosis, 27% of patients presented with altered general status (KPS , 70). The majority of them had multiple cerebral metastases with uncontrolled systemic disease. Few patients presented with only a cerebral metastatic site (12.6%), and only 3 patients (2%) had synchronous BM at the initial diagnosis. The majority of patients underwent radiotherapy or combined local treatments with adjunct systemic treatment. Among HER2positive patients, 60% received an anti-HER2 therapy (trastuzumab, n ¼ 27; lapatinib, n ¼ 7) after BM diagnosis. In HR-positive patients, 51% received hormonal therapy after the onset of BM.

Patient Outcomes Thirty-one patients (20.4%) were alive at the last follow-up with a median follow-up since BM onset of 30.2 months (range, 2.2–81.8 months). Median OS (MOS) and 1- and 2-year survival rates since BM diagnosis were 14.0 months (95% CI, 10.8–17.2), 53%, and 25%, respectively.

Prognostic Scores Repartition of patients into prognostic classes of the 6-scoring system is shown in Table 2. Some classes were poorly represented in certain classifications; for example, there were only 2 patients in Park class 4 and 1 in Niwinska class 1. Analyses of the 6 breast prognostic scores are reported in Table 2. SperdutoGPA (P , .001), Nieder (P , .001), Park (P , .001), Niwinska (P , .001), Claude’ (P , .001), and Le Scodan’ (P ¼ .034) scores all exhibited significant prognostic value for OS. The Kaplan – Meier curves are shown in Figure 1. In order to explore the predictive value of each scoring system, we made comparisons with the C-index and the Brier score. The best C-index was found for the Nieder score (0.672), whereas the worst were for the Le Scodan and Claude scores. Relative error reduction analyzed by Brier score was maximal for the Nieder’ score with an error reduction of 16.3%. Use of the Sperduto’ model reduced relative error of only 9.7% in our population. These results suggest the best predictive and prognostic values for the Nieder score (Table 3; Supplementary data, Figure 1). PPVs were approached for the 6 models. The Nieder’ score produced the best PPV at 71%, while the Sperduto score PPV was 69% (Supplementary data, Figure 1).

Prognostic Factors (Table 4 and Figure 2) In univariate analysis, patients .60 years (P ¼ .015); KPS , 70, (P , .001), radiotherapy alone (P , .001), absence of systemic

Neuro-Oncology

Downloaded from http://neuro-oncology.oxfordjournals.org/ at Aston University on January 26, 2014

Materials and Methods

Tabouret et al: Prognostic score for brain metastases

Table 1. Patient characteristics Characteristics

Number

percents

53.8 (27.2– 83.0) 40.0 (0 –289.4)

30.5 (22.7– 38.4) 65.8 (51.4– 80.1) 29.2 (13.2– 45.2) 41.3 (36.3– 48.9) 11.3 (0 –198) 24/98 80 4/152 13/152 24/152 30/152 56/152 25/152

24.5 2.6 8.6 15.8 19.7 36.8 16.5

44/149 40/149 65/149

29.5 26.8 43.6

19/151 46/151 86/151

12.6 30.5 57

85/146 57/130 27/130 46/130 27/130 30/130

58.2 43.8 20.8 35.4 20.8 23.1

78/151 12/151 6/151 47/151

51.7 7.9 4.0 31.1

8/151 122/151 34/57

5.3 80.8 59.6

44/85

51.7

*1Two or 3 BM. *2More than 3 BM. Abbreviation: BM, brain metastases.

Scores Sperduto (n 5 129) class 3.5–4 (n ¼ 22, 17.1%) class 2.5–3 (n ¼ 49, 38%) class 1.5–2 (n ¼ 44, 34.1%) class 0– 1 (n ¼ 14, 10.9%) Nieder (n 5 150) 1 (n ¼ 20, 13.3%) 2 (n ¼ 66, 44%) 3 (n ¼ 35, 23.3%) 4 –5 (n ¼ 29, 19.3%) Park (n 5 129) I (n ¼ 43, 33.3%) II (n ¼ 64, 49.6%) III (n ¼ 20, 15.5%) IV (n ¼ 2, 1.6%) Niwinska (n 5 152) I (n ¼ 1, 0.7%) II (n ¼ 124, 81.6%) III (n ¼ 27, 17.8%) Le Scodan (n 5 100) I (n ¼ 31, 31%) II (n ¼ 43, 43%) III (n ¼ 26, 26%) Claude (n 5 132) I (n ¼ 113, 85.6%) II (n ¼ 19, 14.4%)

Median OS (months)

95% CI

P value < .001

30.0 20.2 12.3 4.7

1.7–58.1 12.8 –27.5 2.3–22.3 2.6–6.9

20.2 19.5 11.5 2.7

6.8–33.5 11.6 –27.5 8.7–14.3 2.3–3.2

18.0 20.2 2.6 0.4

11.5 –24.5 16.2 –24.1 2.2–3.0 -

< .001

< .001

< .001 16.4 2.5

12.5 –20.2 2.2–2.9

12.1 14.7 2.7

10.7 –13.4 4.1–25.3 0 –5.4

15.7 3.2

11.4 –20.1 0.5–5.9

.034

< .001

decreased OS. Shorter interval between primary and BM diagnosis (P ¼ .079), systemic disease status (P ¼ .070), and number of BMs (P ¼ .074) tended to be significant. No prognostic value was found for subtype of BC (P ¼ .420); hormonal receptor (0.760), HER2 (P ¼ .152), or TN status (P ¼ .545); or interval between first metastatic event diagnosis and BM diagnosis (P ¼ .665). MOS of TN, luminal A, HER2, and luminal B were 15.1 (95% CI, 5.6 –24.7), 12.3 (95% CI, 9.5–15), 20.2 (95% CI, 17.1 –23.2), and 20.5 months (95% CI, 11.4 –29.5), respectively (Supplementary data, Table 1). In multivariate analysis, independent prognostic predictors for better survival were KPS . 60 (P , .001; HR, 0.292), absence of systemic disease (P , .045; HR, 0.454), interval between initial diagnosis and BM . 40 months (P ¼ .044; HR, 0.637), use of systemic treatment (P , .001; HR, 0.125) and combined local treatment and/or SRS and/or surgery (P ¼ .001; HR, 0.464). Number of BM, BM as first metastatic site, and age did not have prognostic value.

Discussion therapy (P , .001), absence of anti-HER2 therapy for HER2-positive subtype (P , .010), absence of hormonotherapy for ER/PR positive patients (P ¼ .002), lymphopenia (P ¼ .022), and BM as the first metastatic site (P ¼ .022) were significantly associated with

Neuro-Oncology

The prognosis for BC patients with newly diagnosed BM is highly variable, and median survival (MS) can range from 3 months to over 2 years.10 Therapeutic decisions closely depend on estimated prognosis. The establishment, as well as validation, of prognostic

3 of 8

Downloaded from http://neuro-oncology.oxfordjournals.org/ at Aston University on January 26, 2014

Age (median, range) Interval between primary tumor diagnosis and BM (months, range) Triple negative (months, 95% CI) Luminal A (months, 95% CI) HER2 (months, 95% CI) Luminal B (months, 95% CI) Interval between first metastasis and BM (months, range) Lymphopenia (n, %) KPS (median) 40 (n, %) 50 (n, %) 60 (n, %) 70 (n, %) 80 (n, %) 90– 100 (n, %) Number of BM (n, %) Single Oligometastases*1 Multiple*2 Systemic disease (n, %) Absent Controlled Uncontrolled Subtype (n, %) ER/PR positive HER2 positive Triple negative Luminal A Luminal B HER2 Local treatment (n, %) WBRT alone Stereotactic radiosurgery Surgery Combined treatment (WBRT and surgery or SRS) No local treatment Systemic treatment (n, %) Anti-HER2 therapy for HER2 positive patients (n, %) Hormonotherapy for AR/PR positive patients (n, %)

Table 2. Log-rank analyses of the 6 scoring systems

Tabouret et al: Prognostic score for brain metastases

classification is therefore a prerequisite for making decisions regarding treatment. Previously published prognostic classifications were based on retrospective data and therefore depended on the studied population characteristics and required validation in an independent series. For the establishment of such classifications, although systemic treatments (particularly anti-HER2 therapies) show proven prognostic value, they should not be included in prognostic factors since they affect patients’ outcome after the therapeutic decision, which would be the optimal time to use prognostic scores that support therapeutic choice. Nonetheless, physicians should be aware of the crucial positive impact on survival of anti-HER2 therapies administered after the onset of BM for HER2-positive patients

4 of 8

and recognize that omission of this systemic treatment would downgrade this good prognostic factor to a poor one and consequently invalidate the prognostic value of a score that includes HER2 status. The boundary between prognostic factor and predictive factor becomes narrow because HER2 status is a favorable prognostic factor provided that a specific treatment is applied, which is the current standard of care. For the same reasons, HER2positive patients who do not receive anti-HER2 therapies at the onset of BM (due to unavailability before the anti-HER2 era) should not be included in prognostic factor studies because they bias the results; this is likely the reason why Park et al13 found that HER2-positive patients have poorer prognoses, which is the opposite of what has been reported by multiple other

Neuro-Oncology

Downloaded from http://neuro-oncology.oxfordjournals.org/ at Aston University on January 26, 2014

Fig. 1. Overall survival according to 6 scoring systems: (A) Nieder (P , .001); (B) Park (P , .001); (C) Niwinska (P , .001); (D) Sperduto (P , .001); (E) Le Scodan (P ¼ .034); and (F) Claude (P , .001).

Tabouret et al: Prognostic score for brain metastases

Table 3. Predictive value analyses of the 6 scoring systems (C-index and Brier score). Score

Classes

Number of Patients

C-index

Relative Error Reduction (Brier score)

Brier Score Reference

Brier Score Calculated

Park Nieder Sperduto Niwinska Le Scodan Claude

1/2/3– 4 1/2/3/4 1/2/3/4 1 – 2/3 1/2/3 1/2

43/64/22 20/66/35/29 22/49/44/14 125/27 31/43/26 113/19

0.635 0.672 0.638 0.610 0.585 0.560

-0.092 -0.161 -0.097 -0.101 -0.024 -0.047

0.198 0.187 0.198 0.189 0.149 0.192

0.180 0.157 0.179 0.169 0.146 0.183

studies.18 – 20 In our study, anti-HER-2 therapies were available for all HER-2 positive patients at the time of BM diagnosis, and patients who did not receive this treatment were unsuitable for this treatment due to being in generally poor condition.

Neuro-Oncology

In our prognostic classification validation study, the 6 prognostic scoring systems10 – 15 were all able to discriminate patients in groups with different survivals, but their discriminative ability and positive and negative predictive values were somewhat different.

5 of 8

Downloaded from http://neuro-oncology.oxfordjournals.org/ at Aston University on January 26, 2014

Fig. 2. Overall survival according to: (A) KPS; (B) use of anti-HER2 therapy; (C) type of local BM treatment; (D) time between initial and BM diagnosis; (E) use of systemic therapy; and (F) extracerebral disease status.

Tabouret et al: Prognostic score for brain metastases

Table 4. Univariate and multivariate analyses of potential prognostic factors in our population Factors

Univariate OS (months)

95% CI

P value

HR

95% CI of HR

.015 16.0 8.0

11.5 – 20.5 2.7– 13.3

3.0 18.0

1.9– 4.1 14.2 – 21.9

16.4 15.3 9.8

9.9– 22.8 9.0– 21.5 4.1– 15.5

31.9 11.8 12.3

0 – 65.3 6.8– 16.8 7.8– 16.5

6.1 20.8

3.4– 8.8 15.0 – 26.5

1.7 18.0

0.3– 3.1 14.2 – 21.8

7.0 30.0

0 – 14.8 17.3 – 42.6

4.5 17.5

2.8– 6.2 4.9– 30.1

15.1 12.3 20.2 20.5

5.6– 24.7 9.5– 15.1 17.2 – 23.2 11.5 – 29.5

15.9 12.7

11.6 – 20.2 9.3– 16.1

13.3 20.2

9.5– 17.0 16.6 – 23.7

15.0 15.1

10.1 – 19.8 5.6– 24.7

10.1 15.9

5.0– 15.2 10.5 – 21.2

14.7 11.5

10.4 – 19.0 5.7– 17.3

11.1 18.6

8.8– 13.5 12.1 – 25.1

11.8 4.5

9.1– 14.5 0.25 – 8.7

,.001

P value 0.322

0.292

0.180 – 0.474

.074

, .001

.378

.070

0.454

0.210 – 0.984

.045

,.001

0.464

0.291 – 0.737

.001

,.001

0.125

0.066 – 0.234

,.001

0.637

0.411 – 0.988

.044

.010

.002 .420

.760

.152

.545

.079

.665

.022

.408

.022

*Excluded of multivariate analysis because of the lack of patients. Abbreviations: BM, brain metastases; HR, hazard ratio; OS, overall survival; WBRT, whole-brain radiotherapy.

6 of 8

Neuro-Oncology

Downloaded from http://neuro-oncology.oxfordjournals.org/ at Aston University on January 26, 2014

Age (years) ,60 ≥60 KPS ,70 ≥70 Number of BM Single Oligometastases Multiple Systemic disease Absent Controlled Uncontrolled Local treatment WBRT alone Combined treatment Systemic treatment No Yes Anti-HER2 therapy* No Yes Hormonotherapy* No Yes Subtype of breast cancer Basal like Luminal A HER 2 Luminal B Hormonal receptor status Negative Positive HER2 status Negative Positive Triple negative No Yes Interval between primary tumor and BM , 40 months ≥ 40 months Interval between first metastasis and BM , 11.5 months ≥ 11.5 months BM as first metastatic site No Yes Lymphopenia* No Yes

Multivariate

Tabouret et al: Prognostic score for brain metastases

Neuro-Oncology

prognostic factors. This classification may therefore be refined, perhaps by adjusting weights given to each prognostic factor. It is also possible that using histologic subtypes as a surrogate of genetic subtypes led to inaccurate prognostic subgroups. In conclusion, the prognostic classification from Nieder et al.12 had the best C-index and Brier score. The PPV was 71%, which remains insufficient. The superiority of the Nieder’ score in the discrimination of patients with newly diagnosed BM from breast cancer could attributed to the prognostic factors included in this score. Indeed, the Nieder’ score is one of the few to consider ECM status as an independent prognostic factor, which also demonstrated significant correlation with survival in our multivariate analysis. The Niwinska’ score also included ECM status as a prognostic factor but stratified patients less accurately. This was likely due to the highly restrictive criteria assigning patients to classes I and III, so that most patients (89%) were classified in the intermediate group that therefore became highly heterogeneous. Among the 4 other BCBM scores, 3 did not include ECM status in their prognostic factor analysis.13 – 15 Sperduto et al established a BC-GPA based on a population of more than 800 patients with newly diagnosed BM from breast cancer treated in 11 institutions.23They included ECM status in prognostic factors analyzed for survival, but the definition of presence or absence of ECM was lacking. Did all institutions consider the same definition of “absence of ECM”? This definition is crucial, and it could be possible that patients having ECM present but nonprogressive (stable or regressive under treatment) had been classified as without or with ECM according to the investigators. The publication from Sperduto et al does not address this question. Our results compare favorably with those of Braccini et al.,27 who compared performances of several prognostic indices for BC patients with BM and found that RPA (Recursive Partitioning Analysis) based on KPS, age, and systemic disease status28 performed better than the new prognostic index developed for BCBM. Another concern that could contribute to the limited value of these scoring systems is that none of them include the neurologic condition of the patient. However, some authors have already highlighted its impact on survival of patients with BM.29,30 The inclusion of this item in prognostic factors analyses should therefore be systematic because it would influence the prognostic classifications arising.

Conclusion Independent prognostic factors differ among studies, and validation of prognostic scores in independent series is mandatory to establishing a useful tool for treatment decision making and refining prognostic classifications.

Supplementary Material Supplementary material is available online at Neuro-Oncology (http://neuro-oncology.oxfordjournals.org/).

Funding None declared.

7 of 8

Downloaded from http://neuro-oncology.oxfordjournals.org/ at Aston University on January 26, 2014

The classification from Le Scodan et al14 provided 3 prognostic classes with MOS of 19.5, 12.5, and 3.5 months for classes I, II, and III, respectively. The positive predictive value of this classification was 47%. The main weakness of this prognostic classification is that TN patients always belong to class III regardless of other factors. Twenty-five patients were TN in our study, and their MOS was 15 months, a particularly high MOS for this subtype perhaps biased by small sample sizeonetheless, despite unfavorable, TN subtype in itself is probably not sufficient to predict a poor outcome. Moreover, TN subtype is heterogeneous breast cancer subtype, and not all TN have the outcome of basal tumors, which were frequently associated to TN subtype. This could explain the poor Brier score (reduction of 2.4%). Using the prognostic classification from Claude et al.15 based on 2 prognostic factors (KPS and lymphopenia), 2 prognostic groups could be differentiated (poor and good) with 3.2 and 15.7 months MOS, respectively. The positive predictive value was 54% (Supplementary data, Table 2), and only 31% of patients who lived less than 4 months were classified in the poor prognostic group. Therefore, the 2 prognostic factors for this classification appear insufficient to determine BMBC patient outcome. Because only 1 patient belonged to the best prognostic class in the classification described by Niwinska et al,11 patients in class II were considered to have good prognoses. The MOS of patients assigned to class II was 16.4 months. The positive predictive value was 59%. Patients who belonged to the worse prognostic group had mostly KPS ≤ 50 and were defined as class III; all but 4 of these patients had a survival time less than 3.5 months. This criterion therefore appears a good one to highlight patients with a very poor prognosis, with the goal being to avoid aggressive treatments and perhaps consider the best supportive care.21,22 Nonetheless, 73% of patients with poor prognosis had a KPS . 50 at the onset of BM and were therefore misclassified in class II. Despite its relevance, KPS as single prognostic factor for classification appears insufficient to identify patients with poor prognosis. The patients in the population-based study by Park et al.13 did not receive any anti-HER2 therapy at the onset of BM, explaining why having HER2-positive status appeared to be an adverse prognostic factor, which is the opposite of all other studies.7,23 Moreover, 1 of the 3 independent prognostic factors was systemic treatment-related, making this classification less relevant for therapeutic decision. Indeed, patients assigned to the best prognostic class (class I) had a worse MOS than patients classified in prognostic class II. The last 2 classifications weighted prognostic factors, with greater importance being given to the KPS. Sperduto et al.10 investigated 6 prognostic factors (age, KPS, genetic subtype, number of BM, presence of extracranial metastases) and time from primary diagnosis to development of BM (TPDBM), which was found to be correlated with tumor subtype24 – 26 (as we also highlighted). Prognostic factors were weighted relative to the magnitude of their regression coefficients. MOS for patients assigned to classes I, II, III, and IV were 30.0, 20.2, 12.3, and 4.7 months, respectively. The PPV was 69%, which leads to consequent misclassifications. BM treatment modalities applied to our population were close to those applied in the population-based study by Sperduto et al. and could not explain the lack of power of the PPV. The PPV rose only to 70% if we excluded HER2-positive patients who did not undergo anti-HER2 therapies at the onset of BM. This is somewhat disappointing because the authors investigated all possible known

Tabouret et al: Prognostic score for brain metastases

Acknowledgments ETparticipated in study design, acquired and analyzed data, and drafted the manuscript. AG, PM, EC, PV, and ATparticipated in the acquisition of data. BE participated in the design of the study and performed the statistical analysis. ATconceived of the study, participated in its design and analyses of data, and coordinated and helped draft the manuscript. All authors read and approved the final manuscript.

Conflict of interest statement. P. Metellus: honoria from Roche; A Goncalves: honoria from Roche, GSK, and Astra Zeneca; A. Tallet: honoria from Roche. All other authors declare no conflict of interest.

15. Claude L, Perol D, Ray-Coquard I, et al. Lymphopenia: a new independent prognostic factor for survival in patients treated with whole brain radiotherapy for brain metastases from breast carcinoma. Radiother Oncol. 2005;76:334– 339. 16. Golden DW, Lamborn KR, McDermott MW, et al. Prognostic factors and grading systems for overall survival in patients treated with radiosurgery for brain metastases: variation by primary site. J Neurosurg. 2008;109(Suppl):77–86. 17. R Core Team 2013. R Foundation for Statistical Computing, Vienna, Austria. 18. Nam BH, Kim SY, Han HS, et al. Breast cancer subtypes and survival in patients with brain metastases. Breast Cancer Res. 2008;10:R20.

1.

Jemal A, Bray F, Center MM, Ferlay J, Ward E and Forman D. Global cancer statistics. CA Cancer J Clin. 2011;61:69–90.

19. Dawood S, Broglio K, Esteva FJ, et al. Defining prognosis for women with breast cancer and CNS metastases by HER2 status. Ann Oncol. 2008; 19:1242–1248.

2.

Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2. N Engl J Med. 2001;344:783–792.

20. Kirsch DG, Ledezma CJ, Mathews CS, et al. Survival after brain metastases from breast cancer in the trastuzumab era. J Clin Oncol. 2005;23:2114 –2116. author reply 2116 –2117.

3.

Tabouret E, Chinot O, Metellus P, Tallet A, Viens P and Goncalves A. Recent trends in epidemiology of brain metastases: an overview. Anticancer Res. 2012;32:4655– 4662.

4.

Barnholtz-Sloan JS, Sloan AE, Davis FG, Vigneau FD, Lai P and Sawaya RE. Incidence proportions of brain metastases in patients diagnosed (1973 to 2001) in the Metropolitan Detroit Cancer Surveillance System. J Clin Oncol. 2004;22:2865– 2872.

21. Dawood S, Gonzalez-Angulo AM, Albarracin C, et al. Prognostic factors of survival in the trastuzumab era among women with breast cancer and brain metastases who receive whole brain radiotherapy: a single-institution review. Cancer. 2010;116:3084 –3092.

5.

Smedby KE, Brandt L, Backlund ML and Blomqvist P. Brain metastases admissions in Sweden between 1987 and 2006. Br J Cancer. 2009; 101:1919– 1924.

22. Rades D, Lohynska R, Veninga T, Stalpers LJ and Schild SE. Evaluation of 2 whole-brain radiotherapy schedules and prognostic factors for brain metastases in breast cancer patients. Cancer. 2007;110: 2587 –2592. 23. Sperduto PW, Kased N, Roberge D, et al. Effect of tumor subtype on survival and the graded prognostic assessment for patients with breast cancer and brain metastases. Int J Radiat Oncol Biol Phys. 2011;82:2111 –2117.

6.

Patchell RA, Tibbs PA, Walsh JW, et al. A randomized trial of surgery in the treatment of single metastases to the brain. N Engl J Med. 1990; 322:494–500.

7.

Niwinska A, Murawska M and Pogoda K. Breast cancer subtypes and response to systemic treatment after whole-brain radiotherapy in patients with brain metastases. Cancer. 2010;116:4238– 4247.

24. Sperduto PW, Kased N, Roberge D, et al. The effect of tumor subtype on the time from primary diagnosis to development of brain metastases and survival in patients with breast cancer. J Neurooncol. 2013; 112(3):467 –472.

8.

Aoyama H, Shirato H, Tago M, et al. Stereotactic radiosurgery plus whole-brain radiation therapy vs stereotactic radiosurgery alone for treatment of brain metastases: a randomized controlled trial. Jama. 2006;295:2483– 2491.

25. Dawood S, Lei X, Litton JK, Buchholz TA, Hortobagyi GN and Gonzalez-Angulo AM. Incidence of brain metastases as a first site of recurrence among women with triple receptor-negative breast cancer. Cancer. 2012;118:4652–4659.

9.

Lin NU, Carey LA, Liu MC, et al. Phase II trial of lapatinib for brain metastases in patients with human epidermal growth factor receptor 2-positive breast cancer. J Clin Oncol. 2008;26:1993– 1999.

26. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci USA. 2001;98:10869–10874.

10. Sperduto PW, Kased N, Roberge D, 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. 2012;30:419 –425.

27. Braccini AL, David A, Thezenas S, Gilles R, Jean-Marc F and William J. Comparative performances of prognostic indexes for breast cancer patients presenting with brain metastases. BMC Cancer. 2013;13:70.

11. Niwinska A and Murawska M. New breast cancer recursive partitioning analysis prognostic index in patients with newly diagnosed brain metastases. Int J Radiat Oncol Biol Phys. 2012;82:2065–2071. 12. Nieder C, Marienhagen K, Astner ST and Molls M. Prognostic scores in brain metastases from breast cancer. BMC Cancer. 2009;9:105. 13. Park BB, Uhm JE, Cho EY, et al. Prognostic factor analysis in patients with brain metastasesfrom breast cancer: how can we improve the treatment outcomes?. Cancer Chemother Pharmacol. 2009;63:627–633. 14. Le Scodan R, Massard C, Mouret-Fourme E, et al. Brain metastases from breast carcinoma: validation of the radiation therapy oncology group

8 of 8

28. Gaspar L, Scott C, Rotman M, et al. Recursive partitioning analysis (RPA) of prognostic factors in three Radiation Therapy Oncology Group (RTOG) brain metastases trials. Int J Radiat Oncol Biol Phys. 1997;37: 745– 751. 29. Brown PD, Buckner JC, O’Fallon JR, et al. Importance of baseline mini-mental state examination as a prognostic factor for patients with low grade gliomas. Int J RadiatOncolBiol Phys. 2004;59:117 –125. 30. Meyers CA, Smith JA, Bezjak A, et al. Neurocognitive function and progression in patients with brain metastases treated with whole brain radiation and motexafin gadolinium: Results of a randomized phase III trial. J ClinOncol. 2004;22:157– 165.

Neuro-Oncology

Downloaded from http://neuro-oncology.oxfordjournals.org/ at Aston University on January 26, 2014

References

recursive partitioning analysis classification and proposition of a new prognostic score. Int J Radiat Oncol Biol Phys. 2007;69:839– 845.