Int. J. Radiation Oncology Biol. Phys., Vol. 43, No. 4, pp.

795– 803, 1999

Copyright © 1999 Elsevier Science Inc.
Printed in the USA. All rights reserved
0360-3016/99/$–see front matter

PII S0360-3016(98)00442-8

CLINICAL INVESTIGATION Brain

IDENTIFICATION OF PROGNOSTIC FACTORS IN PATIENTS WITH BRAIN

METASTASES: A REVIEW OF 1292 PATIENTS

FRANK J. LAGERWAARD, M.D.,* PETER C. LEVENDAG, M.D., PH.D.,*

PETER J. C. M. NOWAK, M.D., PH.D.,* WILHELMINA M. H. EIJKENBOOM, M.D., PH.D.,*
PATRICK E. J. HANSSENS, M.D.,* AND PAUL I. M. SCHMITZ, PH.D.†
Departments of *Radiation Oncology and †Trials and Statistics, Daniel den Hoed Cancer Center, University Hospital Rotterdam,
Rotterdam, The Netherlands

Purpose: Prognostic factors in 1292 patients with brain metastases, treated in a single institution were identified
in order to determine subgroups of patients suitable for selection in future trials.
Materials and Methods: From January 1981 through December 1990, 1292 patients with CT-diagnosed brain
metastases were referred to the Department of Radiation Oncology, Daniel den Hoed Cancer Center, Rotterdam.
The majority of patients were treated with whole brain radiotherapy (84%), the remainder were treated with
steroids only or surgery and radiotherapy. Information on potential prognostic factors (age, sex, performance
status, number and distribution of brain metastases, site of primary tumor, histology, interval between primary
tumor and brain metastases, systemic tumor activity, serum lactate dehydrogenase, response to steroid treat-
ment, and treatment modality) was collected. Univariate and multivariate analyses were performed to determine
significant prognostic factors. Results were compared with literature findings using a review of prognostic factors
in 18 published reports.
Results: Overall median survival was 3.4 months, with 6-month, 1-year, and 2-year survival percentages of 36%,
12%, and 4% respectively. Survival was statistically significantly different between treatment modalities, with
median survival of 1.3 months in patients treated with steroids only, 3.6 months in patients treated with
radiotherapy, and 8.9 months in patients treated with neurosurgery followed by radiotherapy (p < 0.0001).
Multivariate analysis confirmed literature findings of the major prognostic value of treatment modality on
survival of patients with brain metastases. Performance status, response to steroid treatment, systemic tumor
activity, and serum lactate dehydrogenase were independent prognostic factors with the strongest impact on
survival, second only to treatment modality. Site of primary tumor, age, and number of brain metastases were
also identified as prognostic factors in our material, although with lesser importance. In patients with lung
primaries, sex was found to have significant impact on survival. In patients with breast primaries, interval
between primary tumor and development of brain metastases appeared to be a statistically significant prognostic
factor. Histology in patients with lung primaries and distribution of brain metastases were not found to be
statistically significant in multivariate analysis.
Conclusions: In this large database, the value of established prognostic factors was confirmed and, furthermore,
some less well-recognized parameters such as response to steroid treatment, serum lactate dehydrogenase, age,
sex in lung primaries, and site of primary tumor were established. From the three strongest prognostic
factors—performance status, response to steroids, and evidence of systemic disease—simple identification of
favorable and unfavorable subgroups of patients with brain metastases can be constructed. © 1999 Elsevier
Science Inc.

Brain metastases, Prognostic factors, Radiotherapy.

INTRODUCTION only for a limited period of time. Whole brain irradiation

Appropriate management of patients with brain metastases
from solid tumors has been a challenge for decades. The
development of intracerebral metastases is associated with
considerable morbidity. The natural course without treat-
ment is that of progressive neurological deterioration with
median survival of 1–2 months (1–3). Treatment with ste-
roids can alleviate symptoms in the majority of patients, but
has been the mainstay of treatment for many years, with
response rates in 50 –75% of patients and median survival of
3– 6 months (4 –10). Only for a small subset of patients with
single brain metastasis more aggressive treatment by means
of surgery followed by radiotherapy seems the appropriate
type of treatment. Since up to 50% of patients treated with
conventional fractionation schemes die as a result of intra-

Reprint requests to: Frank J. Lagerwaard, Daniel den Hoed The Netherlands. E-mail: lagerwaard@rtdh.AZR.NL
Cancer Center, University Hospital Rotterdam, Department of Accepted for publication 9 October 1998.
Radiation Oncology, Groene Hiledijk 301, 3075 EA Rotterdam,

795
796 I. J. Radiation Oncology ● Biology
cranial progression (10, 11), there have been several at-
tempts to define a favorable subgroup of patients who might
benefit from higher doses of fractionated radiotherapy.
However, most of these investigations failed to demonstrate
a beneficial effect of higher radiation dose on survival (4,
11–13). Stereotactic radiosurgery seems to have improved
the prognosis for selected patients with brain metastases
(14 –16). Furthermore, in recent years there has been in-
creasing interest in the role of chemotherapy, in particular
for patients with breast and small cell lung primaries. This
evolution in available treatment modalities further stresses
the necessity to identify prognostic factors, with implica-
tions for individual treatment selection as well as for defi-
nition of inclusion criteria for future trials. In the past,
several studies on prognostic factors in patients with brain
metastases have been reported, with partially contradictory
results.
Because these investigations are based on selected patient
groups, treated with varying treatment modalities, definite
conclusions on prognostic significant pretreatment factors in
patients with brain metastases are difficult to assess. The
primary goal of this study was to define independent prog-
nostic pretreatment factors in patients with brain metastases
from solid tumors. We conducted a retrospective study
within a large database of patients with brain metastases,
treated in our institution in a 10-year period. We also
performed a review of the literature, with respect to prog-
nostic factors and compared these results with findings from
our database.
METHODS AND MATERIALS
Retrospective analysis
A retrospective analysis of the medical records of the
Department of Radiation Oncology of the University Hos-
pital Rotterdam, Daniel Den Hoed Cancer Center, identified
1333 patients with documented brain metastases from solid
tumors, diagnosed between January 1981 and December
1990. Only patients with computed tomography (CT) diag-
nosis of brain metastases were analyzed, i.e., 35 patients
with mere clinical suspicion or isotope scanning of the brain
were excluded. In six patients insufficient follow-up data
were available, thus leaving a total of 1292 patients for
evaluation. For each patient, age, sex, performance score,
number and distribution of brain metastases, site and status
of primary tumor, metastatic extent, treatment parameters,
and response to treatment were recorded. Patients were
considered having a primary tumor of unknown origin if
within 2 months of diagnosis of brain metastases no primary
tumor had been determined on routine workup. For analysis
of prognostic factors, three different age cohorts were eval-
uated: 59 years and younger, 60 through 69 years, and 70
years and older. Retrospectively, based on the medical
records, the performance status at diagnosis of brain metas-
tases and at regular intervals after treatment using the East-
ern Cooperative Oncology Group (ECOG) grades (17) was
defined. Systemic tumor activity was deduced from two
● Physics Volume 43, Number 4, 1999
separately coded variables, metastatic extent and primary
tumor status. As a general rule, no extensive work-up for
determination of systemic tumor activity was performed,
unless neurosurgery was contemplated. Patients were con-
sidered without evidence of active systemic disease, if,
based on the medical record, there was no evidence of
metastases outside the brain and the primary tumor was
absent. Systemic tumor activity was considered “limited” if
there were known systemic metastases, with primary tumor
absent or controlled or if there were no systemic metastases,
but progression of primary tumor. In cases of progressive
primary tumor growth and systemic metastases, systemic
tumor activity was encoded as “extensive.”
Response to treatment with steroids was judged retro-
spectively as “good” (marked improvement of neurological
symptoms), “moderate” (some improvement), or “little”
(little or no improvement) from the medical records. With
respect to distribution of brain metastases, patients were
grouped as having supratentorial or infratentorial only, su-
pratentorial and infratentorial brain metastases, and combi-
nations including brain stem. Patients were stratified by
level of serum lactate dehydrogenase (LDH) at time of
diagnosis of their brain metastases. Analysis of this factor
was complicated by the fact that LDH determinations were
conducted at different hospitals with different upper limits
of normal. As upper limit of normal LDH a value of 350
U/L was established, thus having 728 patients with “nor-
mal” (i.e., #350 U/L) and 200 patients with “elevated”
LDH (i.e., .350 U/L). For analysis of prognostic value of
interval between primary tumor and brain metastases, pa-
tients were divided into two groups: those with synchronous
occurrence of brain metastases and those with metachro-
nous development of brain metastases. Patients with prima-
ries of unknown origin were excluded from this analysis.
The survival time was measured from the date of the diag-
nostic CT scan; clinical response was measured 6 weeks
after treatment since most patients had their first follow-up
visit at this time. Follow-up CT scans were not routinely
obtained and were available in only 288 patients (22%).
Statistical methods
Overall survival from time of diagnosis of brain metas-
tases was calculated with the Kaplan-Meier method. Uni-
variate analyses were performed using the Kaplan-Meier
method (18) and log rank tests. For multivariate survival
analyses, Cox’s proportional hazards model (19, 20) was
used.
For a final multivariate analysis only 562 (43%) of these
patients had complete observations for all factors. There-
fore, we used an imputation method to estimate missing
values, based on separate multivariate regression methods
with one prognostic factor as dependent variable and all
other factors as independent variables. Because multivariate
results with and without imputation proved comparable,
multivariate results are presented only for the total dataset
of 1292 patients with missing values replaced by imputa-
tion.
 Prognostic factors in patients with brain metastases ● F. J. LAGERWAARD et al. 797

Literature review Table 1. Patient characteristics (n 5 1292)

Theoretically a formally designed meta-analysis of the
Sex
reported literature would give the best insight into the Male 810 (63%)
prognostic value of the various suggested factors; however, Female 482 (37%)
this analysis is time-consuming and complex because stud- Age
ies have major differences with regard to treatment modal- 17–59 623 (48%)
ity, primary tumor under investigation, and inclusion crite- 60–69 425 (33%)
.70 244 (19%)
ria of patients. As an alternative, we pursued a simplified Site of primary
approach in order to obtain a global impression of the Lung 721 (56%)
consistent significance of prognostic factors in brain metas- Breast 213 (16%)
tases. A review of the literature on prognostic factors in Renal 48 (4%)
brain metastases from 1980 to 1996 was performed, reveal- Unknown 102 (8%)
Other 208 (16%)
ing 18 reports with multivariate analyses on prognostic Performance status (ECOG)
parameters. Information was gathered on the same factors as 0 131 (10%)
used in our retrospective study. 1 503 (39%)
For all variables we calculated a weighted prognostic 2 504 (39%)
score by dividing the total sum of patients in studies in 3 68 (5%)
Unknown 86 (7%)
which the investigated factor was significant (p , 0.05) by Systemic tumor activity
the total sum of patients in all studies in which the factor None 171 (13%)
was investigated. These scores range from 0 through 1, with Limited 701 (54%)
higher scores to be interpreted as a higher consistency of Extensive 302 (23%)
being statistically significant. Unknown 118 (9%)
Number of brain metastases
1–2 843 (65%)
RESULTS $3 429 (33%)
Unknown 20 (2%)
Retrospective analysis
Patient characteristics. A total of 1292 patients with
brain metastases were evaluated; the median age of the more often in patients with lung and breast primaries than in
patient cohort at diagnosis of brain metastases was 59 years patients with melanoma and renal cell carcinoma. Metasta-
(range: 17– 84 years). Eight hundred and ten patients were ses were distributed evenly between the left and right hemi-
male and 482 were female (Table 1). spheres.
The majority of patients had lung or breast primaries, Treatment. All patients were treated with high doses of
followed by primaries of unknown origin and malignant steroids initially, usually dexamethasone (dose range 4 –16
melanoma. The median interval between diagnosis of pri- mg/day, mean 14.6 mg/day), which were tapered slowly and
mary tumor and brain metastases was 8.5 months and varied discontinued in the weeks after treatment. Response to
widely with the site of primary tumor, i.e., 4 months in lung steroid treatment was recorded to be present when there was
cancer, 29 months in breast cancer primaries, and 37 months substantial improvement in neurological condition. Patients
in malignant melanoma. Presenting signs and symptoms with seizures at presentation were treated with anticonvul-
were headache (53%), motor weakness (38%), cerebellar sant drugs. As shown in Table 1, the majority of patients
dysfunction (26%), seizures (22%), and dysphasia (14%). were treated with steroids and whole brain radiotherapy,
Two hundred eleven patients (16%) presented with signs using lateral opposed fields with a 4 MV or 6 MV linear
and symptoms from brain metastases without prior history accelerator. Most commonly used fractionation schedules
of malignancy; 93% of these patients were found to have (76%) were 30 Gy in 10 fractions or 20 Gy in 5 fractions
bronchogenic primary tumor. Thirty-one patients (2%), al- (dose range 8 –56 Gy, mean 31.5 Gy; SD 5 6.9 Gy). One
most exclusively patients with malignant melanoma and hundred and forty-eight patients with single brain metastasis
renal cell carcinoma, had asymptomatic brain metastases received an additional boost dose (range 5–20 Gy; mean
detected on routine CT scanning performed in prechemo- 11.5 Gy; SD 5 3.4 Gy) to a limited field. In this series
therapeutic screening procedures. Almost 40% of patients 95 patients with single brain metastasis were treated with
with brain metastases from malignant melanoma had sei- surgery followed by radiotherapy (dose range 20 –55 Gy;
zures at the time of diagnosis of brain metastases. Five mean 38.7 Gy; SD 5 8.5 Gy), started within 6 weeks after
hundred and eighty-one patients had a single brain metas- operation. One hundred and eighteen patients (9.1%) were
tasis on diagnostic CT scan (46%); 691 patients had multi- treated with steroids only after diagnosis of brain metasta-
ple lesions. In 20 patients (2%) the number of brain metas- ses. Reasons for withholding additional radiotherapy in
tases could not be established retrospectively. Brain these patients were poor performance status at diagnosis,
metastases were located supratentorial in 786 patients patient refusal, and prior prophylactic cranial irradiation.
(68%), infratentorial in 136 patients (12%), with combina- Survival. Overall survival of the 1292 patients with brain
tions in the remainder. Cerebellar metastases were found metastases was 12% at 1 year and 4% at 2 years, with a
798 I. J. Radiation Oncology ● Biology
Fig. 1. Overall survival (n 5 1292).
median survival time of 3.4 months (Fig. 1). To establish the
influence of treatment periods, patients were divided in
1-year cohorts according to year of treatment. No difference
in survival could be demonstrated between these groups
(p 5 0.67). Cause of death was known for only 807 patients;
57% of these succumbed because of intracranial progression
of disease.
Prognostic factors
Univariate analysis. First, univariate analysis was per-
formed on 12 different variables to evaluate their potential
prognostic value with respect to overall survival. A number
of patient characteristics (age, sex in lung primaries and
performance status at diagnosis of brain metastases), tumor
characteristics (number and distribution of brain metastases,
site of primary tumor, histology in lung primaries, interval
between primary tumor and brain metastases in breast pri-
maries, systemic tumor activity, serum LDH), and treatment
characteristics (response to steroids, treatment modality)
were investigated. The results are summarized in Table 2
and for the four strongest factors shown in Fig. 2a– d.
Analysis within primary tumor groups showed interval
between primary tumor and brain metastases, measured at 2
years, to be a prognostic factor in patients with breast
primaries (p 5 0.03); in pulmonary and renal primaries no
statistical significance could be found according to interval
(p 5 0.37 and p 5 0.80 respectively). Primary tumors other
than breast, renal, and unknown primaries showed no sig-
nificant difference with respect to survival, tested against
lung primaries. All other univariately tested variables were
significant.
Multivariate analysis. All 12 factors tested by univariate
analysis, were also analyzed by the multivariate propor-
tional hazards model after using multivariate regression for
imputation of missing values (see statistical methods). The
results are shown in Table 3. For the calculation of hazard
ratios and their confidence intervals and statistical signifi-
cance we actually used three different proportional hazard
models.
● Physics Volume 43, Number 4, 1999
The first model was applied for all patients (all sites), the
second for patients with lung primaries only, and finally a
separate model for patients with breast primaries. Unless
indicated in Table 3 (sex, histology, interval) the results are
shown for the model including all sites.
Prognostic subgroups
From Table 3 it can be seen that the significant hazard
ratios (hr) with largest values (i.e., hr . 1.5 or , 0.66) were
found for treatment modalities, performance status, re-
sponse to steroids, and systemic tumor activity. Within the
subgroup of patients treated with radiotherapy only, three
prognostic subgroups could be constructed:
“good prognosis”: ECOG 0 or 1 and
no or limited systemic tumor activity and
good response to steroids
“poor prognosis”: ECOG 2 or 3 and
systemic tumor activity limited or extensive and
little response to steroids
“moderate prognosis”: all other patients treated with radio-
therapy.
Survival curves for the prognostic groups are shown in
Fig. 3. The median survival for these groups are 6.3 months
(“good prognosis,” n 5 243), 3.4 months (“moderate prog-
nosis,” n 5 735), and 1.3 months (“poor prognosis,” n 5
101).
Results of literature review
Figure 4 shows the evaluation of 16 factors in 18 studies
using a multivariate approach for assessment of the influ-
ence on survival. The total number of patients in these
studies is 3013. High consistency scores of the significance
of each factor were defined as values on the y-axis of 0.75
or more. Treatment modality, primary tumor status, sys-
temic metastatic extent, performance status, and serum
LDH appear to be the most consistent significant factors.
Because of the relatively low number of patients (x-axis)
LDH will be dropped (it is investigated in only one report).
It is seen too that age is of uncertain prognostic value
(score 5 0.55). From this approach it can be concluded that
apart from treatment and site of primary tumor, performance
status and systemic tumor activity appear as the most con-
sistent statistically significant factors in the literature.
DISCUSSION
For many years management of brain metastases has been
confined to treatment with high doses of steroids, combined
with fractionated whole brain radiotherapy, with only a
small percentage of patients eligible for surgical resection of
a single brain metastasis. The advent of stereotactic radio-
therapy and the implementation of chemotherapeutic strat-
egies in the treatment of brain metastases from breast and
small cell lung primaries has broadened the therapeutic
spectrum available for oncologists. Choice of treatment for
 Prognostic factors in patients with brain metastases ● F. J. LAGERWAARD et al. 799

Table 2. Univariate results

Median
survival 6-month 1-year 2-year
Patients (months) survival (%) survival (%) survival (%) p-Value†

Overall (n 5 1292) 1292 3.4 36 12 4 —

Patient characteristics
Age (n 5 1292)
17–59 623 3.9 38 15 5 p , 0.0001
60–69 425 3.3 30 12 4
.70 244 2.6 20 5 0.4
Sex (lung) (n 5 721)*
Female 120 4.2 38 13 4 p 5 0.006
Male 601 3.1 26 7 2
Performance status (ECOG) (n 5 1206)
0 131 7.8 57 32 12 p , 0.0001
1 503 4.5 39 15 5
2 504 2.3 22 7 2
3 68 1.4 10 3 3

Tumor characteristics
Number of brain metastases (n 5 1272)
1–2 843 4.0 38 16 6 p , 0.0001
$3 429 2.7 21 5 1
Distribution (n 5 1152)
Cerebrum or cerebellum 922 3.9 37 15 5 p , 0.0001
Both or other combinations 230 2.8 24 6 1
Site of primary (n 5 1292)
Lung 721 3.2 28 8 2
Breast 213 4.5 39 20 7 p , 0.0001 ‡
Renal 48 4.3 42 23 10 p 5 0.003
Unknown 102 5.5 48 22 10 p , 0.0001
Other 208 2.8 28 10 2 p 5 0.21
Histology (lung) (n 5 721)
Other than adenocarcinoma 585 3.0 25 6 1 p , 0.0001
Adenocarcinoma 136 4.1 40 18 7
Interval (breast) (n 5 213)
,2 years 90 4.0 34 17 2 p 5 0.03
$2 years 123 4.9 43 22 10
Systemic tumor activity (n 5 1174)
None 171 6.6 55 26 10 p , 0.0001
Limited 701 3.4 32 10 3
Extensive 302 2.4 16 6 1
Serum LDH (n 5 928)
Normal 728 4.0 37 14 4 p , 0.0001
Elevated 200 2.2 16 6 1
Treatment characteristics
Response to steroids (n 5 934)
Good 647 4.3 38 14 5 p , 0.0001
Moderate 83 2.2 23 10 1
Little 204 1.6 9 2 1
Treatment modality (n 5 1292)
Steroids 118 1.3 4 2 1 p , 0.0001
Radiotherapy 1079 3.6 32 11 3
Surgery and radiotherapy 95 8.9 66 39 18

* Totals below 1292 indicate missing values and/or subgroups.

†
Overall log-rank test, except for site of primary.
‡
Each category tested against lung.

an individual patient with brain metastases is based on a
number of factors: number and localization of brain metas-
tases, systemic tumor activity, performance score, and age
are major determinants for selection of treatment modality.
Future trials in patients with brain metastases depend on
selection of patients with favorable prognosis to allow ad-
equate long-term follow-up to draw conclusions about sur-
vival and late toxicity, further stressing the importance of
prognostic parameters.
Several studies on prognostic factors have been published
800 I. J. Radiation Oncology ● Biology
Fig. 2. Kaplan-Meier survival curves for prognostic factors. (a) Treatment modality. (b) Performance status at diagnosis
(ECOG 0 –3). (c) Systemic tumor burden. (d) Response to steroid treatment.
in the last two decades, partially with conflicting results.
Simplified meta-analysis of multivariate investigations, as
described earlier, showed treatment modality, performance
status, primary tumor status, and systemic metastatic extent
to be the most important factors for survival of patients with
brain metastases, while the importance of age and serum
LDH remained uncertain.
Various prospective and retrospective studies have shown
the superiority of the combination of surgery and radiother-
apy for selected patients with brain metastases (5, 8, 10, 21,
22). In our series, although highly biased on other factors,
patients treated with resection of a single brain metastasis,
followed by radiotherapy had median survival of 8.8
months, compared to 3.8 months for patients treated with
radiotherapy only, and 1.2 months for patients treated with
steroids only.
With multivariate analysis treatment modality proved to
have the highest impact on survival, as shown by vastly
differing hazard rates for different treatment regimes.
Performance status has been shown by various authors to
be of prognostic significance in patients with primary and
secondary brain tumors (4, 9, 10, 23). In our series perfor-
mance status was the major determinant of survival, sec-
ondary only to treatment regime. Survival improved from 6
weeks in poor performance patients through almost 8
months in patients with best performance scores.
In agreement with literature findings, (4, 8, 21, 23) the
● Physics Volume 43, Number 4, 1999
extent of systemic tumor activity was significantly corre-
lated with survival; in particular patients without systemic
tumor progression, although only a small proportion of all
patients, showed superior survival with a median of almost
7 months.
Serum LDH has been clearly established as an important
prognostic factor in patients with lung carcinoma and ma-
lignant melanoma. Chatani et al. in 1989 (24, 25) first
described the prognostic value of LDH in patients with
brain metastases. In our investigation elevation of LDH
proved to be an independent significant unfavorable prog-
nostic factor, as well in univariate as in multivariate analy-
sis. LDH levels are probably correlated to tumor burden in
patients with progressive systemic disease; elevation of
serum LDH without known progressive disease may be a
reflection of occult systemic disease. The independent sig-
nificance of serum LDH in this study, even after correction
for evidence of systemic disease, may be explained in this
fashion.
Several authors (4, 11) have reported improved survival
in patients with brain metastases from breast carcinoma,
compared to other primaries. In our study, this finding could
be confirmed and furthermore, within the subgroup of pa-
tients with brain metastases from breast carcinoma, interval
between primary tumor and brain metastases proved to be of
significant importance.
Response to overall treatment has been linked to survival
 Prognostic factors in patients with brain metastases ● F. J. LAGERWAARD et al. 801

Table 3. Multivariate results

Hazard 95% Confidence

Patients ratio limits hazard ratio p-Value

Patient characteristics
Age (n 5 1292)
17–59 623 1 — —
60–69 425 1.06 0.94–1.21 p 5 0.34
$70 244 1.40 1.20–1.61 p , 0.001
Sex (lung) (n 5 721)
Female 120 1 —
Male 601 1.36 1.11–1.68 p 5 0.004
Performance status (ECOG) (n 5 1292)
0 174 1 — —
1 527 1.24 1.04–1.49 p 5 0.02
2 523 1.74 1.44–2.09 p , 0.001
3 68 2.53 1.88–3.42 p , 0.001
Tumor characteristics
Number of brain metastases (n 5 1292)
1–2 859 1 —
$3 433 1.33 1.17–1.50 p , 0.001
Distribution (n 5 1292)
Cerebrum or cerebellum 1062 1 — —
Both or other combinations 230 1.10 0.94–1.29 p 5 0.22
Site of primary (n 5 1292)
Lung 721 1 — —
Breast 213 0.70 0.60–0.82 p ,0.001
All others 358 0.93 0.81–1.06 p 5 0.27
Histology (lung) (n 5 721)
Other than adenocarcinoma 585 1 —
Adenocarcinoma 136 0.88 0.71–1.08 p 5 0.21
Interval (breast) (n 5 213)
,2 years 90 1 —
$2 years 123 0.65 0.48–0.87 p 5 0.004
Systemic tumor activity (n 5 1292)
None 190 1 — —
Limited 799 1.35 1.13–1.60 p 5 0.001
Extensive 303 1.60 1.31–1.96 p , 0.001
Serum LDH (n 5 1292)
Normal 1081 1 — —
Elevated 211 1.55 1.32–1.81 p , 0.001
Treatment characteristics
Response to steroids (n 5 1292)
Good 986 1 — —
Moderate 83 1.43 1.14–1.80 p 5 0.002
Little 223 1.74 1.48–2.04 p , 0.001
Treatment modality (n 5 1292)
Steroids 118 1 — —
Radiotherapy 1079 0.38 0.31–0.46 p , 0.001
Surgery and radiotherapy 95 0.21 0.15–0.28 p , 0.001

of patients with brain metastases by different authors (8,
26). Little is known, however, about the importance of the
response to treatment with steroids on survival of patients
with brain metastases. In daily clinical practice, patients in
poor neurological or general condition not responding to
adequate doses of steroids frequently are not referred for
further treatment. In 934 patients, in whom response to
steroid treatment could be determined retrospectively, we
found significantly poorer survival of 1.6 months in patients
who did not show substantial neurological improvement to
treatment with steroids, compared to 4.3 months in patients
who did. This effect was confirmed in multivariate analysis
and was not due to treatment with inadequate doses of
dexamethasone, since median dose of dexamethasone was
similar in both groups. Perhaps even more importantly, in
addition to poor survival, response rates in patients without
improvement after steroid treatment were much lower; only
42% of these patients experienced neurological improve-
ment after subsequent radiotherapy. Failure to respond to
dexamethasone treatment is not obligatorily linked to the
amount of edema surrounding the brain metastases. In our
retrospective study, with only limited information on diag-
nostic CT scans, only a minority of patients not responding
to steroid treatment showed absent or limited surrounding
802 I. J. Radiation Oncology ● Biology
Fig. 3. Kaplan-Meier survival by prognostic group (good, moder-
ate, poor prognosis).
edema on diagnostic CT scan. Based on these findings there
is no evidence for changing the current clinical practice of
withholding further treatment in patients in poor condition,
not responding to steroid treatment. Further prospective
studies on the influence of response to steroids may be
warranted.
Several authors (27–29) have shown the superiority of
gadolinium–diethylenetriaminepentaacetic acid (Gd-DTPA)
enhanced magnetic resonance imaging (MRI) over CT-based
diagnosis of brain metastases, particularly in the detection of
smaller metastases and infratentorial lesions. The implementa-
tion of MRI as “gold standard” in diagnosis of brain metastases
has considerable implications on treatment selection and prog-
nosis of patients with cerebral metastases. In this series, a
number of brain metastases were significantly correlated
with survival. Since diagnosis of brain metastases in our
study was made in the CT era, theoretically the prognostic
value of the number of brain metastases when based on MRI
findings may have been underscored. The distribution of
brain metastases was not of significant importance in our
material.
In our study, patients aged 70 years and older fared less
well than younger patients, even after correction for treat-
ment differences, an observation described earlier by Nieder
et al. (30). This observation is of special importance since
the subset of patients 70 years and older represents a rela-
tively large proportion of patients with brain metastases,
i.e., 19% in our material.
Multivariate analysis furthermore showed female sex in
patients with brain metastases from lung primaries to be a
favorable significant prognostic factors. Histology failed to
show a correlation with survival in multivariate analysis of
the subset of patients with lung primaries.
From the most important prognostic factors, as judged by
their hazard rates, three prognostic subgroups could be
determined within patients treated with whole brain radio-
therapy. These prognostic groups, derived from the current
study, are comparable to the results of a recently reported
recursive partitioning analysis of prognostic factors in a
database of 1200 patients from three consecutive trials of
● Physics Volume 43, Number 4, 1999
Fig. 4. Consistency score derived from literature findings using
multivariate analyses.
the Radiation Therapy Oncology Group (RTOG) (31). De-
spite differences in pretreatment characteristics, with a more
favorable group of patients in the RTOG database, both
studies identify performance status and systemic tumor ac-
tivity as the most important prognostic factors in patients
with brain metastases, while both studies confirm the neg-
ative effect of higher age. Although in most studies perfor-
mance status is determined directly prior to radiotherapy,
we chose a more basic approach by investigating perfor-
mance status at diagnosis of brain metastases and response
to steroid treatment separately. Using this approach, a
“good” prognosis group consisting of patients with good
performance status (ECOG 0 or 1), limited or absent sys-
temic tumor activity, and substantial neurological improve-
ment after treatment with steroids could be identified with a
median survival of 6 months. This relatively good prognosis
group, suitable for future trials in patients with brain me-
tastases, represents about one-quarter of all patients with
brain metastases in this series. A “poor” prognosis group
can be described as patients with poor performance status
(ECOG 2 or 3), little or no response to steroids, and limited
or extensive systemic tumor activity. These patients, about
10% of patients in our series, have median survival of only
1.3 months after palliative radiotherapy. A conservative
approach in the treatment of this group of patients with
brain metastases seems justified.
CONCLUSION
The major prognostic significance of treatment modality,
performance status, and systemic tumor activity, derived
from literature findings, could be confirmed in a large da-
tabase of patients with brain metastases. Multivariate anal-
ysis also showed response to treatment with steroids to be a
significant prognostic factor. With these variables, simple
identification of prognostic favorable and unfavorable sub-
groups of patients with brain metastases can be achieved.
 Prognostic factors in patients with brain metastases
REFERENCES
1. Hazra T, Mullins GN, Lott S. Management of cerebral metas-
tases from bronchogenic carcinoma. John Hopkins Med J
1972;130:377–383.
2. Markesbery WR, Brooks WH, Gupta GD, et al. Treatment for
patients with cerebral metastases. Arch Neurol 1978;35:754 –
756.
3. Posner JB. Management of central nervous system metastasis.
Sem Oncol 1977;4:81–91.
4. Borgelt B, Gelber R, Kramer S, et al. The palliation of brain
metastases: Final results of the first two studies by the Radi-
ation Therapy Oncology Group. Int J Radiat Oncol Biol Phys
1980;6:1– 8.
5. Egawa S, Tukiyama I, Akine Y, et al. Radiotherapy of brain
metastases. Int J Radiat Oncol Biol Phys 1986;12:1621–1625.
6. Komarnicky LT, Phillips TL, Martz, K, et al. Randomized
phase III protocol for the evaluation of misonidazole com-
bined with radiation in the treatment of patients with brain
metastases (RTOG-7916). Int J Radiat Oncol Biol Phys 1991;
20:53–58.
7. Nieder C, Niewald M, Schnabel K. The radiotherapy of brain
metastases in bronchial carcinoma. Strahlenther Onkol 1994;
170:335–341.
8. Pladdet I, Boven E, Nauta J, et al. Palliative care for brain
metastases of solid tumour types. Neth J Med 1989;34:10 –21.
9. Ryan GF, Ball DL, Smith JG. Treatment of brain metastases
from primary lung cancer. Int J Radiat Oncol Biol Phys
1994;31:273–278.
10. Zimm S, Wampler GL, Stablein D, et al. Intracerebral metas-
tases in solid-tumor patients: Natural history and results of
treatment. Cancer 1981;15:384 –395.
11. Kurtz JM, Gelber R, Brady LW, et al. The palliation of brain
metastases in a favorable patient population: A randomized
clinical trial by the Radiation Therapy Oncology Group. Int J
Radiat Oncol Biol Phys 1981;7:891– 895.
12. Haie-Meder C, Pellae-Cosset B, Laplanche A, et al. Results of
a randomized clinical trial comparing two radiation schedules
in the palliative treatment of brain metastases. Radiother On-
col 1993;26:111–116.
13. Nieder C, Nestle U, Niewald M, et al. Accelerated radiother-
apy for brain metastases. Radiother Oncol 1997;45:17–22.
14. Engenhart R, Kimmig BN, Höver K-H, et al. Long-term
follow-up for brain metastases treated by percutaneous stereo-
tactic single high-dose irradiation. Cancer 1993;71:1353–1361.
15. Flickinger JC, Kondziolka D, Lunsford LD, et al. A multi-
institutional experience with stereotactic radiosurgery for sol-
itary brain metastasis. Int J Radiat Oncol Biol Phys 1994;28:
797– 802.
16. Mehta MP, Rozental JM, Levin AB, et al. Defining the role of
radiosurgery in the management of brain metastases. Int J
Radiat Oncol Biol Phys 1992;24:619 – 625.
17. Zubrod CG, Schneiderman M, Frei E Jr, et al. Appraisal of
● F. J. LAGERWAARD et al. 803
methods for the study of chemotherapy of cancer in man:
Comparative therapeutic trial of nitrogen mustard and trieth-
ylene thiophosphoramide. J Chron Dis 1960;11:7–33.
18. Kaplan EL, Meier P. Nonparametric estimation from incom-
plete observations. J Am Statist Assoc 1958;53:457– 481.
19. Cox DR. The analysis of binary data. London: Methuen and
Company; 1970.
20. Cox DR. Regression models and life tables. J Roy Statist Soc
B 1972;34:187–220.
21. Noordijk EM, Vecht CJ, Haaxma-Reiche H, et al. The choice
of treatment of single brain metastasis should be based on
extracranial tumor activity and age. Int J Radiat Oncol Biol
Phys 1994;29:711–719.
22. Patchell RA, Tibbs PA, Walsh JW, et al. A randomized trial
of surgery in the treatment of single brain metastases to the
brain. N Engl J Med 1990;8:494 – 499.
23. Diener-West M, Dobbins TW, Phillips TL, et al. Identification
of an optimal subgroup for treatment evaluation of patients
with brain metastases using RTOG study 7916. Int J Radiat
Oncol Biol Phys 1989;16:669 – 673.
24. Chatani M, Teshima T, Hata K, et al. Prognostic factors in
patients with brain metastases from lung carcinoma: Prelimi-
nary result of the second trial. Strahlenther Oncol 1989;166:
562–567.
25. Chatani M, Matayoshi Y, Masaki N, et al. Radiation therapy
for brain metastases from lung carcinoma: Prospective ran-
domized trial according to the level of lactate dehydrogenase.
Strahlenther Oncol 1994;170:155–161.
26. Swift PS, Phillips T, Martz K, et al. CT characteristics of
patients with brain metastases treated in RTOG study 79-06.
Int J Radiat Oncol Biol Phys 1993;25:209 –214.
27. Kuhn MJ, Hammer GM, Swenson LC, et al. MRI evaluation
of “solitary” brain metastasis with triple-dose gadoteridol:
Comparison with contrast-enhanced CT and conventional-
dose gadopentetate dimeglumine MRI studies in the same
patient. Comput Med Imaging Graph 1994;18:391–399.
28. Nomoto Y, Miyamoto T, Yamaguchi Y. Brain metastasis of
small cell lung carcinoma: Comparison of Gd-DTPA en-
hanced magnetic resonance imaging and enhanced computer-
ized tomography. Jpn J Clin Oncol 1994;24:258 –262.
29. Sze G, Milano E, Johnson C, et al. Detection of brain
metastases: Comparison of contrast-enhanced MR with un-
enhanced MR and enhanced CT. Am J Neuroradiol 1990;
11:785–791.
30. Nieder C, Niewald M, Schnabel K. The results of the radio-
therapy of brain metastases in patients at an advanced age.
Strahlenther Onkol 1995;171:646 – 648.
31. Gaspar L, Scott C, Rotman M, et al. Recursive partitioning
analysis (RPA) of prognostic factors in three Radiation Ther-
apy Oncology Group (RTOG) brain metastases trials. Int J
Radiat Oncol Biol Phys 1997;37:745–751.