Annals of Oncology 12 81-87.2001
© 2001 Kluwer Academic Publishers Primed in the Netherlands
Original article
Prognostic factors and long-term survival in 585 patients with metastatic
breast cancer treated with epirubicin-based chemotherapy
M. Ryberg, D. Nielsen, K. 0sterlind, T. Skovsgaard & P. Dombernowsky
Department of Oncology, Herlev Hospital, University of Copenhagen, Denmark
Summary
Background Analysis of prognostic factors in patients with
metastatic breast cancer treated with epirubicin-based chemotherapy
Patients and methods Data from 469 patients treated with
epirubicin-based chemotherapy for metastatic breast cancer
were used Prognostic factors were identified (Cox multivanate
analysis). A prognostic index was compiled and risk groups
were established accordingly. The applicability of the index was
investigated in a series of 116 patients
Results The prognostic factors identified were, liver, pleural,
soft tissue, lung and bone metastases, performance status > 2,
advancing age, abnormal elevation of serum lactate dehydrogenase and negative/unknown oestrogen receptor status. Four
risk groups were established good, intermediate 1, intermediate
II and poor The median and five-year survivals in percentage
were, good 34 months (26%), intermediate I: 19 months (6%),
intermediate II. 12 months (0%); poor. 7 months (1%) The
corresponding values in the applicability group were 32 months
(23%), 28 months (22%), 18 months (5%); and 6 months (0%)
Conclusions It is more the number and impact on the
organs involved, that predict the patients' survival The construction of a prognostic index could be helpful in assessing the
outlook for patients, especially the quite dramatic difference in
long-term survival between the good and poor risk patients.
Key words epirubicin, metastatic breast cancer, prognostic
factors
Introduction
Patients and methods
The treatment of metastatic breast cancer is a tremendous
challenge, and, despite intensive research, the treatment
is still only palliative. In patients given chemotherapy, a
median survival of 1.5-2 years has been found [1-4].
Prior studies have demonstrated that survival depends
on several factors, such as the number and location of
metastases, oestrogen receptor status (ER status) and
performance status (PS) [2, 3, 5-7]. Metastases to the
liver have been especially correlated with a poor survival
[2, 3, 5, 8, 9].
Other factors also have a prognostic impact such as
the stage of the primary disease [5, 8], prior adjuvant
chemotherapy [3, 9, 10], a short disease-free interval
(DFI) [2, 5, 6, 8-10], and advancing age [11-13]. Overall,
these results are contradictory.
The aim of the present study was to identify risk
factors in patients with metastatic breast cancer at the
time of initiation of chemotherapy, and thus to detect
risk groups. A survival analysis was done on data from a
cohort of 469 anthracycline-naive patients with metastatic breast cancer carried out in a single institution
prospectively included in epirubicin-based chemotherapy
trials from 1983 to 1992 followed by an applicability test
in the subsequent period 1992 to 1995
Four hundred sixty-nine patients, consecutively treated with cpirubicinbased chemotherapy for metastatic breast cancer at Herlev Hospital
during the period November 1983 to September 1992 were entered
(old series) (Table I) Subsequently, data were accessed on 116 patients
treated during the period October 1992 to May 1995 (new series) None
of the patients had received anthracycline-based chemotherapy before
The characteristics of all the patients are given in Table 1
Data on the 469 patients were used in a Cox multivanale analysis
Two hundred sixty patients were entered in two prospective randomised phase III trials 1) epirubicin 60 mg/m 2 (days 1 + 8) or
epirubicin 45 mg/m 2 (days 1 + 8) + vindesine 3 mg/m 2 (day 1) every 4
weeks (122 patients) [14], 2) epirubicin 70 mg/m 2 (days 1 + 8) or
epirubicin 60 mg/m 2 (days 1 + 8) + cisplatin 100 mg/m 2 (day 1, the
first six cycles) every 4 weeks (138 patients) [15] The inclusion criteria
have been published earlier [14, 15] Patients who refused randomisation were treated with epirubicin 70 mg/m 2 (days 1 + 8) (115) every
4 weeks A further 94 patients received epirubicin 130 mg/m 2 (day 1)
every 3 weeks
Data on the new series with 116 patients were used for validation
Eighty-five patients participated in a phase III study epirubicin 130
mg/m 2 every three weeks vs epirubicin 130 mg/m 2 and cyclophosphamide 2500 mg/m 2 alternately every three weeks for a total of eight
cycles Thirty-one patients who refused randomisation were treated
with epirubicin 130 mg/m 2 every three weeks
Previous treatment
Patients classified as high risk at the primary operation received
adjuvant treatment Patients with positive regional lymph nodes and/
82
or a tumour size > 5 cm were considered to be high nsk patients Premenopausal patients aged <50 years received cyclophosphamide,
methotrexate, 5-fluorouracil (CMF) ± adjuvant irradiation From
1989 some premenopausal women, who were oestrogen receptor-posUive (ER-pos), were treated with ovarian irradiation Post-menopausal
women and aged <50 years, were grVen the same treatment as high
risk patients > 50 years these with an ER-pos or oestrogen receptor
unknown tumour (ER-unknown) received tamoxifen for one year, and
those with a receptor negative-tumour (ER-neg) received CMF A
small group of patients were treated with tamoxifen or with CMF ±
tamoxifen, (irrespective of age <50 years) Ninety-two patients
included in the study had received first-line chemotherapy with CMF
at the first relapse
Pretreatment evaluation of all patients included a complete history
and physical examination, blood cell counts, serum chemistry profiles.
X-ray of the chest, and bone scans If the bone scan was abnormal, the
axial skeleton was further evaluated with roentgenograms to determine
the nature of the abnormalities Pleural metastases were documented
either by a positive cytology or biopsy, and metastases to the bone
marrow by aspirate from the iliac crest An ultrasound scan of the liver
was performed in all cases with abnormal liver biochemistry After
cessation of chemotherapy, the patients were followed-up every three
months
Potential prognostic characteristics
The following factors were examined metastases to the liver, lung,
pleura, bone, contralateral breast, lymph node(s), bone marrow, brain,
soft tissue including skin with or without underlying connective tissue/
muscle (patients with only local soft tissue metastases underwent
surgery/irradiation), other sites (abdomen, except the liver) and the
number of metastatic sites Other factors examined were the age of
the patient at the initiation of epirubicin treatment, PS according to
WHO's definition, pretreatment biochemistry values such as haemoglobin (hgb), and serum lactate dehydrogenase (S-LDH), ER status,
primary stage (lymph node status/tumour size), adjuvant chemotherapy/radiotherapy, DF1, time from the diagnosis of primary disease to
treatment with epirubicin (TTE), the number of relapses and previous
CMF treatment for recurrent disease
Prognostic index
A prognostic index was compiled from the prognostic factors identified by the Cox multivanate analysis and each factor in the final Cox
model was assigned a prognostic score calculated as the rounded off
regression coefficient x 10
Statistical analysis
Analysis of the relation between pretreatment characteristics and
survival was based on Kaplan-Meier's curves and Cox's proportional
hazards model
Categorisation of variables followed clinically objective criteria
Thus, metastases to a certain site were assigned 0 for 'no' and 1 for
'yes' Non-examined and unconfirmed suspected sites were entered as
missing Values within the normal range of hgb concentration scored as
0, while anaemia (hgb < 7 0 mmol/l) was assigned 1 S-LDH in normal
range (upper limit 450 U/l) was assigned 0, values of 451-900 U/l
scored 1, and values higher than 900 U/l scored 2 Distribution of the
patients' age at the time epirubicin was initiated was divided into
quartiles The age of the individual patient was assigned 0 for the
youngest age group and 3 for the oldest The influences of age and
LDH were tested for linearity in a separate Cox analysis DFI and TTE
were divided as early < 3 years vs late 3 years With ER-pos as
reference, ER-neg and ER-unknown were assigned to separate variables
The Cox analysis was carried out as a stepwise entering analysis
with Wald test statistics P > 0 10 to exclude a variable and P < 0 05 to
Table I Characteristics of 585 patients
Patients
available for
prognostic
index
n (%)
validation
n (%)
Registered
469(100)
116(100)
Primary local disease
422(90 0)
110(94 9)
Lymph node negative
< 4 positive lymph nodes
> 4 positive lymph nodes
Excact lymph node status unknown
71(15
177(39
98(19
123(26
1)
2)
4)
2)
37(31 9)
43(37 1)
27(23 3)
9(7 8)
Tumour size < 5 cm
Tumour size > 5 cm
Unknown
218(46 4)
95 (20 2)
156(33 4)
72(62 1)
17(147)
27(23 3)
Adjuvant therapy"
None
Chemotherapy
Radiation therapy
Ovanal irradiation
Tamoxifen
Chemotherapy and tamoxifen
120(25 6)
150(32 0)
176(7 5)
0(0)
60(128)
22 (4 5)
43(37 1)
39 (33 6)
31(26 7)
5(4 3)
21 (18 0)
1(0 9)
ER-positive
ER-negative
ER-unknown
169(36 3)
102(21 5)
198(42 2)
74(63 8)
34 (29 3)
8(6 9)
Performance status (WHO)
0-1
2-3
4
Unknown
322(69 7)
113(24 5)
2(0 4)
32(5.4)
107(92 2)
9(7 8)
0(0)
0(0)
Primary advanced disease
First relapse after primary disease
Second relapse after primary disease
Unknown
47(10 0)
233(49 7)
182(38 8)
7(1 5)
6(5 1)
79(68 1)
30 (25 9)
1(0 9)
Age
<46
46-52
53-58
>58
118(25
108(23
125(26
118(25
33(28
32(27
34 (29
17(14
Characteristics
1)
1)
7)
1)
Patients
available
for
4)
6)
3)
7)
DFI (months), median (range)
21 (0-343)
21 (0-268)
TTE (months), median (range)
30(0-376)
25(0-284)
Abbreviations ER - oestrogen receptor status, DFI time from the
primary diagnosis to first recurrence, TTE - time from the primary
diagnosis to the initiation of treatment with epirubicin
° Some patients have more than one treatment modality
keep it in the model Interaction among these variables, e g , age and
ER status was investigated The proportionality assumption was
checked by log minus log plots
The database and the statistical analyses were handled by the
computer programme SPSS 90 for Windows (Statistical Solutions
limited, Cork, Ireland)
Ethics
All the patients entered in the treatment trials gave oral and written
informed consent to treatment and to the use of personal data for
research purposes
83
Table 2
Distribution of individual sites, number of sites, LDH, and
age in DBD pauenis
Variables
Factor
Patients
available for
prognostic
index
n (%)
Patients
available
for
validation
n (%)
Individual site
Liver
Lung
Pleura
Bone
Bone marrow
Soft tissue
Lymph node
Contralateral breast
CNS
Other sites"
111(23 7)
113(24 1)
70(14 9)
248 (52 9)
80(17 1)
106(22 6)
179(39 0)
26(5 5)
7(2 0)
29 (6 2)
26 (22 2)
26 (22 2)
7 (6 0)
49 (41 9)
14 (12 0)
26 6 (22 2)
61 (52 1)
11 (9 4)
0 (0)
4 (3 4)
Number of sites
One site
161 (34 3)
54 (46 6)
152(32 4)
87(18 6)
42 (9 0)
27(5 8)
32
18
10
2
(27 5)
(15 5)
(8 6)
(18)
249(53 1)
111 (23 7)
69(14 7)
40 (8 5)
83
17
13
3
(71 5)
(14 6)
(112)
(2 6)
Two sites
Three sites
Four or more sites
One or more sites unknown
LDH
Normal (4450 U/l)
451-900 U/l
>900U/1
Missing
Table 3 The final Cox model with 469 patients
" Other sites abdominal sites except the liver
Results
The data of 469 patients with advanced breast cancer
were used in the survival analysis Table 2 shows the
distribution of individual metastatic site(s), the number
of sites, LDH, and age.
The median survival time for all 469 patients was 14.7
months (range 0-122.3 months) with a median follow-up
time of 76.3 months (range 0-122.3 months). The series
included 14 patients younger than 35 years of age. Median
survival in this subset of patients was 13.7 months. Liver
metastases were more frequent (37.7%) among these
patients, while metastases to other sites occurred with
same incidence as in the entire series.
Coefficient
Coeff/SE"
P-value
Relative
i lsk
Soft tissue
Lung
Pleura
Liver
Bone
PS
S-LDH
Age
ER-negative
ER-unknown
0444
0 363
0 542
3 35
2 86
3 58
< 0 002
0004
<0001
~ n fin i
£T)
1 AQ
3 Hy
< U UU 1
0 362
0 377
0 580
0 185
0 527
0 544
2 83
2 69
6 89
3 70
3 35
4 47
<0 001
<0 009
<000l
<0001
<0 002
<0001
ft
U bZZ
56
44
72
oO
39
46
79
20
69
72
° The Cox regression coefficient = coefficient/standard error
Table 4 Calculation of score for prognostic factors
Factor
Score
Soft tissue
Lung
Pleura
Liver
Bone
PS (0-1 vs 2-4)
LDH (450-900 vs normal)
LDH ( > 9 0 0 v s normal)
Age (years)
46-52
53-58
>58
ER-negative
ER-unknown
4
4
5
6
4
4
6
12
2
4
6
5
5
" Score the regression coefficient x 10, rounded off to the nearest
integer, maximum score = 50
example, the hazard rate for early compared to late
relapse was 1.15, P = 0.22.
The variables, lymph node status and tumour size,
were excluded from the final Cox model, because data
were only available on 288 patients. A separate analysis,
however, was carried out on the data from the 228
patients, and this showed a significant influence of > 4
positive lymph nodes (RR = 1.35, P = 0.02)
The final Cox model is presented in Table 3.
Cox multivariate analysis
Prognostic index
Individual metastastic sites such as the skin, pleura,
liver, and bone, had a significant negative impact on
survival (Table 3) There was no major correlation between impacts of lung and pleura metastases (r = -0.10).
The influence of bone marrow was correlated to that of
S-LDH, and therefore bone marrow was not kept in the
final Cox model. Preclinical variables such as S-LDH,
age, and PS had an influence on survival (Table 3)
Variables, such as number of metastastic sites,
anaemia, adjuvant chemotherapy, adjuvant irradiation,
relapse number (first vs. subsequent), TTE, and DFI, did
not have a significant influence on survival. For DFI, for
Each prognostic factor was assigned a score defined as
the rounded off value of the Cox regression coefficient
x 10 (Table 4). This enabled us to calculate a patient's
personal prognostic score. For instance, a young patient
( < 4 6 years of age) with an ER-pos tumour, normal
S-LDH, good PS, whose relapse was not located in
skin, lung, pleura, liver or bones would score 0 for all
factors in Table 4, the sum score would be 0, and the
patient would represent the most favourable category in
this series. When the sum score is applied as a prognostic
index (PI) it can deduced from Table 4 that a similar
patient with skin and liver metastases would have a PI of
84
Table 5 Distribution of 585 patients with metastatic breast cancer into risk groups and their corresponding PI, median, one-, two, and five-year
survival
Risk groups
Series
Number of
patients (%)
Median survival
(months)
One-year
survival (%)
Two-year
survival (%)
Five-year
survival (%)
0-10
Old
New
113(29)
21(19)
34
32
87
90
63
67
26
23
Intermediate 1
11-15
Old
New
95(25)
38(34)
19
28
68
76
34
52
6
22
Intermediate II
16-22
Old
New
86(22)
38 (33)
12
18
51
71
17
35
0
5
Poor
>22
Old
New
95(24)
16(14)
19
31
4
13
1
0
Good
PI score"
70
57
" PI score the individual prognostic sum score (seeTable 4)
10. The maximum sum in this system is 50, although 44
was the highest value found
Risk groups
The patients were categorised into one of the four risk
groups: good, intermediate I, intermediate II, or poor
according to their PI (Table 5). The PI ranges were 0-10
for good; 11-15 for intermediate I; 15-22 intermediate
II; and > 22 for poor The median and five-years survival;
were good (113 patients)' 34 months and 26%, intermediate I group (95 patients): 19 months and 6%; intermediate II (86 patients) 12 months and 0%; and poor
(95 patients) 7 months and 1% (Table 5).
New series
A new series consisted of 116 patients. The median
survival was 20 6 months (range 0-85.5 months) and the
median follow-up time was 74.8 months (range 0-85.5
months). No patient was younger than 35 years. One
hundred thirteen patients were available for a survival
analysis based on the PI score. All metastastic sites
necessary for the PI score were represented and the
incidence rates were similar to those in the original
series The median and five-years survival, were good
(21 patients): 32 months and 23%; intermediate I (38
patients)- 29 months and 22%; intermediate II (38 patients): 18 months and 5%, and poor (16 patients) 6
months and 0% The survival data were similar to that
of the old series in the good and poor risk groups, but
differed considerably in the intermediate groups I + II
(Table 5). The Kaplan-Meier survival plots for the four
risk groups in both series are given in Figure la and b
Discussion
Studies on prognostic factors in patients with metastatic
breast cancer vary considerably with respect to the
selection of patients, availability of clinical parameters,
and methods of analysis. Differences in patient selection
is momentous, when the results of different multivanate
analyses are to be interpreted. Definition of risk groups
is only appropriate, if patients are selected on similar
criteria [16] Some studies, like ours, enter any patient
with metastatic breast cancer selected for treatment
with chemotherapy [3, 7, 9, 10], other studies enter
patients only at first recurrence [2, 5, 8, 12]. One study
solely concerns first-relapse patients earlier treated with
adjuvant chemotherapy. One must assume, that such
patients have a poor prognosis compared to that of
others, because the disease is more aggressive at the
time of diagnosis [6]. Information about the location of
metastatic sites varies considerably. Some specify the
location of the metastases to liver, lung, etc. [5, 7, 10],
whereas others classify the metastases for instance in
'classical dominant sites', which include three groups: 1)
visceral dominant sites (liver, lung, brain, pleura or
pleural effusion); 2) bone, and 3) soft tissue dominant
site; (local skin, distant skin, subcutaneous masses, and
lymph nodes) [2, 3]. Details are lost, when a variable
comprises many different locations, and the contribution
from a specific metastatic site therefore tends to be
overridden by the less informative variables included in
'classical dominant sites'. Consequently, each individual
site must be considered separately in the analysis, thereby
permitting the role of interactions between variables to
be investigated.
In the present study, we analysed the influence of
different metastatic sites and found that liver, lung,
pleural, bone and soft tissue metastases had negative
impact. Liver metastases have been identified as a prognostic factor by many authors [3, 5, 8-10], whereas this
could not be confirmed by Hortobagyi et al. [7]. However, alkaline phosphatase, S-LDH, and extension of
disease were analysed together with liver metastases,
and interactions between these four factors may have
influenced the outcome of the analysis [7]. Lung metastases as a prognostic factor was proved by Clark et al.
and Hortobagyi et al. [5, 7]. When lung, liver and
pleural metastases were all comprised in the variable
'visceral dominant site', they became a strong negative
prognostic factor [2, 5] Bone metastases were identified
85
(a) Percent surviving
100
50
Old Good
New Good
1
(b)
100
2
3
4
5
9
10
Years
6
Percent surviving
50
\
\
Old Inter II
1 New Inter II "L
;
-L
New Inter 1
. Old Inter 1
1
Years
Figure 1 (a) Kaplan-Meier survival plot The survival for boths series
in the good and the poor groups, x-axis percentage of patients alive,
y-axis survival in years,
the old series,
the new series
(b) Kaplan-Meier survival plot The survival for boths series in the
intermediate 1 + II groups, X-axis percentage of patients alive, Y-axis
survival in years,
the old series intermediate I and intermediate
II,
the new series intermediate I and intermediate II
as a prognostic factor in the study by Clark et al. [5].
A negative impact of soft tissue metastases in patients
being treated with chemotherapy has been demonstrated
by other authors [2, 9]. Conversely, metastases solely
confined to soft tissue as the first recurrence is not a
significant negative prognostic factor [5, 8]. Abnormally
increased pretreatment S-LDH was the strongest single
prognostic factor in our study, although there was interactions with both liver and bone metastases. This findings
has been confirmed by other authors [7, 10, 17]. Furthermore, several authors have demonstrated that a high
tumor burden was associated with a short survival [3, 7,
18, 19]. In the present study only the amount of metastatic sites were included, not a measurement of the
patients tumour mass.
A PS > 2 at the initiation of chemotherapy resulted
in a significantly decreased survival, as has also been
observed by two other authors [7, 10], although PS was
only included in univariate analysis in the study of
Yamamoto et al. [10]. The present study restricts the
considerations to a cohort of patients aged less than 74
years, however, advancing age at the time of initiation of
chemotherapy was a prognostic factor. This has been
confirmed in a further three studies [10-12, 18], but not
in others [3, 7]. Because of the lack of patients with age
under 35 years we were not able to individualise the
prognostics for these young patients as done by the study
ofFalkson[20].
The prognostic impact of the disease-free interval is
still under discussion. The definitions of DFI, patient
selection criteria and methods to estimate the survival
rate are essential for this discussion. Are patients with
primary advanced disease included or not? It is reasonable to assume that patients with primary advanced
disease have a more aggressive disease than other patients. Furthermore, the proportions of patients with
first vs later relapse should be given. In our study, we
omitted patients with primary advanced disease in the
analysis of DFI and 38.8% had second relapse. The DFI
was not significant in the final multivariate analysis, as
was also found by other authors, who used a similar
definition of DFI and patient selection criteria [7, 9]
DFI was identified as a strong prognostic factor in the
study of Yamamoto et al. However, they included patients with primary advanced disease. Furthermore, there
was no information about the number of relapses. In
that study, a group of patients was used to confirm the
strong influences of the duration of DFI, but the estimation of survival was not the same in the Cox series and
the validation series. Survival was estimated from the
day of randomisation in the Cox analysis but from the
date of first relapse in the validations series [10]. However,
the duration of DFI may carry some basic biological
information, provided that the DFI is defined solely as
the time to the first recurrence In studies which entered
patients only at their first relapse, DFI was significant in
three multivariate analyses [2, 6, 8] and in one univariate
analysis [9]. We also examined the prognostic impact
TTE and found that TTE < 3 years was associated with
a poor survival in the univariate analysis, but insignificant
in the final Cox model.
Prior adjuvant chemotherapy was a prognostic factor
in our univariate analysis, but became insignificant
when the primary stage was included in the Cox multivariate analysis, an observation also done by others [3,
5]. In three studies adjuvant chemotherapy was a significant prognostic factor after relapse in a multivariate
analysis [3, 9, 10]. However, the primary stage was not
included in two of these [3, 9], and the type of adjuvant
treatment and clinical characteristics of the patients
were missing in the third [10]. The proportion of patients
receiving adjuvant chemotherapy was 74% in the Yamamoto study as compared to 36.7% in ours [10]. Treatment
related infections and cardiotoxicity had no impact on
86
the survival [21]. The influence of dose schedules, dose
intensity, response rates and second/third-line chemotherapy were not evaluated. The impact of these factors
on survival is going to be investigated.
A prognostic index can be used to allocate patients
into risk groups We defined four risk groups: good
(score ^ 10) with a median survival of 34 months and a
five-year survival of 26%; intermediate I (score 11-15)
with a median survival of 19 months and a five-year
survival of 6.0%, intermediate II (score 16-22) with a
median survival of 12 months and a five-year survival of
0%; and poor (score > 22) with a median survival of
7 months and a five-year survival of about 1%
In the new applicability series, the survival data were
similar in the good and the poor groups, but not in the
intermediate groups I + II, where the survival figures
were better The reason could be staging of patients with
a more sensitive procedure in the new series and/or the
result of a more effective treatment. Other studies have
defined risk groups with similar survival in the groups
[4, 10]. Thus, Yamamoto et al. defined three risk groups
with a median duration of survival: 45.5 months, 24.6
months and 10.6 months [10]. Rahman et al. investigated
patients complying with the institution's eligibility criteria (normal S-LDH, PS 0-1, and younger age) for
high-dose chemotherapy with stem-cell rescue [4]. These
patients had a median survival of 30 months and a fiveyear survival rate of 21%, when treated with conventional
doxorubicin. From a clinical point of view, it is important to be aware of the substantial difference in the
survival between good and poor risk patients Patients
with no or few negative prognostic factors have a long
median survival of 3-3.5 years, which is probably longer
than usually expected in patients with metastatic breast
cancer given conventional chemotherapy. We therefore
suggest the use of a prognostic index comprising all the
influential prognostic variables and the application of
their regression coefficients as weighting factors, in order
to give us a more balanced view of the outlook for
patients
Conclusions
The present study has substantiated, that a number of
prognostic factors, and not just one factor, is predictive
for the survival of patients treated with epirubicin for
metastatic breast cancer. The construction of a scoring
system, allows us to characterise risk groups with a
considerable difference in median survival and a quite
dramatic difference in long-term survival rates, especially
between the good and the poor risk patients We recommend the use of these risk categories for pre-treatment
stratification in randomised trials, but must at the same
time warn against comparison of results derived from
different clinical trials or from trials undertaken in
different periods of time at the same institution.
Acknowledgements
This work was supported by a grant from the foundation
of Elly Valborg and Niels Mikkelsen The authors are
grateful to K. Kjaer and H. Larsen, MD, for their excellent
technical assistance
References
1
Tormey DC Adnamycin (NSC-123 127) in breast cancer An
overview of studies Cancer Chemother Rep 1975, 6 319-27
2 Vogel CL, Azevedo S, Hilsenbeck S et al Survival after first
recurrence of breast cancer Cancer 1992, 70 129-35
3 Falkson G, Gelman R, Falkson CI et al Factors predicting for
response, time to treatment failure, and survival in women with
metastastic breast cancer treated with DAVTH A prospective
Eastern Cooperative Oncology Group study J Clin Oncol 1991,
9 2153-61
4 Rahman ZU, Frye DK, Budzar AU et al. Impact of selection
process on response rate and long-term survival of potential highdose chemotherapy candidates treated with standard-dose doxorubicin-containing chemotherapy in patients with metastatic
breast cancer J Clin Oncol 1997, 15 3171-7
5 Clark GM, Sledge GW, Osborne KC et al Survival from first
recurrence Relative importance of prognostic factors in 1015
breast cancer patients J Clin Oncol 1987, 5 55-61
6 Goldhirsh A, Gelber RD, Castiglione M Relapse of breast
cancer after adjuvant treatment in premenopausal and penmenopausal women Patterns and prognosis J Clin Oncol 1988,6 89-97
7 Hortobagyi GN, SmithTL, Legha Set al Multivanate analysis of
prognostic factors in metastatic breast cancer J Clin Oncol 1983,
1 776-86
8 Hietanen P, Miettinen M, Ma'kinen J Survival after first recurrence in breast cancer Eur J Cancer 1986, 22 913-9
9 Dunphy FR, Spitzer G, Fornoff JER et al Factors predicting
long-term survival for metastatic breast cancer treated with highdose chemotherapy and bone marrow support Cancer 1994, 73
2157-67
10 Yamamoto N, Watanabe T, Katsumata N et al Construction and
validation of a practical prognostic index for patients with metastatic breast cancer J Clin Oncol 1998, 16 2401-8
11 De La Monte S, Hutchins GM, Moore GM. Influences of age on
the metastatic behaviour of breast carcinoma Human Pathol
1988, 19 529-34
12 Toikkanen SP, Kujan HP, Joensuu H Factors predicting late
mortality from breast cancer Eur J Cancer 1991, 27 586-91
13 Vincent MD, Powles TJ, Skeet R et al An analysis of possible
prognostic features of long-term and short-term survivors of
metastatic breast cancer Eur J Cancer Clin Oncol 1986, 22
1059-65
14 Nielsen D, Dombernowsky P, Skovsgaard Tet al Epirubicin or
epirubicin and vmdesine in advanced breast cancer A phase III
study Ann Oncol 1990, 1 275-80.
15 Nielsen D, Larsen SK, Hansen OP et al Epirubicin (E) or
epirubicin and cisplatin (C) as first-line treatment in advanced
breast cancer Ann Oncol 1990, 1 (Suppl) 24
16. Parmar MKB, Machin D Survival Analysis A Practical Approach, second edition Cambridge Wiley 1996
17 Swenerton KD, Legha SS, Smith T e t al Prognostic factors in
metastatic breast cancer treated with combination chemotherapy
Cancer Res 1979, 39 1552-62
18 Greenberg PCL, Hortobagyi GN, Smith TL et al Long-term
follow-up of patients with complete remission following combination chemotherapy for metastatic breast cancer J Clin Oncol
1996, 14 2197-205
19 Tampellini M, Berruti A, Gerbino A et al Relationship between
CA 15-3 serum levels and disease extent in predicting overall
87
survival of breast cancer patients with newly diagnosed metastatic disease Br J Cancer 1997, 75 698-702
20. Falkson G, Gelman RS, Pretonus FJ Age as a prognostic factor
in recurent breast cancer J Clin Oncol 1986, 4 663-71
21 Ryberg M, Nielsen D, Skovsgaard Tet al Epirubicin cardiotoxlcity An analysis of 469 patients with metastatic breast cancer J
Clin Oncol 1998, 16 3502-8
Received 4 April 2000, accepted 25 September 2000
Correspondence lo
M Ryberg, MD
Department of Oncology
Herlev Hospital, University of Copenhagen
Herlev Ringvej
DK-2730 Herlev
Denmark
E-mail mananneryberg@dadlnet dk