Saturday 24 April 1993
ARTICLES
Age as prognostic factor in premenopausal breast
carcinoma*
Whether or not young age at diagnosis is an
adverse prognostic factor in breast cancer has long
been controversial, in part because much previous
work has not taken due account of menopausal
status and confounding factors. We have analysed
the influence of age on prognosis in a consecutive
series of 1703 patients with stage I-IIIbreast cancer.
All were premenopausal and all were treated in one
centre (Institut Curie, Paris) between 1981 and
1985. Mean age was 44 years (range 23-55) and
median follow-up was 82 months.
Younger patients had significantly lower survival
rates and higher local and distant relapse rates than
older patients. The hazard rate of relapse decreased
over time in the youngest age group (≤33) to reach
that of older patients after 5 years. The relation
between the hazard of recurrence and age was a
continuous one, best fitted by a log-linear function
and indicating a 4% decrease in recurrence for every
year of age. Multivariate analysis of both survival and
disease-free interval demonstrated that the worse
prognosis of young age was independent of other
factors such as clinical tumour size, clinical node
status, histological grade, hormone receptor status,
locoregional treatment procedure, and adjuvant
systemic therapy.
No 8852
This difference in outlook has yet to be explained
biologically but it does suggest the need for a closer
look at the natural history of breast cancer in young
women.
Introduction
The effect on outcome of age at diagnosis of breast cancer
remains controversial. Large population-based studies
point to poorer survival in patients aged 34 years and under.
The proportion estimated to be cured was 8-20% lower in
this group of patients than in patients aged 40-49.1-6 These
observations were confirmed by clinical studies from
Institut Curie7,8 and other groups, for such end-points as
overall survival,9 disease-free survival.,10,11 and local-regional
failures. 12-14 However, conflicting results have been reported
*Based on a presentation to the 34th annual meeting of the American
Society for Therapeutic Radiology and Oncology (San Diego, November,
1992).
of de
ADDRESSES: Departments Radiotherapy (A. la
Rochefordière, MD, F. Campana, MD, J. Fenton, MD, J. R. Vilcoq, MD, A.
Fourquet, MD), Biostatistics (B. Asselain, MD), Medical Oncology
(S. M. Scholl, MD, Prof P. Pouillart, MD), Surgery (J.-C. Durand, MD),
and Pathology (H. Magdelenat, PHD), Institut Curie, Paris,
France. Correspondence to Dr Anne de la Rochefordière, Department
of Radiotherapy, Institut Curie, 26 rue d’Ulm, 75231 Paris, Cédex 005,
France
1040
TABLE I-TUMOUR CHARACTERISTICS
*For grade III, 40 >40p<0 01;ffor 33 >33p<0 05.
No other differences significant. ND= not determmed.
by others.15-l8 These studies had small numbers in the
youngest age groups and the age cut-offs varied;
furthermore all age groups were compared, irrespective of
menopausal status. The one published study19 on
premenopausal patients alone found that women below 35
years of age did less well than older premenopausal patients,
but the difference was not significant. We have addressed
this controversy in a large group of premenopausal patients
identified in the Institut Curie’s prospective database on
invasive non-metastatic breast cancer.
Patients and methods
Between January, 1981, and December, 1985,3771 women with
clinical stage I-IIIunilateral invasive breast carcinoma were treated
at the Institut Curie. 1703 (45%) were premenopausal.
Patient characteristics
Median age was 44 (range 23-55). To assess the prognostic role of
age at diagnosis we arbitrarily chose three age groupings-namely
100 very young patients aged 33 or less (group 1), 356 patients aged
34-40 years (group 2), and 1247 aged 41-55 (group 3). We noted for
each group clinical tumour size, axillary node status, stage,
histological type, and grade.10 hormone receptor levels, and
pathological axillary node status (tables I and II). Tumour grade was
available in 1437 patients (84%). Oestrogen (ER) and progesterone
(PR) receptor levels were assessed in 728 (43%) and 1208 patients
(71 %), respectively, and 250 finol/mg DNA was the chosen cut-off
value over which receptor status was considered to
Treatment
The general treatment policy was aimed at breast preservation by
combining radiotherapy and limited surgery. 1317 patients (77%)
had a breast-preserving treatment whereas 386 (23%) had modified
mastectomy. Patients older than 40 years were more likely to have
had a mastectomy than younger patients (p< 0-001) (table n).
Axillary dissection was done in 827 patients (49%).
TABLE II-TREATMENT CHARACTERISTICS
622 patients (36%) were treated with wide excision followed by
radiotherapy (RT), and 260 of those with tumours of 3 cm or has
had been included in a prospective randomised trial on axillary
dissection.22 729 patients (43%) with larger tumours were treated
with RT Cobalt-60 as the primary treatment.23 Mean dose to the
breast was 58-8 Gy (range 50-0-67-6), and the 548 patients with a
complete or a partial response received a boost dose to the tumour
site to a mean total dose of 76-6 Gy (range 60-0-90-0) and surgery
was not done. 147 patients with a partial response after 58 Gy were
treated with subsequent wide excision.
Primary modified radical mastectomy was done in 352 patients
(21 %), followed in 169 by RT. A further 34 patients, who had had
primary radiotherapy, underwent a mastectomy because of
insufficient tumour regression after a mean dose of 58 Gy.
Adjuvant chemotherapy was administered to 371 patients (22%)
via an anthracycline-containing regimen (102 patients) or
cyclophosphamide/methotrexate/5-fluorouracil (CMF) (269
patients) for four to six cycles (table n). These patients were more
frequently node positive and hormone-receptor negative.
Tamoxifen was administered to 126 patients (7%), associated in 77
patients with CMF.
Statistical methods
Cause-specific survival was calculated from the date of first
treatment, disease-related death being scored as an event with
censoring of other patients at the time of last follow-up or of
non-disease-related death. Disease-free interval was also calculated
from date of first treatment, first recurrence (local or distant) being
scored as an event, with censoring of other patients at the time of last
follow-up or of death. Local recurrence was defined as tumour
arising in the treated breast or chest wall.
To draw survival curves we used Kaplan and Meier estimates.24
Annual hazard rates of recurrence were calculated as an estimate of
the annual conditional probability of recurrence from breast cancer.
Curves were compared by the log-rank test.25 A relation between
age and survival was sought by two different approaches. The total
cohort was first stratified into the previously described three age
groups. In a separate approach, we assumed a log-linear relation
be positive.21
Fig 1-Kaplan-Meier estimates of probability of cancer-
specific survival.
 1041

Fig 2-Kaplan-Meier estimates of probability of freedom

from relapse (FFR, local or distant).

between hazard and co-variate of the form log A(t) x age. Here
= hazard rate, t = time, and 0 = regression coefficient. Age (in Fig 3-Smoothed yearly hazards plots and local or distant
relapse.
years) was entered into a Cox regression mode}26 as a quantitative
covariate. We also tested against other functions of hazard such as Univariate analyses
log .(t) 0 x (age)2 or log A(t) log (age), to model mathematically
= =

the relation between age and recurrence or death.
The influence of age, adjusted for other prognostic factors, was
assessed in a multivariate analysis by the Cox proportional hazard
model in a forward stepwise regression. Age was entered as a binary
function (ie, 33 years older or as 40 years/older). Confounding
variables with k subgroups were coded from 0 for the group with the
best prognosis to (k-1) for the worst prognostic group. This model
assumed a loglinear relation of relative risks between two
subsequent subgroups when k > 2. Missing values (receptor levels,
tumour grade) were coded as a separate variable (missing/not
missing), and were retained in the model.
Finally, another Cox regression model with time-dependent
covariates was performed to verify the proportional hazards
assumption, amended by Stablein et al.’
Survival rates and relative risks (RR) are presented with their
95% confidence intervals (CI).
Age and survival. Younger patients had significantly
lower survival rates (fig 1). Tests for trend indicated a
significant difference in disease-related survival rates
between the three age groups 1-3 (p < 0-0003), and there
were significant differences between 1 and 2 (p < 0’03) and
between 2 and 3 (p < 0-02). At 5 years, 68 patients remained
at risk in group 1, 267 in group 2, and 1010 in group 3. The
5-year actuarial disease-related survival rates were 71-2%
(62-0-80-4) for patients aged 33 years or less, 85-7%
(81-9-89-6) for those between 34 and 40, and 87-1%
(85 3-88-9) for the oldest age group. After 5 years, the
hazard rate of death seemed to decrease in the youngest
group while it remained constant in the other two. However,
no significant deviation from the proportional hazards
assumption was observed (p 0-29). Unadjusted rate ratios
=

for cancer-specific deaths were: 20 (1-1&mdash;3-5) in patients aged

Results 33 years or less and 1-1(0-8-1-5) in the intermediate age
Median follow-up was 82 months (range 4-132), and was group (1 for the oldest group).
similar for the three age groups. 130 patients (7%) were lost Age and relapse-free interval. Differences between the
to follow-up before 5 years (4/100, 41/356, and 85/1247, three age groups (fig 2) were highly significant (p for trend
respectively). 359 patients have died. 27 deaths were not < 0-0001). The rate was lower in patients aged 33 or less but
cancer-related deaths (0/34, 4/83, and 23/242, respectively). not significantly so when compared with the rate in patients
aged 34-40 (p=0-09). The rate was significantly lower for
Distribution of clinical pathological, and treatment factors the age 34-40 cohort than for patients older than 40
The proportion of patients with a family history of breast (p<0-0001). At 5 years, disease-free rates were 44%
cancer was similar in all three groups (17%, 20%, and 24%, (34-0-53-6) for patients under 34,59% (54-0-64-0) for those
respectively). There were no significant differences in aged 34-40, and 72% (69’5-74’3) for those over 40. The
distribution of mean tumour size, clinical tumour size,
clinical node status, clinical stage, histology of the primary TABLE III-MULTIVARIATE ANALYSIS OF CANCER-SPECIFIC
SURVIVAL
tumour, or ER level (table I). Microscopic axillary node
invasion was present in 56% (27/48) of patients in group 1,
in 46% (73/161) of those in group 2, and in 43 % (263/618) of
those in group 3. These differences were not significant.
However, younger patients differed from older patients by a
higher proportion of grade III tumours in groups 1/2
combined than in group 3 (22 % vs 17 %, p = 0-01), lower PR
levels in tumours of patients in group 1 than in groups 2/3
(26% vs 20%, p < 005), and more frequent use of
breast-conserving treatment in groups 1 and 2 than in group
3 (84% vs 75%, p < 0-001). Also adjuvant chemotherapy
was more frequent in the two youngest age groups
(p < 005); the type of chemotherapy used (anthracycline-
containing regimens or CMF) was equally distributed in the
three age groups. Antioestrogens were more frequently
given to patients over 40 years of age (p < 0 05).
1042
TABLE IV-MULTIVARIATE ANALYSIS OF RELAPSE-FREE
INTERVAL
unadjusted rate ratios for relapse were 2-5 (1-7-3-6) for
patients under 34 years and 1-6 (1-3-19) for those aged
34-40 (1 for the oldest group).
Annual hazard rate of relapse. At 24 months, this rate was
2-5-fold higher in group 1 than in group 3 (fig 3). After 60
months, however, this gap was closing, and the significant
(p = 0-007) deviation from the proportional hazards
assumption for age 33 years or less confirms a decreasing
relative risk of relapse over time.
Multivariate analyses
The following variables were taken into account in the
Cox model: age, clinical tumour size, clinical node status,
tumour grade, ER and PR status, local-regional treatment
(ie, wide excision + RT, yes/no or radical surgery, yes/no),
and adjuvant treatment. We assumed that there would be no
additional information by including histological type and
mean tumour size. Clinical stage was not included because
we analysed tumour size and node status separately. Since
the assumption of a constant hazard rate of relapse over time
diverged from the observed data, we did the multivariate
analysis at 5 years of follow-up censoring patients at 60
months. We confirmed that age at diagnosis independently
contributed to the risk of death in premenopausal patients
(table ill). The RR for death was 1 -8 times higher in group 1
than in group 3 (p<0 04). Other variables independently
associated with survival were clinical tumour stage
(p < 0&deg;0001), clinical node status (p < 0&deg;0001), PR status
(p < 0-0001), and tumour grade (p < 0.0001); ER status, type
of local regional treatment, and use of adjuvant treatment
did not predict survival.
Fig 4-Hazard risk of relapse (local or distant).
Age computed as continuous variable. Broken lines show 95% CI
limits. RR per year of age=096.
Age, when adjusted for all other variables, remained a
strong independent predictor of relapse (table iv), the RR
for recurrence being 22 times higher in group 1 than in
group 3 (p 0 02). For group 2 vs group 3 the RR was 1-55
=
(p< 0-0001).
On the assumption of a log-linear relation between age
and recurrence, we did a multivariate analysis with age as a
continuous variable. RR for recurrence was 0-96 (0-94-0-97)
for every increasing year of age-ie, the risk of recurrence
fell by 4% (2-6-5-4) for every year of age (fig 4). For
cancer-specific death there was a 2% (04-4-3) decreased
risk for every year of age. Application of other functions of
recurrence hazard related to age suggested that the above
log-linear function provided the best fit.
Discussion
In this relatively homogeneous population of
premenopausal women treated for non-metastatic invasive
breast carcinoma we observed a significant adverse effect of
young age on cancer-specific survival and on relapse-free
interval. This increased risk over the first 5 years of
follow-up was independent of confounding variables. The
small difference between adjusted and non-adjusted rate
ratios shows that no confounding variable interfered to any
great extent with the age factor. This study also showed that
older premenopausal patients (aged 41-55) had the best
prognosis.
The prognostic influence of age at diagnosis in breast
cancer has long been controversial. Interpretation of earlier
studies has been complicated by differences in populations,
cut-off ages, follow-up times, and endpoints. Registry-
based studies’-35,6 have demonstrated that women aged
40-49 have the best prognosis while women under 35 years
of age have poorer survival prospects. There were few very
young patients (below 30) in these series, the incidence of
breast cancer in this age group being low. However, one
study1 reported a 5-year relative survival of 62-9% (n 237)
=
compared with 75-8% in patients aged 45-49 . When
Rutqvist et al 17 studied 14 731 patients divided into two
groups at age 55 the younger patients fared significantly
better, but subgroups of premenopausal women were not
analysed separately. None of these registry-based studies
had information on clinical prognostic indicators such as
tumour grade, hormone receptor levels, and differences in
treatment.
Some investigators who, like us, have used non-registry
data with adjustment for clinical variables, have supported
the concept of a poorer prognosis for younger women.9,11 In
a retrospective study of long-term outcome for 99 patients
aged 30 or less with early stage breast cancer, Lee et al10
calculated survival rates to be 10-20% lower than expected
from historical data on older women, and this reduction is
confirmed in our study. Other investigations, however,
found no correlation between age at diagnosis and
outcome,15,16,18 but confounding factors may have prevented
these being definitive assessments of young age as an adverse
prognostic factor. Problems include too short a follow-up,18
the selection of patients by clinical stage or local-regional
treatment,18 the use of historical controls ’16 and crude age
groupings.l5 By contrast, our study was based on a large
group of homogeneously treated patients in one institution.
Our choice of age cut-offs was dictated by the need to have
groups large enough to permit valid statistical inferences (at
least 100 patients). The oldest age group corresponded to
one already known to have a better prognosis ( > 40).
Nonetheless the relation between age and outcome was

continuous being best described by a log-linear function in
which for each increasing year of age the RR of recurrence
and death fell by 4% and 2%, respectively.
Both young and old patients’ tumours have the potential
for rapid growth and invasion, leading to early
dissemination and death, but in very young patients this is
observed more frequently.18 Is the age factor related to an
identifiable, age-specific, biological entity--eg, a genetic
abnormality? Hereditary genetic defects in breast cancer are
more frequent in younger patients28 but no specific gene
abnormality has been reliably associated with more
aggressive disease. Another explanation might be selection
of rapidly evolving disease. We found that young patients
had recurrences earlier than older patients, and the risk of
recurrence decreased over time to reach that of older
premenopausal patients after 5 years. A short latency period
and/or a more rapid doubling-time may indicate earlier
arising of cancer in the breast. High tumour grades or
negative PR status are more frequently found in younger
patients 13,18 29 but this does not explain this rapid evolution.
Rapid growth could explain the higher rate of early local or
distant relapses and adverse early prognosis of younger
patients.7,30 By contrast, slower growth of disease could
mean that recurrence develops more slowly in older
patients, and in the long term the cumulative rate of failure
may be no higher in younger than in older premenopausal
patients. Unlike Adami et all and Sant’s group6 we did not
find a change in risk of death in young patients over time.
We thank Mrs Chantal Gautier and Mrs Marie-Christine Falcou for their
kind technical assistance.
REFERENCES
1. Adami HO, Malker B, Holmberg L, Personn I, Stone B. The relation
between survival and age at diagnosis in breast cancer. N Engl J Med
1986; 315: 559-63.
2. Host H, Lund E. Age as a prognostic factor in breast cancer. Cancer 1986;
57: 2217-21.
3. Lees AW, Jerkins HJ, May CL, Cherian G, Lan EWH, Hanson J. Risk
factors and 10-year breast cancer survival in northern Alberta. Breast
Cancer Res Treat 1989; 13: 143-51.
4. Nemoto T, Vana J, Bedwani RN, Baker HW, MacGregor F, Murphy
GP. Management and survival of female breast cancer: results of a
national survey by the American College of Surgeons. Cancer 1980; 45:
2917-23.
5. Ries LG, Pollack ES, Young JL Jr. Cancer patient survival: surveillance,
epidemiology and end results program, 1973-79. J Natl Cancer Inst
1983; 70: 693-707.
6. Sant M, Gatta G, Micheli A, et al. Survival and age at diagnosis of breast
cancer in a population-based cancer registry. Eur J Cancer 1991; 27:
981-84.
7. Fourquet A, Campana F, Fafrani B, et al. Prognostic factors of breast
recurrence in the conservative management of early breast cancer: a
25-year follow-up. Int J Radiat Oncol Biol Phys 1989; 17: 719-25.
8. Vilcoq JR, Calle R, Stacey P, Ghossein NA. The outcome of treatment by
tumorectomy and radiotherapy of patients with operable breast cancer.
Int J Radiat Oncol Biol Phys 1981; 7: 1327-32.
9. Ribeiro GG, Swindell R. The prognosis of breast carcinoma in women
aged less than 40 years. Clin Radiol 1981; 32: 231-36.
10. Lee CG, McCormick B, Mazumdar M, Vetto J, Borgen PI. Infiltrating
breast carcinoma in patients age 30 years and younger: long term
outcome for life, relapse and second primary tumors. Int J Radiat Oncol
Biol Phys 1992; 23: 969-75.
11. Noyes RD, Spanos WJ, Montague ED. Breast cancer in women aged 30
and under. Cancer 1982; 49: 1302-07.
12. Bartelink H, Borger JH, Van Dongen JA, Peterse JL. The impact of
tumour size on local control of breast-conserving therapy. Radiother
Oncol 1988; 11: 297-303.
13. Kurtz JM, Jacquemier J, Amalric R, et al. Why are local recurrences after
breast-conserving therapy more frequent in younger patients? J Clin
Oncol 1990; 8: 591-98.
1043
14. Recht A, Connoly JL, Schnitt SJ, et al. The effect of young age on tumor
recurrence in the treated breast after conservative surgery and
radiotherapy. Int J Radiat Oncol Biol Phys 1988; 14: 3-10.
15. Mueller CB, Ames F, Anderson GD. Breast cancer in 3558 women: age as
a significant determinant in the rate of dying and causes of death.
Surgery 1978; 83: 123-32.
16. Rosen PP, Lesser ML, Klnne DW, Beattle EJ. Breast carcinoma in
women 35 years of age or younger. Ann Surg 1984; 199: 133-42.
17. Rutqvist LE, Wallgren A, Nilsson B. Is breast cancer a curable disease? A
study of 14,731 women with breast cancer from the Cancer Registry of
Norway. Cancer 1984; 53: 1793-800.
18. Solin LJ, Fowble B, Schultz DJ, Goodman RL. Age as a prognostic
factor for patients treated with definitive irradiation for early stage
breast cancer. Int J Radiat Oncol Biol Phys 1989; 16: 373-81.
19. Langlands O, Kerr GR. Prognosis in breast cancer; the effect of age and
menstrual status. Clin Oncol 1979; 5: 123-33.
20. Bloom HJS, Richardson WW. Histological grading and prognosis in
breast cancer. Br J Cancer 1957; 11: 339-77.
21. Magdelenat H, Lain&eacute;-Bidron C, Merle S, Zajdela A. Estrogen and
progestin receptor assay in fine needle aspirates of breast cancer:
methodological aspects. Eur J Cancer Clin Oncol 1987; 23: 425-28.
22. Cabanes PA, Salmon RJ, Vilcoq JR, et al. Value of axillary dissection in
addition to lumpectomy and radiotherapy in early breast cancer. Lancet
1992; 339: 1245-48.
23. Vilcoq JR, Calle R, Schlienger P. Irradiation techniques for conservative
treatment of localized breast cancer. In: Harris JR, Hellman S, Silen W,
eds. Conservative management of breast cancer. New surgical and
radiotherapeutic techniques. Philadelphia: JB Lippincott, 1983:
213-24.
24. Kaplan EL, Meier P. Non parametric estimation from incomplete
observations. J Am Stat Assoc 1971; 53: 457-81.
25. Peto R, Peto J. Asymptomatically efficient rank invariant test procedures.
J R Stat Soc A 1972; 135: 185-98.
26. Cox DR. Regression models and life table (with discussion). J R Stat Soc
B 1972; 34: 187-220.
27. Stablein DM, Carter WH, Novak JW. Analysis of survival data with non
proportional hazard functions. Controlled Clinical Trials 1981; 2:
149-59.
28. Claus EB, Risch N, Thompson D. Genetic analysis of breast cancer in the
cancer and steroid hormone study. Am J Hum Genet 1991; 48: 232-42.
29. Fillers EMK. Histological grade of breast cancer in younger women.
Lancet 1992; 339: 1483.
30. Fisher B, Anderson S, Fisher ER, et al. Significance of ipsilateral breast
tumour recurrence after lumpectomy. Lancet 1991; 338: 327-31.
From The Lancet
Darwin’s youth
Charles Robert Darwin was bom in 1809, the younger son of a
successful physician in Shrewsbury. His father had been taken to
Shrewsbury at the age of 19 by his father (himself a physician still
well-remembered as Erasmus Darwin) and there left with [.20 to set
him up in practice. This C20 was so many times multiplied by his
industry and talents that his six children, including Charles, were
spared the need to make money for themselves. Charles was sent to
Edinburgh to study medicine but having spent there two sessions
and having seen, as he writes, "two very bad operations, one on a
child, I rushed away before they were completed. The two cases
fairly haunted me for many a long year", he did not like the thought
of becoming a physician. Therefore his father, fearing that his son
would become an "idle sporting man", for he was a fine shot and
devoted to field sports, planned to make him a clergyman. To
Cambridge he went with the intention of taking Holy Orders but, he
notes in his autobiography, "no pursuit at Cambridge was followed
with so much eagerness or gave me so much pleasure as collecting
beetles". Rural rectories have often been occupied by men who were
more naturalist than parson but Darwin’s passion must have
excluded even a weekly sermon and a few pastoral visits; it drove
him instead to exile for five years as naturalist to the voyage of the
Beagle which sailed to Patagonia, to the South Seas and then home
by way of the East Indies.
(Jan 30, 1943)