Content uploaded by Peter Robert Mullins
Author content
All content in this area was uploaded by Peter Robert Mullins on Dec 15, 2020
Content may be subject to copyright.
(CANCER RESEARCH 43, 2985-2990, June 1983]
0008-5472/83/0043-OOOOS02.00
Progesterone and Estrogen Receptors as Prognostic Variables in Breast
Cancer1
Barbara H. Mason, Ian M. Holdaway, Peter R. Mullins, Lye H. Yee, and Ronald G. Kay
Departments of Surgery [B. H. M., L H. Y.] and Endocrinology [I. M. H.], Auckland Hospital and Departments of Community Health [P. R. M.J and Surgery [R. G. K.],
University of Auckland, Auckland, New Zealand.
ABSTRACT
Estrogen receptor (ER) and progesterone receptor (PR) levels
have been measured in 374 tumors from patients with primary
breast cancer and compared with axillary nodal status and other
patient variables to determine their relationship to prognosis.
Nodal status reliably predicted disease-free interval and overall
survival, and both ER and PR status predicted overall survival
both individually and within node-positive and node-negative
subgroups. PR but not ER status was also able to predict
disease-free survival both overall and in the node-positive
subgroup. When the two receptor measurements were used in
combination, a group of receptor-negative, (ER- and PR-nega
tive), node-negative patients were identified with a significantly
worse survival than that for an ER- and PR-positive group of
node-positive patients. It is apparent that receptor status pro
vides useful prognostic information in patients with early breast
cancer and that ER and PR assays used in combination identify
a subgroup of node-negative patients with poor prognosis who
are likely to benefit from adjuvant therapy following mastectomy.
INTRODUCTION
There is a continuing search for methods to identify patients
at risk of developing recurrent breast cancer following mastec
tomy. Accurate estimation of prognosis is of particular impor
tance when considering the necessity for adjuvant therapy after
initial breast surgery. If prognosis could be determined more
precisely than at present, unnecessary treatment would be
avoided in favorable subgroups, and appropriate prophylactic
therapy would be reserved for high-risk patients. It has become
clear that tumor size (10) and grade (2) and, in particular, axillary
nodal status (10) are important prognostic indicators. Recently,
a number of groups have analyzed ER2 status as a further
prognostic variable in early breast cancer. The presence of
receptors in the primary tumor has been shown to either improve
(3, 4, 6, 8, 15, 16) or not influence (1, 11, 12, 20) prognosis. A
study of PR suggested that PR-positive patients had fewer
métastasesthan did PR-negative patients but a similar number
of local recurrences (19).
In the present study, both ER and PR content of primary
breast tumors have been measured and analyzed in an effort to
improve the assessment of prognosis. Other prognostic indica
tors have also been assessed in relation to receptor data in an
effort to determine whether receptor measurements provide
additional information complementing existing methods of pre
dicting disease outcome.
' This study was supported by the New Zealand Medical Research Council and
the Auckland Division of the Cancer Society of New Zealand.
2The abbreviations used are: ER. estrogen receptor; PR, progesterone receptor.
Received March 16, 1982; accepted January 21, 1983.
MATERIALS AND METHODS
Auckland, New Zealand had a population of 797,000 in 1976 and a
population of 829,000 in 1981. On the average, 280 new cases of breast
cancer are diagnosed each year. Data on all new cases of breast cancer
presenting between September 1976 and September 1980 were re
corded by the Auckland Breast Cancer Study Group on a PDP11
computer disc file. Follow-up records are updated every 9 months. The
frequency of clinical examinations and diagnostic procedures is at the
discretion of individual clinicians, but clinical assessments are usually
made quarterly for 1 year, semi-annuallyfor the next 2 years, and then
annually.Patients lost to follow-up, whose cause of death was unknown,
or in whom death was not due to breast carcinoma were included in the
present study but were recorded as only beingon study up to the time
of last contact. Follow-up records to within 6 months of the analysis
were obtained in 97% of all other patients.
ER and PR were measured in tumor tissue by a dextran-charcoal
method as described previously (13). In the present study, ER levels
were considered negativeat <5 fmol/mg, since the response of patients
with advanced breast cancer to endocrinetherapy was very low when
tumor receptor levels were below this value (14).A significant concentra
tion of PR was arbitrarily defined as 1 fmol/mg or greater for actuarial
life-table analyses. Significant values for PR were arbitrarily defined as
either ^\, »3,or »5fmol/mg in the Cox's life-table regressionanalysis.
All patients, except 63, had a modified radical mastectomy or a simple
mastectomy with axillary clearance. The remaining63 patients had local
tumor biopsies before proceeding directly to radiotherapy and are in
cluded only in Chart 1.
StatisticalMethods. Actuariallife-tableanalysis(17) was madeusing
either date of first recurrence or death as end points, and the results
were analyzed using the generalized Wilcoxon statistic (9). Intervals of 1
month were used in the life-table analysis.When comparing the interac
tion of receptor and nodal status, Cox's life-table regression model was
used, as implemented in the COXREGR procedure of the statistical
package SAS (5). These computations were performed on an IBM4341
computer.
RESULTS
Over the period 1976 to 1980,1136 new breast cancer cases
were recorded in Auckland. Measurements of both ER and PR
were made in 437 tumors (39%). Median follow-up time of these
patients from date of diagnosis to date of death or last contact
was 2.6 years. In 231 tumors (20%), only ER was measured,
because the amount of tissue available was too small to permit
analysis of both receptors. Tissue was not received for receptor
assay from 467 patients (41%). The absence of these patients,
however, did not appear to bias the study series. Thus, similar
proportions were contributed from the various hospitals, they
were of the same average age (59 years), and a similar proportion
were node positive (1:1.5) compared to the total group.
To date, 10 patients have died from causes other than breast
cancer. The cause of death was unknown in 2 patients, and 9
patients have been lost to follow-up. The following analysis is
JUNE 1983 2985
Research.
on December 14, 2020. © 1983 American Association for Cancercancerres.aacrjournals.org Downloaded from
S. H. Mason et al.
restricted to those patients for whom both receptors were meas
ured. The pathological status of the axillary nodes was reported
in all cases and confirmed by histology in 374 of the 437 patients.
Pathology reports did not always state the number of nodes
observed, although all reports state that nodes were found and
examined. In the node-negative group, 54% of the reports gave
the number of nodes examined, with a median of 9 nodes.
Various possible prognostic factors are compared with the
receptor and nodal status of the patient group in Table 1. There
were significantly more large tumors in the node-positive com
pared to the node-negative group (p < 0.005). Other variables,
including menopausal status, tumor grade, and receptor status
had the same distribution in both nodal groups. As expected,
menopausal status (p < 0.01) and tumor histological grade (p <
0.005) had a significantly different distribution between ER-
positive and ER-negative groups.
In addition, PR status was clearly related to ER status, as
described previously (13). The 3 variables, menopausal status,
tumor size, and grade, are evenly distributed between PR-
positive and -negative groups. Surgical adjuvant chemotherapy
(alkeran) was received by 38% of the node-positive group (57
patients). These patients are not significantly associated with a
particular receptor group.
The disease-free interval and survival time of patients grouped
by receptor status are shown in Chart 1. There is a significant
relationship between ER status and survival (p < 0.0001), and
both disease-free interval (p < 0.001) and survival (p < 0.0001)
were significantly related to PR status.
The disease-free interval and survival time of patients related
to either ER and nodal status or PR and nodal status combined
are shown in Charts 2 and 3, respectively. It is clear that of the
2 variables, nodal status was the main determinant of disease-
free interval, although PR status had a significant additional
influence restricted to node-positive patients (p < 0.05). Overall
survival, however, was significantly influenced by ER status in
both node-positive (p < 0.005) and node-negative (p < 0.05)
groups. PR status was significantly related to survival only in
node-positive patients (p < 0.05).
The relationship between different combinations of ER, PR,
and nodal status and either disease-free interval or survival is
shown in Charts 4 and 5. There was a significant difference in
disease-free interval between ER-negative/PR-negative and ER-
positive/PR-positive patients within both node-positive and node-
negative subgroups (p < 0.05; Chart 4). Survival was similary
influenced by receptor status in both node-positive and node-
negative groups (p < 0.05; Chart 5). Importantly, a subgroup of
node-negative patients (node-negative/ER-negative/PR-nega-
tive) had a significantly (p < 0.05) worse survival than a node-
positive (node-positive/ER-positive/PR-positive) subgroup (Chart
5).The influence of receptor and nodal status on prognosis was
also analyzed by an alternative method to the actuarial-life table
technique. The interaction between ER, PR, and nodal status
with time to first recurrence or length of survival was studied
using Cox's life-table regression. As with the actuarial analysis,
ER status was not significantly related to time of first recurrence
Table 1
Distribution of patient variables
AxillarynodesTumor
pre
sentMensesPre-
andintramenopausalPostmenopausalUnknownTumor
size«5cm>5cmUnknownTumor
grade123UnknownEstrogen
receptors»5
fmol/mg<5
fmol/mgProgesterone
receptorses3
fmol/mg<3
fmol/mgAxillary
nodesTumor
presentTumor
absentAdjuvant
therapyAlkeranHormonalRadiotherapyChemotherapyNo
therapyNo.51963a12028232719101925871795763183%413937652537404241393842100757510027Tumor
ab
sentNo.731483208142945291411329211411021221%596163357563605859616258252573Estrogen
receptors^5
fmol/mgNo.a621566196253•65522141•14183921323323186%506460585076467644615958257561<5fmol/mgNo.62881321716172610144106589224611118%503640425024542456394142752510039Progesterone
receptors3*3
fmol/mgNo.71110416418363216131a141447111424441152%5745504350443363294751425010010050<3
fmol/mgNo.531342164241640321118310679110334152%4355505750566637725349585050
" Significant difference between groups within cell (p < 0.01).
2986 CANCER RESEARCH VOL. 43
Research.
on December 14, 2020. © 1983 American Association for Cancercancerres.aacrjournals.org Downloaded from
Steroid Receptors and Prognosis in Breast Cancer
12 K~ » 21 24 27 30 33
MONTHS AFTER DIAGNOSIS
Chart 1. Actuarial analysis of disease-free survival and overall survival of breast cancer patients according to the presence of tumor ERs and PRs. x, ER-positive,
n = 252; O, ER-negative, n = 185; •,PR-positive, n = 213; D, PR-negative, n = 224.
but, unlike the actuarial results, PR status also failed to correlate
significantly with disease-free interval (Table 2). This result was
not altered if a positive level of PR was redefined as &3 or &5
fmol/mg. ER and PR, however, each added significantly to nodal
status when determining the duration of survival (Table 3). As
the level of definition of a positive PR concentration was in
creased from 2=1to >3 to >5 fmol/mg, the value of PR status
as a prognostic indicator increased. Thus, when a positive PR
level was defined as =*3 fmol/mg, the addition of PR status to
nodal and ER status led to a significant improvement in assess
ment of survival. Neither receptor was superior to the other in
assessing survival time.
Tumor size and grade were also studied as prognostic factors
in relation to nodal and receptor status. Tumor size did not
further influence survival when nodal status was known (Cox's
life-table regression model), although there was a significant
improvement if only receptor status was known (p < 0.05). The
addition of tumor grade to ER or PR results did not significantly
add to the assessment of survival.
DISCUSSION
The identification of breast cancer patients at risk of tumor
recurrence following initial surgery has become increasingly im
portant in view of adjuvant therapies known to delay the ap
pearance of métastases(7) or improve survival (18). Currently,
the status of the axillary nodes at surgery is considered the most
important variable in predicting tumor recurrence, although at
tention has been drawn to the possibility that receptor assays
may provide additional information (7).
The present study provides further support for a role for
receptor status in assessing prognosis in breast cancer. Only
PR status appeared to relate to time to first recurrence when
assessed by actuarial analysis either alone or in node-positive
patients (Charts 1 and 3). The usefulness of PR status for
determining time to recurrence (Charts 1, 3, and 4) was not,
however, confirmed by Cox's life-table analysis (Table 2). In
contrast, ER status was not significantly related to disease-free
interval when analyzed by either actuarial life-table analysis
(Charts 1 and 2) or Cox's life-table regression analysis (Table 2).
A number of other groups have also found ER status unhelpful
in assessing disease-free interval (1, 11, 12, 20, 21). Pichón ef
al. (19) found PR status to be more important than ER status in
determining prognosis in early breast cancer. However, a number
of groups have found that ER status helps to predict disease-
free interval both in isolation (3-5,8,16) and within nodal groups
(3, 4, 16). Currently, there seems to be no resolution of these
divergent results and, clearly, longer follow-up of carefully cate
gorized patient groups is required.
By comparison with results comparing receptor status to time
to first recurrence, it seems clear from the present study that
receptor levels are clearly important when assessing overall
survival time. Thus, either ER or PR status identified patients
with significantly prolonged survival (Charts 1 to 3 and 5). When
patients were grouped according to nodal status, receptor levels
still indicated subdivisions with significantly longer survival (Table
3; Charts 2 and 3). These findings are in agreement with other
reports in the literature (6, 8).
Knowledge of nodal status and levels of both ER and PR
enabled survival to be further categorized within patient
subgroups (Chart 5). In particular, analysis of both receptors
identified a group of node-negative patients (ER-negative, PR-
negative) with a significantly worse survival than a node-positive
subgroup (ER-positive, PR-positive). This type of information
JUNE 1983 2987
Research.
on December 14, 2020. © 1983 American Association for Cancercancerres.aacrjournals.org Downloaded from
B. H. Mason et al.
_
/
PNS
o--*—o-—o-—o
—X 1 X X ._
12 » lì 21 2»
MONTHS AFTCR DIAGNOSIS
30 33 36
Chart 2. Actuarial analysis of disease-free survival and overall survival of breast
cancer patients according to the presence of tumor ERs and axillary nodal status.
, node-negative tumors (x, ER-positive, n = 130; O, ER-negative, n = 94);
, node-positive tumors (x, ER-positive, n = 91 ; O, ER-negative, n = 59).
could be of considerable importance when planning strategies
for adjuvant therapy in high-risk groups following mastectomy.
From the present data, it could be questioned whether analysis
of PR provides useful information on early breast cancer over
and above that obtained with ER assays. In general, either
receptor when used alone or together with nodal status has
provided similar data (Charts 2 and 3). However, the combination
of ER and PR results appears to provide useful additional infor
mation in a subset of patients (Chart 5), although whether this
information will provide clinically usable patient stratification
awaits formal testing in suitable prospective trials of adjuvant
therapy.
In this type of analysis, it is obviously of importance to consider
what tissue level of receptor should be considered biologically
significant. The present ER assay has been validated against
response of patients with advanced disease to endocrine ther
apy, and an ER level of 5*5 fmol/mg appears biologically "posi
tive" (14). It is, however, more difficult to describe a similar
biologically significant PR level. We have chosen to classify any
PR level of 1 or more fmol/mg as "positive" (Charts 1 and 3-5).
If, however, positive levels are redefined as »3or »5fmol/mg,
the results are essentially unchanged, although there is a trend
to increasing significance of PR as a prognostic indicator at the
higher cutoff levels.
When assessing receptor levels as prognostic indicators in
breast cancer, it is important to consider other variables known
to influence prognosis which could relate to receptor status. A
number of such factors have been considered in the present
study (Table 1) and do not appear to have influenced the current
—X ».. -K
V—x--?-».
/ *•-*
"\s ---x
~o o--*—o——o-—o
TX
Chart 3. Actuarial analysis of disease-free survival and overall survival of breast
cancer patients according to the presence of tumor PRs and axillary nodal status.
, node-negative tumors (x, PR-positive, n = 118; O, PR-negative, n = 106);
, node-positive tumors (x, PR-positive, n = 74; O, PR-negative, n = 76).
Õ2 II 21 2< 27 30 33 36
MONTHS AFTER DIAGNOSIS
Chart 4. Actuarial analysis of disease-free survival of breast cancer patients
according to the presence of tumor ERs and PRs and axillary nodal status.
, node-negative tumors p, ER-posith/e/PR-positive, n = 86; O, ER-positive/
PR-negative, n = 44; •,ER-negative/PR-positìve, n = 32; x, ER-negative/PR-
negative, n = 62); , node-positive tumore p, ER-positive/PR-positive, n = 58;
O, ER-positive/PR-negative, n = 33; •,ER-negative/PR-positive, n = 16; x, ER-
negative/PR-negative, n = 43).
2988 CANCER RESEARCH VOL. 43
Research.
on December 14, 2020. © 1983 American Association for Cancercancerres.aacrjournals.org Downloaded from
Steroid Receptors and Prognosis in Breast Cancer
Table 2
Cox's life-table regression analysis of Interactionbetween time to first recurrence and receptor and nodal status
Variable8Axillary
nodes
PRC
ER
Nodes and PR
Nodes and ER
Nodes, PR. and ERNo.positive150
185
224No.
negative224
189
150x"30.34 3.02
0.95
32.94
31.34
33.15P<0.001
NS"
NSd.f.1 1
1
2
2
3xa
difference2.60
1.00
0.21PNS NS
NS
8 Independent variables.Note: dependent variable =
6 Based on log likelihood.
c PR positive = level of 3 fmol/mg or greater.
NS, not significant.
disease-freeinterval.
Table 3
Cox's life-table regression analysisof interaction betweenpatient survival and receptor and nodal status
Variable8Axillary
nodes
PRC
ER
Nodes and PR
Nodes and ER
Nodes, PR and ERNo.
positive150185
224No.
negative224
189150x"14.1810.24
8.84
23.29
23.65
27.70P«cO.001
<0.005
<0.01d.f.1 1
1
2
2
3x2
difference9.11
9.47
4.05P<0.01 <0.01
<0.05
8 Independent variables.Note: dependent variable =
6 Based on log likelihood.
c PR positive = level of 3 fmol/mg or greater.
survival.
"1 5 S 5—ï
MONTHS »rlCR DIAGNOSIS
Chart 5. Actuarial analysisof overall survivalof breast cancer patients according
to the presence of tumor ERs and PRs and axillary nodal status. Symbols and
patient numbers as for Chart 4.
results. Many reports of receptor status in early breast cancer
have failed to identify the patient group in whom receptor studies
were performed and to account for possible patient bias. In the
present series, the group under study represented only 39% of
the total population developing breast cancer over the survey
period. It does not, however, appear that bias was introduced
by this selection, although study of a greater proportion of the
total breast cancer group would have been preferable.
In the present study, an interaction was observed between
receptor status and tumor histological grade (Table 1). Tumor
grade measures the differentiation of tumor cells and, since the
breast is a hormone-responsive organ, it is very likely that the
presence or absence of hormone receptors in tumor cells also
reflects differentiation. It is of interest that although receptor
measurements and nodal status were both bound to be signifi
cant prognostic variables, both measurements are distributed
independently of one another within the patient group (Table 1).
These findings are similar to those already reported (1, 4). This
suggests that the 2 variables reflect different biological mecha
nisms in patient survival, and it could thus be desirable to
combine both types of prognostic measure when considering
patient management.
ACKNOWLEDGMENTS
We thank the Auckland surgeons and pathologists for their contribution to this
study and H. Cook for secretarial assistance.
REFERENCES
1. Blarney, R. W . Bishop, H. M.. Blake, J. R. S.. Doyle, P. J., Elston, C. w .
Haybittie.J. L, Nicholson, R. I., and Griffiths, K. Relationship between primary
breast tumour receptor status and patient survival. Cancer (Phila). 46 (Suppl )
2765-2769,1980.
2. Bloom, H. J. G.. and Richardson, W. W. Histological grading and prognosis in
breast cancer. Br. J. Cancer, 11:359-377,1957.
3. Cooke, T., George, W. D . and Griffiths, K. Possible tests for selection of
adjuvant systemic therapy in eariycancerof the breast. Br. J. Surg.,67: 747-
750,1980.
4. Cooke, T., George, D . Shields, R., Maynard, P., and Griffiths, K. Oestrogen
receptors and prognosis in eariy breast cancer. Lancet, 1:995-997,1979.
5. Cox, D. R. Regression models and life tables (with discussion).J. R. Stat. Soc.
Part B, 34: 182-220, 1972.
6. Croton, R., Cooke, T., Holt, S., George, W. D., Nicholson, R., and Griffiths, K.
Oestrogen receptors and survival in breast cancer. Br. Med. J., 283: 1289-
1291,1981.
7. Fisher, B., Slack, N., Katrych, D., and Wolmark, N. Ten year follow-up results
of patientswith carcinomaof the breast ina co-operativeclinicaltrial evaluating
surgicaladjuvantchemotherapy.Surg. Gynecol Obstet., 740:528-534,1975.
8. Furmanski, P.,Saunders,D. E., Brooks, S. C., and Rich, M.A. The prognostic
value of oestrogen receptor determinations in patients with primary breast
cancer. Cancer (Phila.),46: 2794-2796,1980.
9. Gehan, E. A. A generalized Wilcoxon test for comparing arbitrarily singly-
censored samples. Biometria, 527:203-222,1965.
10. Gorski, C. M., Niepolomska. W., Nowak, K., Gebe!, B., Plewa. T., Pysz. H.,
and Adamus. J. Clinical evaluationand pathological grading in relation to other
prognostic factors In: A. P. M. Forrest and P. B. Kinkier (eds.). Prognostic
Factors in Breast Cancer, pp. 311-312. Edinburgh: E & S Livingstone, Ltd.,
1968.
11. Hannei. R., Woodings, T., and Vivian, A. B. Prognostic value of oestrogen
receptors in breast cancer. Cancer (Phila.),44: 671-675,1979.
12. Hilf, R., Feldstein, M., Gibson, S. L., and Saviov. E. D. The relative importance
of oestrogen receptor analysis as a prognostic factor for recurrence or re
sponse to chemotherapy in women with breast cancer. Cancer (Phila.), 45:
1993-2000,1980.
13. Hoidaway, I. M., and Mountjoy, K. G. Progesterone and oestrogen receptors
in human breast cancer. Aust. N. Z. J. Med., 8: 630-638, 1978.
JUNE 1983 2989
Research.
on December 14, 2020. © 1983 American Association for Cancercancerres.aacrjournals.org Downloaded from
S. H. Mason et al.
14. Hoidaway. I.M., Mountjoy. K. G., Harvey.V. J., Aten, E. P., andStephens,E. after mastectomy for breast cancer. Cancer¡Phiia.),41:2088-2098,1978.
J. Clinical applications of receptor measurements in breast cancer. Br. J. 19. Pichón,M., Pallud, C., Brunei, M., and Miigrom, E. Relationship of presence
Cancer,40:136-139,1980. of progesterone receptors to prognosis in early breast cancer. Cancer Res.,
15. Korn.W. H. Morphologic and clinical aspects of ER in carcinoma of the breast. 40: 3357-3360,1980.
Surg Gynecol Obstet. 148: 240-242,1979. 20. Samaan, N. A., Buzdar. A. U.. Aldmger, K. A., Schultz, P. N., Yang, K. P.,
16. Knight, W. A., Livingstone, R. B . and Gregory, E. J. Predicting risk of Romsdahl. M. M., and Martin, R. Oestrogen receptor a prognostic factor in
recurrence after surgery. Cancer Res., 37: 4669-4671,1977. breast cancer. Cancer (Phila). 47:554-560,1981.
17. Merrell. M., and Shulman, L. E. Determination of prognosis in chronic disease 21. Shapiro, C. M., Schifeiing.D., Bitran,J. D., Oesser,R. K. Rochman, H., Michel,
illustrated by systemic lupus erythematosus. J. Chronic Dis., 12-32,1955. A., Shapiro, R., Evans, R., Kozloff, M. F., Recant, W., and Billings, A. A.
18. Nissen-Meyer, R., KjeHgren,K., Maimio, K., Mansson.B.,and Norin, R. Surgical Prognostic value of the oestrogen receptor level in pathologic stage 1 and 2
adjuvant chemotherapy:results with one shortcourse with cydophosphamide adenocarcinomaof the breast. J. Surg. Oncol.. 79:119-121,1982.
2990 CANCER RESEARCH VOL. 43
Research.
on December 14, 2020. © 1983 American Association for Cancercancerres.aacrjournals.org Downloaded from
1983;43:2985-2990. Cancer Res
Barbara H. Mason, Ian M. Holdaway, Peter R. Mullins, et al.
in Breast Cancer
Progesterone and Estrogen Receptors as Prognostic Variables
Updated version
http://cancerres.aacrjournals.org/content/43/6/2985
Access the most recent version of this article at:
E-mail alerts related to this article or journal.Sign up to receive free email-alerts
Subscriptions
Reprints and
.pubs@aacr.orgDepartment at
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Permissions
Rightslink site.
Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)
.http://cancerres.aacrjournals.org/content/43/6/2985
To request permission to re-use all or part of this article, use this link
Research.
on December 14, 2020. © 1983 American Association for Cancercancerres.aacrjournals.org Downloaded from