REPORTS

vival analyses. Time-to-relapse com- therapy in patients with axillary lymph

parisons and survival comparisons for node-positive breast carcinoma. These tri-
Postchemotherapy Adjuvant women in the two treatment groups als were designed to test 1) the value of a
were made by use of the Kaplan-Meier doxorubicin-containing regimen in pre-
Tamoxifen Therapy Beyond
method and the logrank test Reported menopausal patients (E5181) and treat-
Five Years in Patients P values are two-sided. Results: Five ment duration in postmenopausal patients
With Lymph Node-Positive years after the randomization, no (E4181) and 2) the value of 5 years ver-
Breast Cancer statistically significant differences were sus 1 year of postoperative tamoxifen

Downloaded from https://academic.oup.com/jnci/article/88/24/1828/890199 by guest on 04 August 2024

noted in either time to relapse or sur- therapy. In the trial involving pre-
Douglass C. Tormey, Robert vival between women continuing to menopausal patients, all patients were
receive tamoxifen and those on obser- randomly assigned to one of two chemo-
Gray, Hendre C. Falkson* vation. Eighty-five percent of the therapy regimens plus tamoxifen for 1
women receiving tamoxifen were dis- year. At the end of that year, the patients
For the Eastern Cooperative ease free at this time compared with were randomly reassigned either to stop
Oncology Group 73% of those on observation (P = .10); all therapy or to continue tamoxifen treat-
survival was 86% for those continuing ment for a total of 5 years. The patients in
to receive tamoxifen and 89% for those the trial involving postmenopausal
on observation (P = .52). Differences in women also received tamoxifen with their
Background: Data from a pilot study
the time to relapse and survival be- combination chemotherapy; however,
published in 1984 suggested that
tween premenopausal and post- they were randomly assigned postopera-
tamoxifen administration (as adjuvant
menopausal women assigned to the two tively either to stop all therapy after 4 or
hormonal therapy) for more than 5
treatment groups were also not statisti- 12 months or to continue tamoxifen treat-
years after initial breast cancer surgery
cally significant (time to relapse: P = ment through 5 years after completing the
might have therapeutic benefit. Pur-
3 8 and P = .16 for premenopausal and first 12 months of combination therapy.
pose: A randomized trial was per-
postmenopausal patients, respectively; Results from these trials were pre-
formed to assess the efficacy of
survival: P = .18 and P = .72 for pre- viously reported at median follow-up
maintaining tamoxifen therapy beyond
menopausal and postmenopausal pa- times of 5.1 (E5181) and 4.1 (E4181)
5 years in women with axillary lymph
tients, respectively). There was an years {1,2). The median follow-up in the
node-positive breast cancer who had
indication that women with estrogen trial involving postmenopausal patients is
been treated with surgery followed by
receptor-positive tumors may ex- now 10.8 years, and the three-way com-
1 year of chemotherapy and 5 years of
perience a longer time to relapse with parison continues to demonstrate a trend
tamoxifen. Methods: One hundred
continued tamoxifen therapy (P = in disease-free survival (DFS) favoring
ninety-four women (87 postmenopausal
.014); however, the survival difference women receiving chemotherapy plus hor-
and 107 premenopausal) enrolled in
for this subgroup was not statistically mone therapy for 1 year followed by
two concurrent Eastern Cooperative
significant (P = .81). The toxicity pat- maintenance tamoxifen therapy through 5
Oncology Group adjuvant trials
terns in the two treatment groups were years in comparison with women receiv-
(E4181 for postmenopausal patients
similar. Conclusions and Implications: ing treatment for 4 or 12 months (P =
and E5181 for premenopausal patients)
Our results suggest that further evalu- .08). However, no overall survival ad-
were randomly assigned to continued
ation of adjuvant tamoxifen therapy vantage has been found (Falkson HC,
tamoxifen therapy or observation. Data
beyond 5 years in women with axillary Gray R, Tormey DC: unpublished obser-
for 193 women (87 postmenopausal
lymph node-positive, estrogen recep- vations). The median follow-up in the
and 106 premenopausal) were avail-
tor-positive breast cancer who have
able for analysis. Median follow-up is
also been treated with adjuvant
5.6 years since the randomization at 5
chemotherapy would be appropriate.
years, with the longest follow-up being
[J Natl Cancer Inst 1996;88:1828- 'Affiliations of authors: D. C. Tormey, AMC
8.0 years. The major analyses mea-
33] Cancer Research Center, Denver, CO, R. Gray,
sured events from the time of ran- Dana-Farber Cancer Institute, Boston, MA; H. C.
domization until relapse or death; Falkson, University of Pretoria, South Africa.
these included time-to-relapse analyses, In 1982, the Eastern Cooperative On- Correspondence to: Douglass C. Tormey, M.D.,
Ph.D., AMC Cancer Research Center, 1600 Pierce
with new opposite-breast cancers cology Group (ECOG) initiated two trials SL, Denver, CO 80214.
counted as treatment failures, and sur- of adjuvant chemotherapy plus hormone See "Notes" following "References."

1828 REPORTS Journal of the National Cancer Institute, Vol. 88, No. 24, December 18, 1996
trial involving premenopausal patients is stratified on the basis of induction therapy regimen Patient Evaluation
now 10.7 years, and patients receiving the for the second and third randomizations either to
continue or to discontinue tamoxifen at years 1 and Patient evaluations during the first postoperative
doxorubicin-containing regimen continue 5. Postmenopausal patients (E4181) were again year and the subsequent 4 years on tamoxifen were
to show a significant DFS advantage (P = stratified on the basis of initial lymph node status detailed previously (1.2). The evaluation schedule
.0017), but they now also show an overall and ER status before random assignment at year 5. for the patients randomly assigned at 5 years in-
Treatments were assigned by use of permuted cluded a clinic visit with blood chemistry and
survival advantage (P = .031). The results
blocks within strata, with a block size of 2 (5). This hematology studies done every 3 months, a chest x
of this trial also continue to show an ad- ray every 6 months, and a bone scan and a mam-
information was not available to the institutions
vantage with tamoxifen maintenance (P - while the study was under way. Dynamic balancing mographic examination every 12 months.
.0082) for DFS but not for overall survival (5) was used to ensure reasonable treatment balance
(Tormey DC, Gray R, Falkson HC: un- within institutions, where each "institution" in- Patient Distribution
published observations). cluded several distinct clinics. Treatment assign-
ments for the South African patients were made via A total of 87 postmenopausal and 107 pre-
Shortly after the trials described above sealed envelopes and stratified only on the basis of menopausal patients were randomly assigned at 5
were initiated, data suggesting that there study because of the limited number of entries that years either to continue tamoxifen treatment or to
could be potential benefit from continu- were expected. There were five South African observation only. Data from one premenopausal
patients from E4181 (two assigned to observation patient were excluded from all analyses because of a
ing tamoxifen treatment for more pro- recurrence of cancer prior to the randomization.
and three assigned to tamoxifen therapy) and seven
longed periods of time became available

Downloaded from https://academic.oup.com/jnci/article/88/24/1828/890199 by guest on 04 August 2024

from E5181 (three assigned to observation and four Thus, data from a total of 193 patients were avail-
from a pilot study (3,4). Accordingly, the assigned to tamoxifen therapy). able for analysis. This patient population represents
designs of the trials were altered to in- 69% of the premenopausal (E5181) and 64% of the
postmenopausal (E4181) patients completing 5
clude the random reassignment of pa-
Treatment Regimens years of tamoxifen therapy on the parent protocols.
tients who had completed 1 year of It is difficult to compare reliably the patients who
chemotherapy plus 5 years of tamoxifen The randomization at year 5 for all patients was were not randomly assigned with those who were;
treatment and were disease free either to either to continue treatment with tamoxifen (10 mg however, the characteristics of the two groups were
twice daily) until relapse or to discontinue tamox- similar, i.e., 26% of the patients who were not ran-
discontinuation of tamoxifen or to con-
ifen therapy. domly assigned continued tamoxifen treatment, and
tinuation of the drug until relapse. This Previous treatment for premenopausal patients the time to relapse (TTR) for this cohort was similar
report focuses on findings from this latter consisted of random assignment to 12 cycles of to that of the randomly assigned cohort (P = .53).
randomization. either CMFPT or ALTER induction regimens. The The median follow-up for this report is 5.6 years
CMFPT regimen consisted of 28-day cycles of after the randomization at 5 years, with the longest
cyclophosphamide (100 mg/m2 orally on days 1 follow-up being 8.0 years. The median duration of
Patients and Methods through 14), methotrexate (40 mg/m intravenously tamoxifen therapy from the start of induction treat-
[IV] on days 1 and 8), fluorouracil (600 mg/m2 IV ment for those who were randomly assigned to con-
Patient Eligibility on days 1 and 8), prednisone (40 mg/m2 orally on tinue tamoxifen is 9.3 years, with the longest being
Premenopausal and postmenopausal women who days 1 through 14), and tamoxifen (10 mg orally 13.1 years.
were disease free after treatment with 1 year of twice per day). The ALTER regimen consisted of
chemotherapy plus tamoxifen and 4 additional years the 28-day CMFPT regimen with the addition of
fluoxymesterone (10 mg orally twice per day) in
Statistical Methods
of tamoxifen on ECOG protocols E4181 and E5181
were eligible for random assignment. Criteria for cycles 1, 3, 5, 7, 9, and 11, with prednisone omitted The major analyses measured events from the
entry into the parent E4181 and E5181 studies were after cycle 3; the even-numbered cycles consisted of time of random assignment at 5 years until relapse
detailed previously {1,2). Briefly, women were in- 22-day courses of vinblastine (4.5 mg/m IV on day or death, respectively, for the 193 assessable
cluded if they had infiltrating carcinomas pathologi- 1), doxorubicin (45 mg/m IV on day 1), thiotepa patients. Analyses of the TTR used the date of the
cally less than or equal to 5 cm in diameter, one or (12 mg/m IV on day 1), fluoxymesterone (10 mg first documented evidence of relapse, judged by pre-
more histopathologically involved ipsilateral axil- orally twice per day), and tamoxifen (10 mg orally viously published guidelines (6), as the date of
lary lymph nodes; a known estrogen receptor (ER) twice per day) (also known as the VATHT relapse and included new opposite-breast primary
assay; normal hematologic function and biochemi- regimen, where A = Adriamycin [doxorubicin] tumors as treatment failures; patients who died
cal profiles; a normal bone scan; definitive surgery and H = Halotestin [fluoxymesterone]). After 12 without disease recurrence were censored in these
performed within the preceding 6 weeks; and writ- cycles of induction therapy, the patients were ran- analyses at the date of their death. Separate DFS
ten informed consent according to Department of domly assigned either to continue tamoxifen or to comparisons were also done, with disease-unrelated
Health and Human Services, state, and institutional observation for an additional 4 years. Patients who deaths counted as treatment failures. The results ob-
guidelines. In these trials, being premenopausal was were disease free on tamoxifen 5 years after study tained from these DFS comparisons were similar to
defined as having one or more menses in the 12 entry were again randomly assigned either to ob- those obtained from the major analyses. Survival
months before diagnosis or being under 52 years of servation only or to continue tamoxifen therapy analyses included deaths from all causes.
age and having no menopausal symptoms in patients until relapse.
Fisher's exact test (7) was used to compare the
with a prior hysterectomy. All other patients were Previous treatment for postmenopausal patients distribution of categorical prognostic factors be-
defined as being postmenopausal. consisted of random assignment to induction therapy tween treatments. The method of Kaplan and
with four cycles of CMFPT, 12 cycles of CMFPT, Meier (8) was used to estimate TTR and survival
or 12 cycles of CMFPT followed by continuation of curves; logrank tests were used to evaluate the
Randomization
tamoxifen therapy through 5 years. Patients in this equality of the TTR and survival curves (9). When
Random assignments to treatment for North latter treatment arm who were disease free on data from both E4181 and E5181 were included in
American patients were made via telephone calls to tamoxifen 5 years after study entry were again ran- a comparison, the test was stratified on the basis
a central office. These patients were stratified for in- domly assigned either to observation only or to con- of study. The toxicity analyses were based on all
duction therapy in their original protocols on the tinue tamoxifen therapy until relapse. analyzed entries for each toxic effect. All P values
basis of having one to three versus four to 10 versus Treatment administration and dose modifica- are from two-sided tests, and the term "sig-
more than 10 positive axillary lymph nodes and ac- tions for the induction regimens have been de- nificant" refers to two-sided P values less than or
cording to ER-positive (£10 fmol/mg protein) ver- scribed in detail previously (1,2). There were no equal to .05. There was no accrual goal for this ex-
sus ER-negative (<10 fmol/mg protein) status. dose modifications specified for the administration ploratory study, since the sample sizes were deter-
Premenopausal patients (E5181) were additionally of tamoxifen. mined by objectives from earlier steps; therefore,

Journal of the National Cancer Institute, Vol. 88, No. 24, December 18, 1996 REPORTS 1829
 the number of patients available was fixed by other Table 1. Characteristics of patients randomly assigned after 5 years of tamoxifen merapy to continued
considerations. tamoxifen treatment or to observation only

No. of patients (%)

Results
Tamoxifen Observation
Characteristic (n= 100) (n = 93)
Patient Characteristics
Menopausal status*
The treatment regimens (tamoxifen Premenopausal 57 (57) 49 (53)
versus observation) were well balanced Postmenopausal 43(43) 44(47)
with respect to the stratification charac- Induction regimen
teristics used (Table 1). The major patient CMFPTt 68(68) 66(71)
ALTER* 32 (32) 27(29)
characteristics were premenopausal
No. of axillary lymph nodes positive
status, prior treatment with CMFPT in- 1-3 60(60) 58 (62)
duction therapy, one to three positive 4-10 32 (32) 31 (33)
axillary lymph nodes, and ER-positive >10 8(8) 4(4)
tumor status. Estrogen receptor status §

Downloaded from https://academic.oup.com/jnci/article/88/24/1828/890199 by guest on 04 August 2024

Positive 73(73) 67 (72)
Negative 27 (27) 26(28)
TTR Treatment Comparisons
•Premenopausal patients were enrolled in Eastern Cooperative Oncology Group (ECOG) trial E5181, and
The proportions of patients who were postmenopausal patients were enrolled in ECOG trial E4181 (see "Patients and Methods" section for
disease free at 5 years after the ran- details).
domization at 5 years were 85% for those fCMFPT = cyclophosphamide, methotrexate, fluorouracil, prednisone, arid tamoxifen (see "Patients and
Methods" section for details).
continuing to receive tamoxifen and 73% tALTER = cycles of CMFPT and fluoxymesterone (Halotestin; Upjohn Co., Kalamazoo, MI) alternating
for those on observation (P = .10; Fig. 1, with cycles of vinblastine, doxorubicin (Adnamycin; Adria Laboratories, Columbus, OH), thiotepa,
A). This trend was similar for pre- fluoxymesterone, and tamoxifen; prednisone was omitted after cycle 3 (see "Patients and Methods" section
for details).
menopausal patients (P - .38; Fig. 1, B), §Estrogen receptor positive was defined as 210 fmol/mg protein; estrogen receptor negative was defined
postmenopausal patients (P = .16; Fig. 1, as < 10 fmol/mg protein.
C), and patients with ER-positive tumors
(P = .014; Fig. 2). Among patients with
ER-negative tumors, there has been only seven local-regional only, 14 distant, and trointestinal tract, ovary, and endo-
one relapse among 26 patients on obser- two mixed. metrium for the patients on observation;
vation compared with three relapses There have been eight deaths without and 2) ovary (two patients) and pancreas
among 27 patients receiving tamoxifen the recurrence of breast cancer among all for the patients receiving tamoxifen.
(P=.24). patients (six deaths among those assigned The toxicity patterns observed in both
to receive tamoxifen and two among treatment groups were similar. Moderate
Survival Treatment Comparisons patients on observation). Both deaths in toxicity events occurred in eight patients
the observation group involved post- on observation and in nine patients
Overall survival at 5 years was 86% for menopausal patients and were due to receiving tamoxifen. These events in the
those receiving tamoxifen and 89% for chronic obstructive pulmonary disease observation group were as follows: pain,
those on observation; the difference in (one patient) and congestive heart failure central neurologic toxicity, liver toxicity
these values was not statistically sig- (one patient). Among the patients receiv- (three patients), central neurologic
nificant (P = .52; Fig. 3). Similarly, the ing tamoxifen, three of the deaths in- toxicity plus weight gain, hypertension
trends were not significantly different for volved premenopausal patients; two of plus infection plus skin toxicity plus
premenopausal patients (P = .18), these deaths were caused by ovarian car- pain, and hyperglycemia; in the tamox-
postmenopausal patients (P = .72), cinoma, and the cause of the third death ifen treatment group, these were nausea,
patients with ER-positive tumors (P = was unknown, although the latter patient infection (three patients), liver toxicity
.81), or patients with ER-negative tumors had been disease free 2 months earlier. (two patients), hypertension, pain, and
The three deaths among postmenopausal skin toxicity. Severe or life-threatening
patients assigned to receive tamoxifen events occurred in four patients on obser-
Sites of First Relapse and Causes of
were due to cirrhosis (one patient), vation (liver toxicity, pain, hypoglycemia,
Death
pancreatic carcinoma (one patient), and and infection) and in four patients receiv-
The distributions of sites of first re- cardiac fibrosis (one patient). ing tamoxifen (two deep vein throm-
lapse for patients assigned to the two boses, central neurologic toxicity, and
Side Effects During Treatment edema).
treatment groups were similar. The num-
bers and sites of relapse for patients There have been reports of second can-
receiving tamoxifen were as follows: four cers involving four patients on observa- Discussion
local-regional only, eight distant, and tion and three patients receiving
three mixed; for patients on observation, tamoxifen. The sites of these cancers The trial described above was de-
the numbers and sites were as follows: were as follows: 1) left eye in situ, gas- veloped to enlarge previous experience

1830 REPORTS Journal of the National Cancer Institute, Vol. 88, No. 24, December 18, 1996
 A 1.0 -

0.8 -

c 0.6 -
o
•a
Propon

0.4 -

0.2 -
Obs (23 relapses/93 patients)
Tam (15 relapses/100 patients)
0.0 -

0 2 4 6 8

Years

Downloaded from https://academic.oup.com/jnci/article/88/24/1828/890199 by guest on 04 August 2024

# at risk
Obs 93 80 59 26 7
Tam 100 91 69 34 4

B 1.0 -

08 -
Propo rtlon

0.6 -
Fig. 1. Time to relapse curves according to treatment for all
patients (A), for premenopausal patients (B), and for
0.4 -
postmenopausal patients (C). The total number of
P = .38 events/number of patients is indicated for each treatment
along with the comparative P value (two-sided; logrank test).
0.2 - The number of patients at risk with each treatment is dis-
Obs (9 relapses/49 patients)
played at 2-year intervals below the graph. Obs = patients on
Tam (7 relapses/ 57 patients)
observation; Tam = patients receiving tamoxifen.
0.0 -

0 2 4 6 8

Years
* at risk
Obs 49 43 29 14 4
Tam 57 51 37 18 4

c 1.0 -

0.8 -
ProponDon

0.8 -

0.4 -

P = .16
0.2 -
Obs (14 relapses/44 patients)
Tam (8 relapses/ 43 patients)
0.0 -

0 2 4 6 8

Years
* at risk
Obs 44 37 30 12 3
Tam 43 40 32 16 0

Journal of the National Cancer Institute, Vol. 88, No. 24, December 18, 1996 REPORTS 1831
 which to view this information is to con-
1.0 -
sider the average hazard for recurrences
for patients with ER-positive status/per-
0.8 - son-year. This hazard from our historical
database (70) for patients with node-posi-
o 0.6 - tive disease and ER-positive tumors with
c more than 5 years follow-up is .065
(standard error [SE] = .005), which is
A" 0.4 A
comparable to that of the observation
P = .O14 group in the present study (.075; SE =
0.2 .016). However, the hazard for the
Obs (22 relapses/67 patients)
Tam (12 relapses/ 73 patients) tamoxifen-treated patients in the present
0.0 - study was significantly lower, at .033 (SE
= .010). Since this latter finding emerged
4 6
from a subset analysis in the present
Years study, it must be considered with caution

Downloaded from https://academic.oup.com/jnci/article/88/24/1828/890199 by guest on 04 August 2024

* at risk
37 16
and requires confirmation in future trials.
Obs 67 56
Recently published findings from the
Tam 73 69 53 25
National Surgical Adjuvant Breast and
Bowel Project (NSABP) B-14 trial (77)
Fig. 2. Time to relapse curves according to treatment for all patients with estrogen receptor-positive tumors. and the Scottish Adjuvant Tamoxifen
The total number of events/number of pauents is indicated for each treatment along with the comparative P
value (two-sided; logrank test). The number of patients at risk with each treatment is displayed at 2-year in- Trial (72), predominantly involving
tervals below the graph. Obs = patients on observation; Tam = patients receiving tamoxifen. patients with node-negative disease, have
suggested that the administration of
(3,4) with tamoxifen given for more than analyses between the treatment ap- tamoxifen beyond 5 years without induc-
5 years and to do so in a randomized en- proaches for either TTR or overall sur- tion chemotherapy is unlikely to provide
vironment. It was recognized at the trial's vival. In addition, there were no a therapeutic benefit. In fact, the data sug-
inception that the number of patients significant differences observed in toxic gest that it may be detrimental. Clearly,
entered would not allow detection of a side effects or second cancers. as is the case with any other drug,
therapeutic effect unless there was a It is of interest that the TTR for tamoxifen could have long-term side ef-
major difference in outcome. It is there- patients with ER-positive tumors was sig- fects that would pose a therapeutic hazard
fore not surprising that no significant dif- nificantly in favor of those receiving that overrides at least some of the poten-
ferences were observed in the main tamoxifen. An alternative manner in tial benefit of continued therapy. If
tamoxifen does have a detrimental effect
after 5 years of therapy, the associated
hazard rate needs to be defined. This in-
1.0 - formation would facilitate an assessment
of whether continued tamoxifen therapy
might have a theoretically beneficial
0.8 -
therapeutic ratio for higher-risk patient
cohorts. For example, ECOG has found
Proportion

0.6 - the 5- to 10-year annual hazard rates for

recurrence of breast carcinoma, or,
0.4 - equivalently, the percent failure per per-
P = .52 son-year, to range from 0.4 to 1.9 for
0.2 -
patients with node-negative disease, from
Obs (10 deaths/93 patients) 3.4 to 6.0 for patients with one to three
Tam (14 deaths/100 patients)
positive nodes, and from 5.7 to 10.0 for
0.0 -
patients with more than three positive
0 2 4 6 8 nodes, irrespective of prior therapy (70).
Years
Conceivably, therefore, tamoxifen could
# at risk have a negative therapeutic ratio for
Obs 93 88 75 31 7 patients with node-negative disease and
Tam 100 97 77 41 5 still have a beneficial therapeutic ratio for
higher-risk cohorts of patients who have
Fig. 3. Overall survival according to treatment for all patients. The total number of events/number of patients node-positive disease. Alternatively, it is
is indicated for each treatment along with the comparative P value (two-sided; logrank test). The number of possible that tamoxifen administration
patients at risk with each treatment is displayed at 2-year intervals below the graph. Obs = patients on obser-
vation; Tam = patients receiving tamoxifen. has no further therapeutic benefit after 5

1832 REPORTS Journal of the National Cancer Institute, Vol. 88, No. 24, December 18, 1996
years, in which case additional treatment 73% with observation. It is also estimated (4) Tormey DC, Rasmussen P, Jordan VC. Long-
term adjuvant tamoxifen study: clinical update
would be without benefit for the control that the comparison of the treatments with [letter]. Breast Cancer Res Treat 1987;9:157-
of breast carcinoma. In addition, the use regard to overall survival would have a 8.
of induction chemotherapy in our trial power of 80% for detecting a hazards (5) Zelen M. The randomization and stratification
of patients to clinical trials. J Chronic Dis
could be hypothesized to reduce the ratio of 3.1, which corresponds to 5-year 1974;27:365-75.
tumor burden and to provide a more survival proportions of 96% and 87%, (6) Tormey D, Fisher B, Bonadonna G, Carter S,
favorable setting for long-term tamoxifen respectively. Given the size of this trial, Davis L, Glidewell O, et al. Proposed guide-
lines—report from the Combined Modality
therapy to be effective. As such, the we do not expect further follow-up to im- Trials Working Group in breast cancer. In:
NSABP results in patients exclusively prove substantially our ability to detect Carbone PP, Sears ME, editors. Suggested
with node-negative disease and our differences between the treatments in protocol guidelines for combination chemo-
therapy trials and for combined modality trials.
results in patients exclusively with node- terms of either overall survival or TTR. Bethesda (MD): DHEW Publ No. (NIH)77-
positive disease may be equally valid be- Clearly, the analysis of results from other 1192, 1977:20-35.
cause of differences in either therapy, current or planned studies must be under- (7) Mehta CR, Patel NR. A network algorithm for
performing Fisher's exact test in rxc contingen-
patient populations, or other unknown taken to settle the issue of the optimal cy tables. J Am Stat Assoc 1983;78:427-34.
variables. Irrespective of the biologic duration of treatment with tamoxifen for (8) Kaplan EL, Meier P. Nonparametric estima-
hypothesis generated to explain potential appropriate subgroups of patients. tion from incomplete observations. J Am Stat

Downloaded from https://academic.oup.com/jnci/article/88/24/1828/890199 by guest on 04 August 2024

Assoc 1958;53:457-81.
differences in the results from cohorts of In summary, this report provides (9) Mantel N. Evaluation of survival data and two
patients with predominantly node-nega- evidence that it appears to be reasonable new rank order statistics arising in its con-
tive disease and results from patients with to continue to evaluate the use of tamox- sideration. Cancer Chemother Rep 1966;50:
163-70.
node-positive disease, it should be pos- ifen therapy beyond 5 years in patients (10) Saphner T, Tormey DC, Gray R. Annual
sible to define which alternative is correct with axillary lymph node-positive, ER- hazard rates of recurrence for breast cancer
by appropriate trials involving only positive breast carcinoma who have after primary therapy. J Clin Oncol 1996; 14.
2738^6.
patients with higher-risk, node-positive received induction chemotherapy. (//) Fisher B, Dignam J, Bryant J, DeCillis A,
disease. Wickerham DL, Wolmark N, et al. Five versus
Although our trial was exploratory in more than five years of tamoxifen therapy for
nature, the data do not at this time support References breast cancer patients wth negative lymph
nodes and estrogen receptor-positive tumors. J
a potential therapeutic or adverse-events Natl Cancer Inst 1996;88:1529-42.
(/) Tormey DC, Gray R, Abeloff MD, Roseman (12) Stewart HJ, Forrest AP, Everington D, Mc-
disadvantage for long-term tamoxifen use DL, Gilchrist KW, Barylak EJ, et al. Adjuvant Donald CC, Dewar JA, Hawkins RA, et al.
among patients with node-positive dis- therapy with a doxorubicin regimen and long- Randomised comparison of 5 years of adjuvant
ease. Additionally, the significant thera- term tamoxifen in premenopausal breast can- tamoxifen with continuous therapy for operable
cer patients: an Eastern Cooperative Oncology breast cancer. The Scottish Cancer Trials Breast
peutic advantage observed in patients with Group trial. J Clin Oncol 1992; 10:1848-56. Group. Br J Cancer 1996;74:297-9.
ER-positive tumors must be cautiously (2) Falkson HC, Gray R, Wolberg WH, Gilchnst
interpreted because it was detected in a KW, Harris JE, Tormey DC, et al. Adjuvant
subset analysis, whereas the overall com-
trial of 12 cycles of CMFPT followed by ob- Notes
servation or continuous tamoxifen versus four
parison did not yield a significant result. cycles of CMFPT in postmenopausal women Conducted in part by the Eastern Cooperative On-
It is currently estimated that the com- with breast cancer: an Eastern Cooperative cology Group (R. L. Conns, Chairman).
Oncology Group phase III study [published er- Supported in part by Public Health Service grants
parison of the two treatments in our study ratum appears in J Clin Oncol 199O;8:I6O3], J CA21076, CA21692, CA23318, CA66636, and
would have a power of 80% for detecting Clin Oncol 1990;8:599-607. CA21115 from the National Cancer Institute, Na-
a hazards ratio of 2.5 (tamoxifen/observa- (3) Tormey DC, Jordan VC. Long-term tamoxifen tional Institutes of Health, Department of Health and
adjuvant therapy in node-positive breast can- Human Services.
tion), which corresponds to 5-year DFS cer: a metabolic and pilot clinical study. Breast Manuscript received September 26, 1996; revised
proportions of 88% with tamoxifen and Cancer Res Treat 1984;4:297-302. October 25, 1996; accepted November 13, 1996.

Journal of the National Cancer Institute, Vol. 88, No. 24, December 18, 1996 REPORTS 1833