A randomized placebo-controlled study of tamoxifen after adjuvant chemotherapy in premenopausal women with early breast cancer (National Cancer Institute of Canada—Clinical Trials Group Trial, MA.12)

Abstract

Background: In the early 1990s, the role of adjuvant tamoxifen in premenopausal women with early breast cancer (EBC) was not established. Similarly, optimum timing relative to adjuvant chemotherapy and efficacy of tamoxifen in hormone receptor-negative tumors were unclear.

Patients and methods: Premenopausal women with EBC, any hormone receptor status, after surgery received standard adjuvant chemotherapy [doxorubicin (adriamycin)/cyclophosphamide, cyclophosphamide/methotrexate/5-fluorouracil, or cyclophosphamide/epirubicin/5-fluorouracil] followed by randomization to tamoxifen or placebo for 5 years. Outcomes were overall survival (OS), disease-free survival (DFS), toxicity, and compliance with therapy.

Results: Median follow-up for 672 women was 9.7 years. Multivariate analysis showed improved DFS [78.2% versus 71.3% at 5 years; hazard ratio (HR) 0.77; P = 0.056] and a trend for improved OS (86.6% versus 82.1% at 5 years; HR 0.78; P = 0.12). There was no evidence of greater benefit for the receptor-positive subgroup. Compliance with treatment was suboptimal in both arms, with 103 (31%) women on tamoxifen and 70 (21%) on placebo-stopping therapy early because of toxicity, refusal, or other choices.

Conclusions: Adjuvant tamoxifen, given after chemotherapy to premenopausal women with EBC, improved 5-year DFS. Poor compliance may have reduced treatment efficacy.

introduction

In 1992, the Early Breast Cancer Trialists’ Collaborative Group (EBCTCG) published the 1990 meta-analysis of 75 000 women in randomized trials evaluating systemic adjuvant therapies for early breast cancer (EBC) [1]. This established the benefits of polychemotherapy alone, and tamoxifen alone, in reducing recurrence and improving survival. However, the meta-analysis indicated little advantage for tamoxifen added to polychemotherapy [6362 women—risk reduction (RR) recurrence 7%, RR mortality 3%] in women <50 years. This meta-analysis [1] also established 5 years of tamoxifen as an optimum duration, but its role in women with estrogen receptor (ER)-negative tumors remained unresolved.

Superiority of sequential rather than concurrent use of tamoxifen with chemotherapy was not confirmed until publication of Intergroup 0100 study results in 2002 [2]. However, a theoretical concern [3], together with experimental data [4], indicating that tamoxifen given concurrently with chemotherapy could decrease rates of cell division reducing susceptibility to cell cycle active chemotherapy agents, provided the basis for sequential use of tamoxifen in several adjuvant chemohormonal trials started in the 1990s. The interaction of chemotherapy and tamoxifen may be quite complex in premenopausal women. Chemotherapy-induced amenorrhea is often associated with better outcomes, although controversy remains [5]. In 1992, there was substantial uncertainty as to whether tamoxifen administered with or after chemotherapy in premenopausal women was redundant or conferred benefit similar to that seen in postmenopausal women.

Thus, in 1992, the National Cancer Institute of Canada—Clinical Trials Group (NCIC CTG) Breast Committee initiated Mammary (MA).12 in premenopausal women with node-positive breast cancer, evaluating the benefit of 5 years of adjuvant tamoxifen versus placebo following completion of adjuvant chemotherapy. Participants could opt for one of three well-established chemotherapy regimens, cyclophosphamide/methotrexate/5-fluorouracil (CMF), cyclophosphamide/epirubicin/5-fluorouracil (CEF) as used in an NCIC CTG trial that had just been completed [6], or doxorubicin (adriamycin)/cyclophosphamide (AC) [7].

patients and methods

patient population

Women with histologically confirmed breast cancer who had undergone complete or segmental mastectomy plus axillary node dissection were included. Pre-/perimenopausal status was characterized by one of the following: normal menstruation, amenorrhea <1 year, biochemical evidence of ovarian function, amenorrhea for 1–3 years in women <52 years, or hysterectomy without oophorectomy in women <56 years [criteria of the International Breast Cancer Study Group (IBCSG)]. Patients with pathological T1–4, N0–2, and M0 tumors were eligible. Initially, entry was restricted to women with node-positive disease but, from February 1995, women with high-risk node-negative tumors (tumor ≥1 cm and high histological grade or lymphovascular invasion or both) were included. Determination of levels of at least one hormone receptor [ER and/or progesterone receptor (PgR)], by biochemical (positive ≥10 fmol/mg protein) or immunohistochemical assay (as determined by the local pathologist), was required but patients with any receptor status were eligible.

Patients were excluded if they had distant metastases, residual disease in the breast or axilla, other serious medical illnesses, or a previous cancer. Women considering pregnancy or using hormones were excluded. Brief exposure to tamoxifen (≤2 weeks) was permitted.

treatment plan

The study was approved by local Research Ethics Boards, and patients provided written informed consent. Staging investigations included history, physical examination, complete blood count and liver function tests before chemotherapy, and chest radiograph and bone scan within 8 weeks of starting chemotherapy. Patients were required to start chemotherapy within 14 weeks of first histological diagnosis of breast cancer and to complete a course of standard adjuvant chemotherapy with six cycles of CMF or CEF [6] or four cycles of AC [7]. After stratification for type of chemotherapy (CMF versus CEF versus AC), hormone receptor status (ER and/or PgR positive versus ER and PgR negative), and nodal status (0 versus 1–3 versus 4–9 versus 10+), through the central office of NCIC CTG, women were randomized to tamoxifen/placebo on/after day 8 of cycle 6 of CMF/CEF or day 1 of cycle 4 of AC and were required to start study medication within 42 days of completing i.v. chemotherapy. Tamoxifen/placebo was provided by AstraZeneca, Missisauga, Ontario, as 20 mg tablets to be taken daily for 5 years.

follow-up

After starting tamoxifen/placebo, women were seen every 3 months for 2 years, every 6 months for 3 years, and then yearly. Adverse events/toxic effects of tamoxifen/placebo were recorded. When patients stopped blinded treatments, reasons were documented, including treatment completion, recurrence, death, intercurrent illness, toxicity of treatment, refusal (for reasons other than toxicity), major protocol violation (started another hormonal therapy before progression), and others. Other hormonal therapies (including nonstudy tamoxifen) taken before progression (during or after off-protocol treatment) were also recorded.

outcome measures

The primary end point was overall survival (OS), calculated from date of randomization to date of death or censored on the last day the patient was known to be alive. Disease-free survival (DFS), the secondary efficacy end point, was calculated from randomization to the earliest date of recurrence or death or censored on the last date the patient was known to be alive. Contralateral breast cancer was considered a second primary malignancy. The worse grade of adverse event/toxicity according to the NCIC CTG Expanded Common Toxicity Criteria was recorded for each patient at randomization (baseline), during therapy with tamoxifen/placebo (acute) and after completion of therapy (not reported, except second malignancies).

statistical analysis

The trial was designed to detect, with 80% power and a two-sided alpha of 5%, an increase in 5-year OS from 70% for women on placebo to 78% for those receiving tamoxifen [corresponding to a hazard ratio (HR) of 0.69]. This would require 242 deaths at final analysis. It was projected that this number of events would be observed by enrolling 800 patients over 5 years and following them for five additional years. The trial was activated in July 1993. Based on the recommendation of the Data Safety and Monitoring Committee (DSMC), the trial closed in April 2000 after 672 patients were randomized since, because of slow accrual, it was projected that 242 deaths would still be observed with 5 years’ additional follow-up.

Two interim analyses were scheduled when one-third (80) and two-thirds (160) of deaths required were observed. The DSMC recommended continuation after the first interim analysis. After reviewing results of the second interim analysis and a supplemental futility analysis, the DSMC accepted a proposal by the Trial Committee to publish these results.

OS and DFS of the two treatment groups were described by Kaplan–Meier curves and compared by a log-rank test adjusting stratification factors at randomization. The Cox proportional hazards model was used to adjust the observed treatment effect for the three stratification factors and four other potential prognostic factors: age (≤45 versus >45 years); Eastern Cooperative Oncology Group performance status (0 versus 1, 2); time from initial pathological diagnosis to randomization (≤6 versus >6 months); and pathological T stage (≤2 versus >2 cm). The chi-square test was used to compare rates of adverse events/toxicity between treatment groups. All statistical tests were two sided.

results

patient population

From 1993 to 2000, 672 women were randomized: 338 to tamoxifen and 334 to placebo. Median age was 46 years (range 29–58). Other characteristics are summarized in Table 1. ER assays were done by the biochemical method in 202 (60%) of women randomized to tamoxifen and 172 (51%) on placebo. The remaining ER assays were done by immunohistochemistry (one patient on each arm had both assays). All PgR assays were done by biochemistry. Hormone receptor status was positive (ER positive and/or PgR positive) in 505 (75%) women. Eighteen patients did not meet the eligibility criteria (Appendix, Figure A1). On tamoxifen, these included one incomplete resection, six incomplete screening tests, three inappropriate chemotherapy regimens, and three not meeting high-risk criteria for node-negative disease. On placebo, five patients had incomplete screening tests. All randomized patients were included in survival analyses on an intention-to-treat basis.

Table 1.

Patient characteristics according to treatment arms

Characteristic	Tamoxifen		Placebo		Total
	n	%	n	%	N	%
Receptor status
ER and/or PgR positive	252	74	253	76	505	75
ER and PgR negative	50	15	26	8	76	11
ER negative and PgR unknown	36	11	55	16	91	14
ER positive	223	66	231	69	454	68
ER negative	115	34	103	31	218	32
PgR positive	132	39	119	36	251	37
PgR negative	60	18	41	12	101	15
PgR unknown	146	43	174	52	320	48
Nodal status
Node negative	84	25	83	25	167	25
1–3 nodes	190	56	187	56	377	56
4–9 nodes	55	16	53	16	108	16
10+ nodes	9	3	11	3	20	3
Adjuvant chemotherapy
CEF	73	22	73	22	146	22
CMF	153	45	151	45	304	45
AC	112	33	110	33	222	33
Age
≤29	1	0	2	0	3	0
30–39	64	19	66	20	130	19
40–49	210	62	210	63	420	63
≥50	63	19	56	17	119	18
Stage (pathological)
I	33	10	37	11	70	10
II	284	84	278	83	562	84
III	21	6	19	6	40	6
Pathological T stage
1	140	41	149	45	289	43
2	179	53	165	49	344	51
3/4	19	6	20	6	39	6
ECOG status
0	234	69	206	62	440	65
1	100	30	125	37	225	34
2	4	1	3	1	7	1
Menstrual status (at randomization)
Regular menses	85	25	93	28	178	26
Irregular menses	36	11	39	12	75	11
No menses >3 months	129	38	121	36	250	37
Previous hysterectomy	17	5	20	6	37	6
No data	71	21	61	18	132	20

ER, estrogen receptor; PgR, progesterone receptor; CEF, cyclophosphamide/epirubicin/fluorouracil; CMF, cyclophosphamide/methotrexate/5-fluorouracil; AC, doxorubicin(adriamycin)/cyclophosphamide; ECOG, Eastern Cooperative Oncology Group.

outcomes—OS and DFS

At data cut-off for this analysis, March 2007, median follow-up was 9.7 years (range 0.2–13.1 years), and 169 women had died, 77 (23%) on tamoxifen and 92 (27.5%) on placebo (Table 2). The Kaplan–Meier curves for OS (Figure 1A) show 5-year OS rates of 86.6% on tamoxifen and 82.1% on placebo. The HR from the stratified Cox proportional hazards model was 0.78 [95% confidence interval (CI) 0.57–1.06; P = 0.12]. The effect of tamoxifen was explored in subgroups defined by overall hormone receptor and ER receptor-only status (Table 3). There was no evidence of greater efficacy for tamoxifen in the hormone receptor-positive or ER receptor-positive subgroups (P values of interaction tests 0.71 and 0.14, respectively).

Table 2.

Outcome events—deaths and recurrences by treatment arm

	Tamoxifen		Placebo
	n	%	n	%
Patients	338		334
OS—total deaths	77	23	92	28
Breast cancer	72		86
Other malignancy	2		4
Othera	3		2
DFS—total events	105	31	121	36
Recurrences	101	30	116	35
Locoregional only	17	5	14	4
Breast	11		10
Chest wall	2		1
Regional nodes	4		3
Distant only	51	15	57	17
Multiple sites	33	10	45	13
Boneb	59		69
Liver	45		54
Lung	40		53
Brain	24		27
Ascites/effusions	38		47
Other	30		34
Death without recurrence	4	1	5	1

^aTamoxifen: nonprotocol treatment complication (1), intracranial bleed (1), unknown, but extensive metastases (1); placebo: motor vehicle accident (1), suspected suicide (1).

^bSingle patient can have multiple sites of recurrence.

OS, overall survival; DFS, disease-free survival.

Figure 1.

(A) Kaplan–Meier curves for overall survival, tamoxifen versus placebo; (B) Kaplan–Meier curves for disease-free survival, tamoxifen versus placebo.

Table 3.

Effect of tamoxifen on OS and DFS according to hormone receptor subgroups

	Tamoxifen		Placebo
	Total patients	% Surviving 5 years	Total patients	% Surviving 5 years	HR (95% CI)
OS
Receptor status
ER and/or PgR positive	252	90	253	85	0.83 (0.58–1.18)
ER and PgR negative	50	76	26	73	0.78 (0.34–1.81)
ER negative, PgR unknown	36	78	55	72	0.67 (0.29–1.56)
ER status
Positive	223	90	231	86	0.94 (0.65–1.38)
Negative	115	81	103	72	0.61 (0.37–1.01)
DFS
Receptor status
ER and/or PgR positive	251	80	253	73	0.82 (0.61–1.11)
ER and PgR negative	50	70	26	69	0.83 (0.38–1.82)
ER negative, PgR unknown	36	78	55	65	0.65 (0.30–1.43)
ER status
Positive	222	79	231	75	0.90 (0.66–1.24)
Negative	115	77	103	64	0.65 (0.41–1.03)

OS, overall survival; DFS, disease-free survival; HR, hazard ratio; CI, confidence interval; ER, estrogen receptor; PgR, progesterone receptor.

For DFS, there were 226 recurrences or deaths without recurrence, 105 (31%) on tamoxifen and 121 (36%) on placebo (Table 2). The Kaplan–Meier curves for DFS (Figure 1B) show 5-year DFS rates of 78.2% on tamoxifen and 71.3% on placebo. The HR from the stratified Cox proportional hazards model was 0.77 (95% CI 0.59–1.01; P = 0.056). Table 3 shows that as for OS, there was no evidence for a greater effect of tamoxifen on DFS in the hormone receptor-positive or ER-positive subgroups (P values of interaction tests 0.87 and 0.26, respectively).

protocol treatment

Ten patients on tamoxifen and seven on placebo never started protocol treatment. Among them, one on tamoxifen and zero on placebo were ineligible. All patients were off-protocol treatment at time of data cut-off. Of women starting treatment, 103 (31%) on tamoxifen and 70 (21%) on placebo stopped early because of toxicity, refusal, protocol violation, and other causes (Table 4). Overall, 29% on tamoxifen and 27% on placebo received ≤2 years of protocol treatment. Forty-one (12.5%) on tamoxifen and 32 (10%) on placebo received other hormones before recurrence, which included progestogens for menopausal symptoms, tamoxifen extended beyond 5 years, and aromatase inhibitors (Table 4). Among them, 17 on tamoxifen and 22 on placebo started off-label tamoxifen before the scheduled end of blinded treatment.

Table 4.

Compliance with study medication

	Tamoxifen		Placebo
	n	%	n	%
Total off treatment	338	100	334	100
Reasons
Treatment completed	175	52	174	52
Recurrence/progression	50	15	78	23
Death	1	0	1	0
Intercurrent illnessa	9	3	11	3
Toxicity treatment	29	8	15	5
Refusal	44	13	35	11
Protocol violation	13	4	12	4
Otherb	17	5	8	2
Number treated patients	328	100	327	100
Duration of study medication (months)
0–12	64	19	45	14
13–24	33	10	41	13
25–36	19	6	31	9
37–48	23	7	21	6
49–60	189	58	189	58
Other hormones before progression and during protocol treatmentc	7	2	2	1
Tamoxifen	0		2
Letrozole	1		0
Progestogen	6		0
Other hormones before progression and after protocol treatmentc	34	10	30	9
Not completed 5 years of blinded treatment
Tamoxifen	17		20
Letrozole	12		1
Anastrozole	1		0
Exemestane	0		3
Progestogen	4		2
Completed 5 years of blinded treatment
Tamoxifen	0		5
Letrozole	11		2
Exemestane	1		0

^aIncluded development of contralateral breast cancer.

^bIneligible, noncompliant, stopped early/unblinded in error, physician advice, desire for pregnancy, left the country.

^cA patient may have received more than one hormonal agent.

adverse events/toxicity

Toxicity was evaluated in 328 patients on tamoxifen and 327 on placebo who started protocol treatment (Appendix, Figure A1). At baseline, 54% and 50% of women subsequently randomized to tamoxifen and placebo, respectively, were complaining of hot flashes. Similar numbers in each arm, 129 (38%) on tamoxifen and 121 (36%) on placebo, had ceased menstruation for at least 3 months, and a further 36 (11%) on tamoxifen and 39 (12%) on placebo were having irregular periods (Table 1).

Large numbers of grade 1/2 acute toxic effects were reported but were similarly distributed across the arms. For grade 3/4 (Appendix, Table A1) toxic effects, the only important difference was in frequency of severe hot flashes, which were significantly worse on tamoxifen (tamoxifen 23% versus placebo 14%; P = 0.005).

Twenty-one (6%) patients on tamoxifen and 33 (10%) on placebo developed 56 second malignancies. There were fewer contralateral breast cancers (14 versus 18) and fewer other malignancies (7 versus 17; P = 0.07) in the tamoxifen arm. The only two cases of endometrial cancer were on placebo.

discussion

In this placebo-controlled study, adjuvant tamoxifen following chemotherapy produced improved DFS (HR 0.77; P = 0.056) and a trend for improved OS (HR 0.78; P = 0.12) in premenopausal women with EBC. Although exploratory subgroup analysis indicated a trend for greater benefit in women with ER-negative tumors, this was not statistically significant. Such a finding has no biological rationale and is not in agreement with results of other similar studies [8–10].

A number of factors delayed analysis of this trial to a time when use of adjuvant tamoxifen has become standard practice after adjuvant chemotherapy in premenopausal women with receptor-positive tumors. In the late 1990s, slow accrual probably reflected an increasing conviction among physicians and patients that tamoxifen was beneficial in this setting, despite limited data actually available at that time [11]. The study was closed with a reduced sample size because of slow accrual after recruitment of 672 women (planned accrual 800) based on recalculation of the statistical assumptions, as outlined in the ‘Patients and Methods’ section. Our primary end point was OS, but over the time span of the trial, mortality was much lower than anticipated, probably because of increasing efficacy of adjuvant therapies and better salvage therapies for metastatic disease. The recent declining mortality for women with EBC has been well documented across many countries [12].

Several studies of similar design (evaluating tamoxifen given after adjuvant chemotherapy in premenopausal women with EBC) have demonstrated improvements in DFS, but are not definitive with respect to OS [8–10]. Our study is the remaining unpublished randomized trial directly addressing this question. A future individual patient data meta-analysis may be more informative concerning the effects of tamoxifen on OS in this population.

The study most comparable to ours was conducted by the IBCSG (13.93), 1993–1999, and recruited premenopausal node-positive women [8]. All received chemotherapy followed by randomization to tamoxifen for 5 years or no endocrine therapy. Of 1246 assessable patients, 735 (59%) had ER-positive tumors, and patients receiving tamoxifen versus control had improved DFS at 5 years (75% versus 62%; HR 0.59; P = 0.0001), although no benefit was seen in the ER-negative cohort. There were no significant differences in OS in ER-positive (HR 0.86; P = 0.36) or ER-negative subgroups (HR 0.92; P = 0.63).

Intergroup protocol INT-0102 [9], 1989–1993, compared 5 years of tamoxifen with no endocrine therapy in 1690 women with high-risk node-negative breast cancer (56% premenopausal) who had completed adjuvant chemotherapy. At 10 years, improved DFS (HR 1.32; P = 0.003) and OS (HR 1.26; P = 0.03) were seen in the ER-positive subgroup, but not overall. Subgroup analysis was not carried out according to menopausal status.

In European Organisation for Research and Treatment of Cancer (EORTC) study 10901 [10], 1863 women recruited 1991–1999, after chemotherapy, were randomized between tamoxifen 20–30 mg daily for 3 years or no endocrine therapy. Analysis was based on 1724 patients, 73% of whom were pre-/perimenopausal. Five-year DFS figures were 67% versus 63%, favoring tamoxifen (HR 0.84; P = 0.035). Benefit was greater in the ER-positive subgroup, but menopausal status did not influence effects of treatment.

In a North American setting, adjuvant chemotherapy remains the standard of care for premenopausal women with high-risk node-negative or node-positive EBC. For women with endocrine-responsive (hormone receptor positive) tumors, chemotherapy is usually followed by 5 years of tamoxifen. However, several European studies have shown that ovarian function suppression with or without tamoxifen can be as effective as standard CMF chemotherapy [13–15]. IBCSG is leading an international collaboration exploring ovarian function suppression with or without adjuvant chemotherapy in premenopausal women with EBC, through two ongoing randomized trials (IBCSG 24.02 Suppression of Ovarian Function, IBCSG 25.02 Tamoxifen and Exemestane Trial) [16].

The placebo-controlled design of our study allowed us to collect detailed comparative data on compliance, and we found that 31% on tamoxifen and 21% on placebo-stopped study medication prematurely for reasons other than recurrence, death, or intercurrent illness. The greatest attrition occurred in the first 24 months. Although data from EBCTCG [1, 17] showed benefit for durations of adjuvant tamoxifen as short as 1 year, results were better with longer therapy and currently, at least 5 years of therapy is regarded as optimal. We know that 17 stopping tamoxifen and 22 stopping placebo went on to receive open-label tamoxifen, but this may be an underestimate given its wide availability. As few patients received aromatase inhibitors, it is unlikely that their use affected efficacy outcomes.

The influence of compliance issues on the main efficacy outcomes will have been complex although a dilution of the potential beneficial effects of tamoxifen seems likely. None of the other comparable studies was placebo controlled, and data on other hormonal therapies used by women in control arms were not reported [8–10]. In INT-0102 [9], 57% of women completed ≥4.5 years of tamoxifen, with 8% stopping due to toxicity and 7% refusing therapy. In IBCSG-93 [8], 8% of women did not start tamoxifen and 33% discontinued drug prematurely. In EORTC 10901 [10], 4.4% stopped tamoxifen due to toxicity. In this study, a shorter planned duration of tamoxifen (3 years) may account for the lower dropout rate, but methodology for documenting compliance was not reported. Early discontinuation of hormonal therapy has also been reported in aromatase inhibitor studies. In ATAC, 22%–26% across the three arms withdrew from therapy [18]. A lower proportion of patients in Intergroup Exemestane Study, 13%–15%, discontinued therapy prematurely after a median of 36 months [19]. In MA.17, 19%–20% of women discontinued treatment early [20]. Compliance data were not reported for BIG 1–98 [21]. In a practice setting, problems with adherence to aromatase inhibitor therapy have also been documented [22].

Poor compliance with adjuvant tamoxifen has been described in clinical practice. Partridge et al. [23], documenting filled prescriptions for adjuvant tamoxifen among 2378 women through New Jersey's Pharmaceutical prescription program, found that 23% of patients missed taking tamoxifen on >20% of days studied, although 87% filled prescriptions in the first year. Overall adherence was 50% by year 4. Also using prescription refill data from a national prescribing database, Barron et al. [24] found that 22% of 2816 women stopped tamoxifen within 1 year, increasing to 35% at 3.5 years. Data from another community-based study of 2080 women demonstrated a correlation between poor adherence to adjuvant tamoxifen and increased risk of death from breast cancer [25]. The gaps between evidence-based recommendations for adjuvant hormonal therapy and clinical practice were the subject of a recent editorial [26]. Genetic variation in cytochrome P-450 2D6 affecting metabolism of tamoxifen may account for different degrees of toxicity and thus compliance [27].

In MA.12, amenorrhea rates were close to 50% after chemotherapy before randomization to tamoxifen/placebo, but were similar between the arms, and this is unlikely to have influenced the efficacy of treatment. Many ‘toxic effects’ recorded in our study, for both tamoxifen and placebo, probably were related to chemotherapy-induced menopause [28] and/or prolonged chemotherapy-related side-effects, particularly in the first year. This may account for higher dropout rates in studies including high proportions of premenopausal women treated with adjuvant chemotherapy, compared with aromatase inhibitor trials in which all women were postmenopausal, regardless of whether they received chemotherapy. Toxic effects clearly worse on tamoxifen included vaginal discharge/fibrosis and hot flashes, as documented in other studies. Our finding that mood changes were more frequent on placebo indicates that these are not specific side-effects of tamoxifen, but have other causes such as premature menopause, the emotional toll of the disease, and effects of multiple treatments.

To date, second malignancy rates are low and do not differ significantly between the arms. The only two cases of endometrial cancer occurred on placebo, and it is probable that women entering the menopause are less susceptible to the carcinogenic effects of tamoxifen on the endometrium [29]. A delayed effect is still possible, although all women have been off therapy for 1–8 years.

In conclusion, in MA.12, DFS at 5 years was marginally improved for premenopausal women receiving 5 years of tamoxifen after adjuvant chemotherapy for EBC. However, poor compliance with therapy is likely to have influenced the efficacy of treatment in this trial and other similar studies. Some of the compliance issues clearly related to the physical and emotional consequences of early menopause induced by adjuvant chemotherapy, but there were other contributory factors (Table 4). These findings have implications for the use of oral anticancer medications in many settings [30].

funding

AstraZeneca, through the National Cancer Institute of Canada Clinical Trials Group; Canadian Cancer Society, through the National Cancer Institute of Canada (# 4448).

appendix

Figure A1.

Consort diagram.

Table A1.

Numbers of acute adverse events/toxic effects (>10% incidence either arm)

Adverse event	Tamoxifen (328 patients)						Placebo (327 patients)
	1	2	3	4	Total (%)	%R	1	2	3	4	Total (%)	%R
Blood
WBC	82	30	3	0	115 (35)	0	74	20	1	0	95 (29)	1
Neutrophils	69	10	5	3	87 (27)	0	44	14	6	0	64 (20)	0
Hb	79	3	0	0	82 (25)	0	45	2	0	0	47 (14)	0
Cardiovascular
Edema	44	11	0	0	55 (17)	6	61	9	0	0	70 (21)	7
Endocrine
Amenorrhea	16	205	0	0	221 (67)	34	26	198	0	0	224 (69)	37
Hot flashes	61	134	75	0	270 (82)	78	87	131	47	0	265 (81)	75
Influenza-like symptoms
Arthralgia	60	13	4	0	77 (23)	4	59	33	4	0	96 (29)	4
Lethargy	125	24	1	0	150 (46)	19	136	27	3	0	166 (51)	23
Myalgia	40	13	3	0	56 (17)	4	43	17	3	0	63 (19)	2
Gastrointestinal
Nausea	45	8	1	0	54 (16)	7	46	8	1	0	55 (17)	8
Abdominal pain	21	6	0	0	27 (8)	1	29	9	3	0	41 (13)	2
Genitourinary
Vaginitis/fibrosis	66	22	0	0	88 (27)	21	55	9	2	0	66 (20)	15
Hepatic
ALP	28	1	0	0	29 (9)	0	90	2	0	0	92 (28)	0
AST	44	8	2	0	54 (16)	1	32	4	1	0	37 (11)	0
Infection
Infection	11	13	1	0	25 (8)	1	16	17	0	0	33 (10)	1
Neurological
Headache	15	20	3	0	38 (12)	4	27	20	1	0	48 (15)	4
Insomnia	20	11	0	0	31 (9)	3	24	13	3	0	40 (12)	4
Mood	60	40	6	0	106 (32)	17	80	46	7	1	134 (41)	21
Pain	39	11	3	0	53 (16)	2	55	16	1	0	72 (22)	2
Sensory	37	5	0	0	42 (13)	2	48	4	1	0	53 (16)	2
Osseous
Bone pain	20	11	2	0	33 (10)	1	29	15	3	0	47 (14)	3
Pulmonary
Cough	22	3	0	0	25 (8)	0	22	8	2	0	32 (10)	0
Skin
Alopecia	17	6	2	0	25 (8)	1	33	10	0	0	43 (13)	2
Skin changes	14	1	0	0	15 (5)	1	29	3	0	0	32 (10)	1
Rash/itch	32	11	1	0	44 (13)	3	31	14	1	0	46 (14)	3

%R = considered by investigator to be ‘possibly’, ‘probably’, or ‘definitely’ related to protocol treatment.

WBC, white blood cell; Hb, hemoglobin; ALP, alkaline phosphatase; AST, aspartate aminotransferase.