Twelve-year mortality results of a randomized trial of 2 versus 5 years of adjuvant tamoxifen for postmenopausal early-stage breast carcinoma patients (SITAM 01)

Abstract

BACKGROUND

This study evaluated the impact on overall survival (OS) of 2 versus 5 years adjuvant tamoxifen in early breast carcinoma patients after 12 years of follow-up.

METHODS

Women with breast carcinoma T1–3, N0–3, M0, aged 50–70 years, were eligible for this multicenter randomized Phase III trial. Patients event-free after 2 years of tamoxifen therapy (TAM) were randomly assigned to stop or continue TAM (20 mg/day) for an additional 3 years. The primary endpoint was disease-free survival. Secondary endpoints included OS and toxicity.

RESULTS

From 1989 through 1996, 1901 patients were randomly assigned either to stop treatment (n = 958) or to continue TAM (n = 943). Overall, 98% of patients alive at the previous report (n = 1611) had updated information about OS, of whom 549 had died. The median duration of postrandomization follow-up was 115 months (interquartile range, 86–137). No statistically significant differences between the two arms were detected in the whole population (hazard ratio [HR], 1.02; 95% confidence interval [CI], 0.86–1.22) and in estrogen receptor (ER)-positive patients (HR, 0.90; 95% CI, 0.72–1.13). In the latter group, survival curves started to diverge after 90 months, showing a trend in favor of the 5-year arm. In younger (age ≤55 yrs) ER-positive patients longer TAM was associated with a 44% decrease in the risk of death (HR, 0.56; 95% CI, 0.31–1.00), while no clear benefit was documented in women older than 55 years of age (HR, 0.98; 95% CI, 0.77–1.25).

CONCLUSIONS

The benefits of longer TAM on OS start to emerge only after 9 years from diagnosis and seem to be more relevant in younger ER-positive women. Cancer 2005. © 2005 American Cancer Society.

Hormonal adjuvant therapy with tamoxifen in early breast carcinoma is widely accepted as a standard treatment because of its documented efficacy in improving disease-free and overall survival (OS). The scientific evidence of its efficacy is strong, as documented by the systematic meta-analyses of all randomized trials in early breast carcinoma coordinated by the Early Breast Cancer Trialists' Collaborative Group (EBCTCG).1–3 However, the optimal duration of tamoxifen therapy (TAM) is still a matter of debate; in fact, the results of the overview suggesting a greater benefit with longer treatment duration are derived from the indirect comparison of different lengths of TAM.

The Italian Study of Adjuvant Treatment in Breast Cancer(SITAM)-01 trial was designed to compare 2 versus 5 years adjuvant treatment with tamoxifen in women aged 50–70 years with early-stage invasive breast carcinoma. In our previous report (with a postrandomization follow-up of 52 mos), we found a statistically significant prolongation of disease-free survival (DFS) related to longer treatment duration in estrogen receptor (ER)-positive patients, while no difference in terms of OS could be detected.4 These findings were in agreement with the results of the Cancer Research Campaign Trial, in which no data on receptor status were reported, and those of the French trial by Delozier et al.5, 6 Conversely, a Swedish trial documented a significant improvement in survival associated with longer tamoxifen duration, but survival curves started to diverge only after 6 years from surgery.7

Therefore, we decided to update our data to a median follow-up period of 115 months (interquartile range, 86–137 mos).

PATIENTS AND METHODS

The SITAM-01 was initiated in 1989 by the Italian Interdisciplinary Group for Cancer Care Evaluation (GIVIO) Group. It was a multicenter, randomized, unblinded clinical trial comparing 2 versus 5 years of TAM in women aged 50–70 years with early-stage invasive breast carcinoma. The study design was pragmatic in entry criteria. This approach was adopted to ensure wide participation and high patient accrual. All women with operable invasive breast carcinoma T1–3 N0–3 M0 (according to TNM Staging)8 aged 50–70 years were eligible, irrespective of tumor grade or ER status. The protocol was approved by the Italian Ministry of Health. The study involved 53 Italian outpatient oncology clinics. Oral informed consent was obtained from all patients. Patients were registered at any time up to 2 years after surgery, but TAM had to be started within 1 month after surgery. In patients receiving chemotherapy, hormonal treatment was postponed until after systemic therapy. Those patients who were alive and event-free after 2 years of TAM were randomized to stop or continue it for an additional 3 years. Random assignment of patients was accomplished through telephone calls by the trialists to the Trial Office.

Between June and December 2004, we requested information about the vital status of the 1611 patients alive at the date of the last analysis. The information was obtained through personal contacts with the oncology units participating in the trial or directly by the registry offices if the patient was lost to follow-up. An updated information on vital status was available for 98% of the cases. No additional details on causes of death could be obtained.

Statistical Analysis

Power and sample size estimation of the study have been described elsewhere.4 OS was defined as the length of time from the date of randomization to death for any cause.

All analyses were performed on an intention-to-treat basis. Probabilities of OS were calculated using the Kaplan–Meier method,9 and comparisons were carried out using the log-rank test. Hazard ratios (HRs) with their 95% confidence intervals (95% CI) were also estimated using the Cox proportional hazards model.10 All reported P-values are two-sided.

Because in the previous report we showed a statistically significant longer DFS among the ER+ patients allocated to the 5-year treatment arm, we decided to perform subgroup analyses within this patient subgroup according to the following variables: age (≤55 yrs – lower quartile, >55 yrs), pathologic tumor size (T1, T2, T3), number of positive axillary nodes (N0, N1–3, N>3), and previous chemotherapy (yes, no).

All analyses were performed by using the SAS Statistical Package v. 8.2 (SAS Institute, Cary, NC).

RESULTS

Overall Findings

From 1989 through 1996, 1901 patients were randomly assigned either to stop treatment (n = 958) or to receive tamoxifen for 3 additional years (n = 943). The characteristics of the patients are shown in Table 1. At the time of the previous analysis (December 2000), 1611 patients were alive. On the whole, 98% of these patients had updated information about OS. The information was provided by participating centers for 21% of the patients and by registry offices for the remainder. In 2% of the cases the residence town on the entry form was missing or wrong and it was not possible to obtain any information. The closing date for this analysis was December, 2004.

Table 1. Patients' Characteristics by Treatment Arm

Characteristicsa	Treatment arm		Total	P valueb
	2 yrs	5 yrs
a Continuous variables are summarized as mean and standard deviation; categorical variables are expressed as frequencies and percentages. b P values (5 vs. 2 yrs tamoxifen) refer to unpaired t- test for continuous variables and contingency tables chi-square test for categoric ones.
No. of patients	958	943	1901
Age in yrs	60.9 ± 5.9	60.8 ± 5.9	60.9 ± 5.9	0.6
Pathologic T				0.5
T0-1	499 (52.1)	478 (50.7)	977 (51.4)
T2	372 (38.8)	390 (41.4)	762 (40.1)
T3	35 (3.7)	28 (2.9)	63 (3.3)
T4	1 (0.1)	3 (0.3)	4 (0.2)
Unknown	51 (5.3)	44 (4.7)	95 (5.0)
No. of lymph nodes				0.6
0	531 (55.4)	516 (54.7)	1047 (55.1)
1-3	285 (29.8)	277 (29.4)	562 (29.5)
> 3	123 (12.8)	137 (14.5)	260 (13.7)
Unknown	19 (2.0)	13 (1.4)	32 (1.7)
Tumor size (mm)	20.8 ± 10.4	21.9 ± 10.7	21.4 ± 10.6	0.04
Estrogen status				0.6
ER+	584 (61.0)	557 (59.1)	1141 (60.0)
ER–	136 (14.2)	146 (15.5)	282 (14.8)
Not measured	238 (24.8)	240 (25.4)	478 (25.2)
Surgery				0.7
Conservative	329 (34.3)	335 (35.5)	664 (34.9)
Mastectomy	611 (63.8)	597 (63.3)	1208 (63.6)
Unknown	18 (1.9)	11 (1.2)	29 (1.5)
Chemotherapy				0.2
Yes	87 (9.1)	101 (10.7)	188 (9.9)
No	871 (90.9)	842 (89.3)	1713 (90.1)
Radiotherapy				1.0
Yes	376 (39.3)	373 (39.5)	749 (39.4)
No	559 (58.3)	557 (59.1)	1116 (58.7)
Unknown	23 (2.4)	13 (1.4)	36 (1.9)

The median duration of postrandomization follow-up was 115 months (interquartile range, 86–137 mos), with no difference between the two treatment arms. The total number of deaths was 276 (28.8%) in the 2-year arm and 273 (29.0%) in the 5-year arm. No statistically significant difference between the two arms was detected in the entire population (HR, 1.02; 95% CI, 0.86–1.22), as reported in Figure 1.

Figure 1. All patients: overall survival (OS) by treatment arm.

Subgroup Analysis

The 8-year and 13-year postrandomization, corresponding to 10 years and 15 years after initial surgery, OS percentages are shown in Table 2.

Table 2. Overall Survival According to Treatment Group and Patient Characteristics

	96 mos % survival (± SE)	156 mos % survival (± SE)	HR	95% CI	P
ER+: estrogen receptor positive; N+: Node positive
All
2 yrs	79 ± 1	64 ± 2	1.02	0.86–1.22	0.79
5 yrs	78 ± 1	65 ± 2
ER+
2 yrs	79 ± 2	61 ± 4	0.90	0.72–1.13	0.38
5 yrs	80 ± 2	66 ± 3
ER+ N0
2 yrs	85 ± 2	68 ± 8	0.77	0.52–1.12	0.17
5 yrs	89 ± 2	78 ± 4
ER+ N+
2 yrs	71 ± 3	53 ± 4	1.00	0.76–1.33	0.99
5 yrs	70 ± 3	54 ± 4
ER+ N = 1–3
2 yrs	75 ± 3	55 ± 5	1.11	0.78–1.58	0.55
5 yrs	72 ± 4	54 ± 5
ER+ N>3
2 yrs	61 ± 6	49 ± 6	0.81	0.50–1.30	0.38
5 yrs	67 ± 5	52 ± 7
ER+ age ≤55
2 yrs	81 ± 4	62 ± 7	0.56	0.31–1.00	0.05
5 yrs	87 ± 3	77 ± 8
ER+ age >55
2 yrs	78 ± 2	67 ± 3	0.98	0.76–1.25	0.86
5 yrs	78 ± 2	66 ± 3
ER+ T1
2 yrs	84 ± 2	60 ± 7	0.85	0.61–1.19	0.34
5 yrs	85 ± 2	73 ± 4
ER+ T2
2 yrs	72 ± 3	58 ± 4	0.90	0.65–1.24	0.52
5 yrs	74 ± 3	59 ± 5
ER+ T3
2 yrs	54 ± 14	54 ± 14	1.18	0.39–3.50	0.77
5 yrs	58 ± 14	31 ± 16
ER+ N+ no chemotherapy
2 yrs	72 ± 3	52 ± 5	0.98	0.72–1.34	0.92
5 yrs	71 ± 3	54 ± 4
ER+ N+ chemotherapy
2 yrs	66 ± 7	60 ± 8	1.13	0.56–2.30	0.73
5 yrs	66 ± 8	32 ± 23

Among ER-positive patients, there was no statistically significant benefit related to prolongation of TAM (HR, 0.90; 95% CI, 0.72–1.13). However, the survival curves started to diverge after 90 months postrandomization, showing a trend in favor of the 5-year arm (Fig. 2).

Figure 2. Estrogen receptor (ER)+ patients: overall survival (OS) by treatment arm.

The benefit of longer TAM was evident only in younger (age ≤55 yrs) ER-positive patients (HR, 0.56; 95% CI, 0.31–1.00) (Fig. 3), but not in older ones (HR, 0.98; 95% CI, 0.76–1.25 for age >55 yrs). Survival curves according to patient age (≤55 yrs vs. >55 yrs) are shown in Figure 4.

Figure 3. Estrogen receptor (ER)+ patients: overall survival (OS) by treatment arm.

Figure 4. Estrogen receptor (ER)+ patients: overall survival (OS) by treatment arm.

To formally explore the differential effect of longer TAM on OS in younger women, we performed a test of interaction between age classes and treatment arms in a multivariate analysis; the P-value for the interaction term did not reach statistical significance (P = 0.09). The analysis according to the other subgroups did not show any statistically significant difference in terms of OS.

DISCUSSION

In the previous report,4 after a median follow-up of 70 months we showed that 5 years of treatment with tamoxifen significantly improved event-free survival only among ER-positive patients, confirming the previous results of the French and Swedish trials.6, 7 Nevertheless, the benefit in DFS in ER+ patients did not translate to prolonged survival, in agreement with the results of the Cancer Research Campaign and French trials (median follow-up of 24 mos and 48–49 mos for the CRC and French trials, respectively).5, 6 On the other hand, in the Swedish trial a significant improvement associated with longer tamoxifen duration was documented after a median follow-up of 42 months.7 The most likely explanation for those findings was the lack of adequate statistical power (i.e., the small number of events) to detect significant differences in terms of OS. As the 1998 EBCTCG overview pointed out,3 the possible carry-over benefit of adjuvant TAM could also interfere with the interpretation of the results. Therefore, we concluded that a much longer follow-up would be needed before drawing definite conclusions about the effects of long-term TAM on mortality.

To our knowledge, this is the first updated analysis of a trial comparing 2 versus 5 years of TAM. In our study the median follow-up was 115 months after randomization, corresponding to more than 11 years from surgery. Despite such a long period of observation, we could not document any difference in terms of OS between the two treatment arms in the whole population and in the ER-positive subgroup. In the latter subgroup, the benefits of longer TAM on OS seem to emerge only after 9 years from diagnosis, when the survival curves start to diverge.

Post-hoc power calculations confirmed that the study had adequate power to detect a difference between the treatment groups in terms of OS. In particular, the power to detect a risk reduction of death by 25% at the 0.05 level of significance in this study was 0.92; for a risk reduction of 20%, the power was 0.74.

The analysis according to patient age suggests that the benefits of longer TAM in terms of OS can be more relevant in younger ER-positive women, for whom survival curves start to diverge 4 years after randomization. In this subgroup, prolonging TAM to 5 years was associated with a 44% reduction in the risk of death. Conversely, no trend suggesting better OS in relation to longer TAM was present in older patients, probably as a consequence of competing causes of death that may have masked the effect of longer TAM in reducing cancer mortality. Another explanation can be represented by the excess risk of death related to TAM adverse events (cardiovascular and thromboembolic events), which could be more relevant in older women. As described elsewhere,4 in our population we reported a significant increase in thromboembolic events in the 5-year arm, but such an excess risk was homogeneous across age strata, and not confined to older women.11 Unfortunately, we could not obtain any details about the causes of death, because most of the updated information derived from the registry offices, which provided only the date of death.

The lack of a clear benefit on OS does not rule out the importance of longer TAM in elderly patients with ER+ early breast carcinoma; in fact, the positive effect on DFS documented in our previous report was evident even in women older than 55 years. A careful evaluation in each patient of the risk–benefit balance derived from a prolongation of hormonal treatment would probably lead to a stricter selection of women requiring longer TAM, which could in turn translate into greater benefits in terms of mortality reduction. When the risks of adverse events outweigh the expected benefits in terms of DFS and quality of life, then stopping tamoxifen and considering alternative treatments can represent a valid alternative.12

From a methodological point of view, our data suggest that the benefit of hormonal treatment on OS requires a long time to become evident. Therefore, trials investigating endocrine therapy require a very long follow-up to provide reliable information about the role of this therapy in modifying the survival of patients. This finding is strengthened by the results of other trials on hormonal treatment. For example, in the ATAC trial (one of the largest trials on endocrine therapy), comparing the effects of anastrozole and tamoxifen in 9366 postmenopausal women, after a median follow-up of 68 months a difference in favor of anastrozole was detected in terms of DFS, but not in terms of OS among ER+ patients.13 Along the same lines, the results of the last EBCTCG overview14 clearly demonstrated that about 5 years of TAM are better than 1–2 years in ER positive, ER unknown disease, both in terms of recurrence and OS. Nevertheless, the protective effect of longer tamoxifen on mortality starts to emerge only after 2–4 years from randomization and becomes more evident after 5–9 years, probably as a consequence of a carry-over effect.

In conclusion, our results show that, after a median follow-up of 115 months from randomization, 5 years of TAM are superior to 2 years of treatment in reducing total mortality only in ER+ patients aged ≤55 years. However, subgroup analyses can be subjected to substantial instabilities because of the small number of patients at risk; furthermore, the test of interaction between age and tamoxifen duration did not reach statistical significance, although this test has inherent limitations due to its low statistical power. Therefore, we believe that our findings should only be used to generate further investigation, rather than to draw definitive conclusions, and to underline the importance of a long follow-up in trials exploring the effects of hormonal therapy in breast carcinoma.