JANUARY 16, 2013
ATLAS: Tamoxifen Far Inferior to AIs in Years 6 to 10
Tamoxifen has no benefit until year 11

Steven Vogl, MD
Medical Oncologist,
New York City

The ATLAS (Adjuvant Tamoxifen: Longer Against Shorter) trial of five additional years of tamoxifen versus observation was presented at the San Antonio Breast Cancer Symposium on Dec. 5, 2012, and published simultaneously in The Lancet to much notoriety and acclaim.1 The popular press ignored the obvious inferiority of tamoxifen to an aromatase inhibitor when it is given for five years after five years of previous tamoxifen and local therapy for breast cancer. Further, the press ignored the obvious exploratory nature of the statistical analysis, which was conceived after reviewing the data, and the very limited number of women who could potentially benefit.

ATLAS Trial Description

ATLAS is an open-label, randomized trial of a second five years of tamoxifen that accrued 12,894 women in 36 countries between 1996 and 2006. Mean follow-up in ATLAS now is 7.6 years from randomization. In 2006, it closed early when the results of the MA.17 trial indicated that prolonged adjuvant therapy with letrozole (an oral aromatase inhibitor [AI]) was clearly effective in the estrogen receptor (ER)-positive patient population given after five years of tamoxifen, with a 48% overall relative reduction in disease recurrences.

The current report was generated when the data were analyzed for the planned Oxford meta-analysis of the benefits of prolonged endocrine therapy in women with known ER-positive tumors. In the entire ATLAS study, only 53% of tumors were known to be ER-positive, with 37% ER status unknown (mainly in Asia). Of those with ER-positive tumors analyzed in the current report, 50% of women were more than 55 years old, 54% were node-negative, 47% had <21 mm tumor diameters, only 15% entered before 2000, only 9% were still premenopausal when randomized, 32% had only four to 4.9 years of prior tamoxifen, 1% had prior local recurrence, and 2% had prior contralateral breast cancer.

For women with ER-positive tumors, the overall rate of recurrence in years 6-15 was reduced 16%, death rate from breast cancer (attributed locally without review) was reduced 17%, and overall death rate was reduced 12.5%. No benefit was noted for women with known ER-negative tumors. Benefits for those whose tumors’ ER status was unknown were intermediate between those noted for women with ER-positive tumors and the absent benefits for ER-negative tumors.

The authors noted that there was no detectable benefit from tamoxifen for patients during the five years of therapy (i.e., years 6-10), but that considerable benefit emerged later. They reanalyzed the data by five-year periods to make the benefits after year 10 appear more impressive, since the first analysis suggested that the benefits changed with time, and since a statistical test for heterogeneity was positive (Table 1). This is reasonable and intelligent, but does not test their original hypothesis, and should be considered only an exploratory analysis.

Table 1. ATLAS Trial: Effects of Prolonged Tamoxifen on Recurrence and Breast Cancer Death
Observation (%) Tamoxifen (%) Rate Ratio
10-Year breast cancer recurrence 14.5 13.1 0.90 (years 6-10)
15-Year breast cancer recurrence 25.1 21.4 0.74 (years 11-15)
>15-Year breast cancer recurrence 0.85 (beyond year 15)
10-Year breast cancer death 6.0 5.8 0.97 (years 6-10)
15-Year breast cancer death 15.0 12.2 0.70 (years 11-15)
>15-Year breast cancer death 0.79 (beyond year 15)
ATLAS, Adjuvant Tamoxifen: Longer Against Shorter

Benefit of Letrozole Switch Huge Compared With Continued Tamoxifen: Letrozole has 48% drop in recurrence and 24% drop in death during years 6 to 10

Table 2 compares the results of 10 years of continued tamoxifen with the results of a switch to letrozole at the start of year 6, as done in MA.17. Letrozole is the obvious winner, although the results are somewhat complicated by the difficulty of correcting for partial crossover of placebo-assigned women to letrozole when the impressive MA.17 results were initially reported. Women who crossed over seemed to have their prognosis substantially improved.2

Table 2. Prolonged Tamoxifen Versus Switch to Letrozole: ATLAS Versus MA.171
Rate Ratio: Tamoxifen Rate Ratio: Letrozole
10-Year breast cancer recurrence (years 6-10) 0.90 0.52
10-Year breast cancer distant recurrence (years 6-10) 0.90 0.51
10-Year death from any cause (years 6-10) 0.871 0.76
ATLAS, Adjuvant Tamoxifen Longer Against Shorter
1. Figure for total death rate reduction with median follow-up of 12.6 years.

Thus, using the figures from the ATLAS trial report, of the 6,454 women who were assigned to prolonged tamoxifen, by year 10 letrozole would have prevented recurrence in 7% (449 women) and death by year 10 of at least 1.4% (93 women). We have no data on how the letrozole patients would do after year 10, but the 7% head start on fewer recurrences will presumably translate to many fewer deaths in years 11 to 15.

Minor Effects That Do Not Nullify The Major Conclusions

Compliance is an issue for all prolonged hormonal therapies and their studies: By year 7, only 84% of women in ATLAS were still taking the tamoxifen to which they had been assigned, down to about 70% by year 9. Favoring a stronger effect of continued tamoxifen, if more of them had taken tamoxifen longer, fewer would have had disease recurrences and died. Favoring a stronger effect of letrozole after tamoxifen, lack of compliance probably similarly reduced the apparent efficacy of letrozole in MA.17. Making tamoxifen in ATLAS seem less effective than it really was, the control group of ATLAS had a 4% rate of non-protocol hormonal therapy at year 7 and a rate of about 6% at year 9, making the control group’s results seem a little better than they would have been if all the control patients had stayed off hormonal therapy.

The small numbers of enrolled women with prior local recurrence and contralateral primaries made the overall prognosis of women entered on ATLAS somewhat worse. While it is not usual to enter such women into adjuvant trials, their small number means that their adverse prognosis affects the overall results very little.

Which Women Are Not Candidates for an Extra Five Years of Tamoxifen?

The obvious superiority of the AI switch to continued tamoxifen (even though this has not been the subject of a prospective randomized trial) means that every woman who is postmenopausal after five years of tamoxifen should get an AI for at least five years. Whether AI therapy should go longer is the subject of four studies yet to report results, although three have completed accrual. Of the women in the ATLAS trial, 90.5% were postmenopausal by the start of year 6. Thus, most women on the trial would best have been treated with an AI after the first five years of tamoxifen—hence the trial’s early closure to accrual.

This recommendation for AI switch after 5 years of tamoxifen includes women who were premenopausal at diagnosis who later became postmenopausal, according to a recently published analysis of MA.17.3 In this analysis by Paul Goss, MD, PhD, of Massachusetts General Hospital in Boston, the “switch” to letrozole in those women premenopausal at diagnosis produced a reduction in relapse rate of 74% compared to observation, while the reduction was only 33% for those post-menopausal at diagnosis.

Until the results of the RxPonder (SWOG S1007) trial become available (randomizing node-positive women with ER-positive tumors and lower Oncotype DX recurrence scores to observation versus chemotherapy), I believe the standard therapy for women with positive nodes and positive ER status in the primary breast cancer includes chemotherapy. The chemotherapy will render a large proportion of node-positive premenopausal women at diagnosis postmenopausal by the start of the sixth year, leaving very few for tamoxifen. Because younger women (<35 years) have a poorer prognosis by virtue of their age alone in many studies that correct for other prognostic factors, many physicians offer these women fairly intensive chemotherapy even when nodes are negative, further reducing the pool of women still menstruating after five years of tamoxifen. This leaves relatively few candidates for prolonged tamoxifen.

Which Women Should Get Another Five Years of Tamoxifen?

Women who are not candidates for an AI because they still have ovarian endocrine function and have sufficient remaining risk should receive tamoxifen. To help define the risk, the Austrian Breast Cancer Study Group (ABCSG) presented a gene expression assay called EndoPredict in San Antonio in 2012.4 The assay is suitable for use on formalin-fixed paraffin-embedded tumor tissue. This assay compares mRNA levels of three proliferation genes and five estrogen receptor–dependent genes to mRNA levels of three reference genes (similar to the techniques used in the Oncotype DX). The proliferation genes drive the predictive value of the test during the first 5 years of endocrine therapy, and the estrogen receptor-dependent genes in the next five years. Using clinical parameters of tumor size and number of involved nodes plus the EndoPredict score, 64% of women in a large validation cohort could be placed into a very low-risk group with almost no relapses after 5 years. If these results are validated, then the EndoPredict score would help spare many women long years of therapy they do not need.

Women considered for a second 5 years of tamoxifen also should not desire pregnancy, since tamoxifen is not established as safe for the embryo and fetus, and they should have tolerated the five years of treatment well. Alternative therapies for these women include ovarian suppression and an AI plus ovarian suppression, neither of which has been adequately studied in this setting. Of note is the slightly inferior activity, especially in more obese women, of anastrozole plus ovarian suppression compared with tamoxifen plus ovarian suppression in the ABCSG–12 trial that also looked at adjuvant zoledronic acid.5

Tamoxifen for the second five years is a reasonable alternative for women who cannot tolerate any of the available AIs, although the benefits seem to be much less. The ATLAS trial demonstrates the safety of continued tamoxifen, with an excess incidence of endometrial cancer of only 1.74%, and only an 0.2% increased mortality from endometrial cancer to year 15. Almost none of this risk is for premenopausal women, who constitute the vast majority of patients who should consider prolonged tamoxifen based on currently available data.

What Is the Nature of the Tamoxifen “Carry-Over” Effect?

The ATLAS analysis makes much of the benefits of the first five years of tamoxifen after it is stopped. The risk per year for relapse in years 6 to 10 after five years of tamoxifen is 69% of the risk after five years of placebo. The mechanism of this effect is unclear: Perhaps an active metabolite is slowly released from pools in body fat, or perhaps there is an endocrine effect on prolonging tumor dormancy that persists beyond administration of the drug. The point should be made, contrary to the ATLAS analysis, that the continued benefit from the first five years of tamoxifen does not preclude further benefit from hormonal therapy, since this is clearly evident in MA.17 and other AI crossover trials. It was also evident in trials of continued tamoxifen beyond one and two years of therapy, in which two and five years of tamoxifen produced immediate reductions in relapse rates.

Where Do We Go From Here?

The time beyond 10 years from diagnosis is now open for therapeutic intervention, since ATLAS showed a substantial benefit. The TAM-02 trial from France showed that delayed adjuvant tamoxifen (median of five years from diagnosis) improves disease-free and overall survival in ER-positive populations who were naive to both tamoxifen and AIs.6 This was a study of delayed tamoxifen in women who should have started earlier, but, for some reason, did not. It makes sense to study tamoxifen after an AI, the latter given for five years either as primary hormonal therapy or after a switch from five years of tamoxifen. Shorter versions of this switch to tamoxifen were as good as continued letrozole in BIG 1-98.

Should Any Woman Get Tamoxifen Now in Years 11 to 15?

Absent any clinical trial results, I would only consider very late tamoxifen for women with ER-positive tumors at very high risk (many positive nodes) who desire the most aggressive therapeutic posture—those who want “everything done”—and who have finished five years of an AI or cannot tolerate additional AI therapy. One would hope that the switch back to tamoxifen after an AI in years 6 to 10 and tamoxifen in years 1 to 5 would yield immediate benefits, unlike the much-delayed benefits seen in the ATLAS trial. We have no data in hand to support this hypothesis.


  1. Davies C, Pan H, Godwin J, et al. Long-term effects of continuing adjuvant tamoxifen to 10 years versus stopping at 5 years after diagnosis of oestrogen receptor-positive breast cancer: ATLAS, a randomised trial. Lancet. 2012 Dec 4. pii: S0140-6736(12)61963-1, PMID: 23219286.
  2. Jin H, Tu D, Zhao N, Shepherd LE, Goss PE. Longer-term outcomes of letrozole versus placebo after 5 years of tamoxifen in the NCIC CTG MA.17 trial: analyses adjusting for treatment crossover. J Clin Oncol. 2012;30:718-721, PMID: 22042967.
  3. Goss PE, Ingle JN, Martino S, et al. Impact of premenopausal status at breast cancer diagnosis in women entered on the placebo-controlled NCIC CTG MA17 trial of extended adjuvant letrozole. Ann Oncol. 2012 Oct 1. [Epub ahead of print], PMID: 23028039.
  4. Dubsky P, Brase JC, Fisch K, et al. The EndoPredict score identifies late distant metastases in ER+/HER2– breast cancer patients. Cancer Res. 2012;72(24 Suppl):Abstract nr S4-3.
  5. Pfeiler G, Königsberg R, Fesl C, et al. Impact of body mass index on the efficacy of endocrine therapy in premenopausal patients with breast cancer: an analysis of the prospective ABCSG-12 trial. J Clin Oncol. 2011;29:2653-2659, PMID: 21555684.
  6. Delozier T, Switsers O, Génot JY, et al. Delayed adjuvant tamoxifen: ten-year results of a collaborative randomized controlled trial in early breast cancer (TAM-02 trial). Ann Oncol. 2000;11:515-519, PMID: 10907942.

More in Vogl, NY
Vogl, NY: Put the Market Back in Drug Pricing
The drug market is not working to keep down the costs of generic ......
NOVEMBER 18, 2015
Extended-Field RT for 1 to 3 Positive Nodes–No!!!
Whelan et al recently reported negative results of the large, ......
Vogl, NY on Diagnostic Expansion in MM: Elegant, Thoughtful, but Risky
In a thoughtful and precise paper, the International Myeloma ......
MAY 18, 2015