ATAC trial update

The ATAC trial has reached an important point in its evolution, with a median followup of 68 months (1.1) (ATAC Trialists’ 2005). Almost all of the patients are now off therapy, and we have one year of follow-up after the therapy was completed.

This is important for two reasons: it makes me comfortable about the efficacy and the hypothetical “carry-over effect” we’ve been hearing about for tamoxifen, and it makes me comfortable with the toxicities and tolerability of anastrozole. I believe this is probably the most important of the three ATAC analyses, and it allows me, as a practicing clinician, to change practice. I speak not only as a practicing clinician but also as the past principal investigator of the trial.

The simplest interpretation of the results is that anastrozole prevents one in four of the relapses we see in patients on tamoxifen. That translates into highly significant improvements in disease-free survival, recurrence-free survival and distant disease-free survival. The absolute number for difference in recurrence-free survival in the patients with receptor-positive disease at six years is close to four percent. It is important to remember that this trial included a group of patients with a relatively good prognosis.

In terms of relative risk reductions, we have no reason to suppose that the relative risk reductions will be different in any subgroup, and if that one in four relative risk reduction is across the board, then in a subgroup of patients with, for example, a 40 percent chance of relapse at six years, the absolute reduction is about 10 percent, not four percent, as was seen in the ATAC trial.

Survival as an endpoint in adjuvant therapy trials

This analysis was triggered by the number of distant recurrences and deaths from all causes. With regard to distant recurrences, our power calculations were correct; the trial was sufficiently powered to detect a significant difference. For overall survival, our power calculations were wrong, because this was a group of elderly women with a good prognosis, and the overall survival analysis is diluted by deaths from other causes before breast cancer recurrence.

I believe we can predict breast cancer survival with a fair degree of precision. I can’t see any reason why we would not eventually see a significant difference in breast cancer deaths. This trend is already present and is close to significance.

Whether that will translate into overall survival is uncertain. I’m not concerned about toxic side effects contributing to other causes of death — we know enough about anastrozole not to be worried about that — but I am concerned that the effect of preventing breast cancer deaths might be diluted by competing morbidities.

I’ve always been a purist, arguing that the only two real outcome measures in medicine are length of life and quality of life. I am on record as saying that all other outcome measures are surrogates, but we have to avoid waiting too long for the length-of-life outcome and to accept that surrogate measures translate into length of life with a fair degree of precision.

Quality of life and toxicity data in the ATAC trial

The use of anastrozole instead of tamoxifen does not impair quality of life. We can also say, with confidence, that the gynecological symptoms linked to tamoxifen have now translated into a fourfold increase in hysterectomy rates compared with anastrozole.

That is a dramatic observation, which we nearly missed. I was persistent about tracking down all the hysterectomies in women who had their wombs at the time of randomization. We came up with an extraordinary figure — I believe it’s the most extreme relative risk I’ve encountered in clinical trials.

The absolute numbers were 1.3 percent versus 5.1 percent (Howell 2004) for anastrozole and tamoxifen, respectively. This has a profound economic impact. I also don’t know how many hysteroscopies are being performed for every hysterectomy or how much the workup costs to decide whether a woman should have a hysterectomy, but these are big cost issues.

The update doesn’t give us any new information with regard to other prespecified adverse events, and no other adverse event is emerging with a frequency of more than one percent.

The fracture rate incidence is becoming a little more reassuring. An excess fracture rate occurs in the first two or three years, but then the lines are beginning to come together (Howell 2004). As patients stop taking anastrozole, the fracture rate returns to that of the patients randomized to tamoxifen. Furthermore, so far no difference has occurred in fractures of the neck or femur, which are of particular concern.

I think the issue of bone is easy to manage. We should be alert to it, monitor bone mineral density, perhaps exclude patients who have established osteoporosis, and then be ready to intervene with a bisphosphonate when the patient becomes osteopenic.

The polyarthralgia with anastrozole remains a problem. We don’t understand it, and it occasionally leads to withdrawal of treatment; however, the bottom line is that a significant difference exists favoring anastrozole for patients withdrawing from treatment because of side effects. If you evaluate the totality of side effects, anastrozole does better. If you consider the issue of the gynecological symptoms leading to hysterectomy, I believe the new drug — anastrozole — has the better tolerability profile.

Hazards for breast cancer recurrence with tamoxifen versus anastrozole

We are familiar with Kaplan-Meier curves, which are useful for the statistical analysis but don’t truly reflect what’s going on as the hazard ratios do. The best recorded data from patients treated by local therapy alone, before systemic therapy, is from Milan. A high and narrow peak for relapse occurs at two years, which then comes down again. A second, much-flatter peak for relapse occurs at about five years.

I’ve been working with Romano Demicheli, Michael Retsky and Bill Hrushesky on modeling and developing hypothetical explanations for this, and our review article will soon be published in the European Journal of Cancer.

Little doubt exists that the initial peak is provoked by the act of surgery. Surgery switches on a suite of genes for healing. The same suite of genes necessary for wound healing is necessary for provoking the growth of cancer, so what’s good for healing is also good for cancer.

In the hazard rate analysis plot from the ATAC trial, we’re seeing two peaks with tamoxifen. Using semiquantitative comparison with the Milan data, the first peak is lower with tamoxifen, but a peak still occurs. In the anastrozole arm, the initial peak is lost and the second peak is flatter. I believe this is the most profoundly important observation in this trial — not only to help make therapeutic decisions, but also to give a fascinating biological insight.

I believe the strongest argument for starting adjuvant endocrine therapy with an aromatase inhibitor is that anastrozole almost ablates that first peak (1.2). If you wait two to three years, as some of the trials are reporting, the effects are wonderful, but meanwhile you’ve lost those patients who will relapse and ultimately die in those first two years.

ER-positive, PR-negative subset in the ATAC trial

This is a nonprotocol evaluation of the data, but it’s very powerful. Professor Mitch Dowsett previously presented these data (Dowsett 2003), and now we have an update (1.3). The findings are even more striking than they were before.

Four phenotypes exist in the hormone receptor groups: double negative, double positive, and one or the other positive. The double negatives show no advantage with anastrozole compared to tamoxifen. Each of the other three phenotypes indicate an advantage with anastrozole, but the hazard rate favoring anastrozole in the estrogen receptor-positive, progesterone receptor-negative subset is 0.4 — almost a 60 percent relative risk reduction.

One would be skeptical about that as a data-derived observation, but a good mechanistic explanation exists. That particular phenotype tends to be Grade III, and tends to overexpress HER2. Earlier observations suggest that patients in whom HER2 overexpresses do favorably on aromatase inhibitors compared to tamoxifen, so I think it’s probably true, although it needs to be explored prospectively in another trial.

Switching patients from adjuvant tamoxifen to aromatase inhibitors

I am now absolutely confident that women who’ve been on tamoxifen for two or three years should switch to an aromatase inhibitor. We have excellent data for both exemestane and anastrozole from three trials. Boccardo’s small ITA trial with anastrozole was the first to report (Boccardo 2003), followed by the large IES study (Coombes 2004) with exemestane and the joint Austrian-German study of anastrozole presented in San Antonio (Jakesz 2004). Overwhelming evidence indicates that a switch to an aromatase inhibitor is beneficial.

I recommend the switch regardless of whether the patient has been on tamoxifen for one year or four years. You can wait forever for refinements, but no one is ever going to do a trial of a switch at one year or a switch at four years. We just have to stretch the available evidence and be sensible about it, and I think it would be reasonable to switch.

The MA17 trial is a well-conducted trial (Goss 2003) in women who have already received five years of tamoxifen. It shows proof of principle that you can influence the natural history of breast cancer after five years of tamoxifen. I’ve gone on record that I’m bitterly disappointed that they closed the trial and then allowed the placebo group to switch to letrozole, because they are treating the placebo group with experimental therapy — five years on tamoxifen, an average of two and a half years placebo, and then letrozole. That is an unproven treatment and I don’t think we’ll ever really learn the long-term benefit and toxicity.

I think we’re going way beyond the data. What worries me is that we cannot correct this situation. We’ll always be left with an area of uncertainty; however, to their eternal credit, the MA17 and NCIC group have redeemed themselves by being prepared to do a second randomization for duration after five years of the aromatase inhibitors.

Bisphosphonates in premenopausal women on tamoxifen or anastrozole

The Austrian study presented in San Antonio analyzed the capacity of zoledronic acid to prevent bone loss (Gnant 2004). The patients are all premenopausal women receiving an LHRH agonist. They are then randomly assigned to anastrozole or tamoxifen, followed by a second randomization to zoledronic acid or not.

In the main-effect analysis, zoledronic acid protects against osteopenia and osteoporosis. In the four-arm analysis, the bone mineral density in the goserelin plus anastrozole arm is the lowest, but the curve for goserelin plus anastrozole plus zoledronic acid runs parallel with the curve for goserelin plus tamoxifen plus zoledronic acid. I find that reassuring. It is evidence that zoledronic acid, a bisphosphonate, can reverse this loss of bone mineral density. The other thing that was somewhat of a surprise was that even the women who received tamoxifen and goserelin were losing bone.

Select publications

Dr Baum is Emeritus Professor of Surgery and Visiting Professor of Medical Humanities at University College, London, United Kingdom.