You’ve been looking here at HER2 status; very interestingly, you’re also concerned with aromatase inhibition and tamoxifen. Can you tell me what was the big issue that you were investigating here? What did you set out to do?

HER2 has long been recognised as a driver of endocrine resistance. Patients who are HER2 positive tend to relapse earlier on endocrine therapy than patients who are not and the question we were addressing was whether or not patients who should receive endocrine therapy are better treated with aromatase inhibitor from day zero if they are HER2 positive or whether they could delay and use a switching strategy for two or three years and potentially extend their endocrine therapy for maybe five or seven years.

So explain what you did in the study, please, with it.

What we performed was a patient level meta-analysis of three trials; in each trial HER2 had been tested in one of three central laboratories. We were able to recruit 12,000 patients with data to the study and to test whether or not HER2 acted as a predictive biomarker for benefit from AIs versus tamoxifen in the first three years of treatment which is what informs that treatment switch – should I start now or should I wait.

From a quick reading it seems to me this is not completely straightforward though, is it?

No, it isn’t completely straightforward. We satisfied the criteria for our primary endpoint, we had a significant treatment biomarker interaction.

Meaning what?

Meaning the benefit from AIs over the first three years represented a 30% risk reduction in HER2 negative patients but actually there was minimal evidence of a risk reduction in HER2 positive patients, in fact the hazard ratio was 1.1 suggesting they might even have some deleterious effect. However, as you said, it wasn’t straightforward. The challenge was that the HER2 population is a small population of the overall test subjects so we only had 1,000 HER2 positive patients. In that population we only had 111 events, that affects the statistical power of that interaction term. Analysing three trials we also saw a degree of heterogeneity, a different effect, in some of the different trials. Again, formally that wasn’t statistically significant, reinforcing our primary conclusion, but it concerned us that we couldn’t necessarily recommend a change in practice because of that heterogeneity. There was a second issue that prevented us from recommending a change in practice and that is at the time the trial was performed monotherapy with Herceptin for HER2 positive cases who had not received chemotherapy was not available. So in the modern era it’s possible perhaps to recommend these patients for Herceptin treatment, that may change our results.

Of course, some people switch from tamoxifen to aromatase inhibition, is that included in your calculations?

This was the primary goal of our calculation was looking at the pre-switch period. So we haven’t yet looked at what happens after the switch, that’s going to be a further analysis in the future.

But it seems that patients with HER2 negative disease, a certain category of patients, do better with an AI.

That seems to be the implication. Again, in so far as the treatment biomarker interaction tests that, we can make a fairly robust conclusion in that space. We’re less able to conclude that the HER2 positive patients do not benefit, although there’s a question mark about that.

So could you sum up the conclusions of this then, for me?

The conclusion we have drawn is that there remains a question mark about the benefit for AIs in the HER2 positive population, that further research is needed before we can make any recommendations to change clinical practice in that group. But we will be going away to do that research over the next year or so.

So as a predictive marker, HER2 does what in this role?

As a predictive marker statistically HER2 would select those patients who should get up-front AI and they would be the HER2 negative population. But with the caveats that I have explained around this study we would not regard that as sufficient at this stage to drive a change in clinical practice.

So your recommendations to doctor right now would be what?

Continue as you have been doing.

So the straightforward strategy normally would be?

The straightforward strategy normally will depend on clinic to clinic. Some clinicians will use a switching strategy for HER2 positive or HER2 negative patients, some clinicians will go up front, that’s based on other meta-analyses of the aromatase inhibitor trials. Clinicians will look at their individual patients and make those decisions and I guess the take home message is don’t change the way you do that from last week to this week.

It’s a bit disappointing that there isn’t a predictive biomarker, what hopes do you hold out for finding one?

It’s a challenging space. The problem that we have is that the biomarker positive groups tend to be small. There is still hope that we can do this; we have evidence that other drivers of this pathway may increase the population that we can test and it may be that in the next year or two we come up with a better analysis. But it’s a very challenging space, predictive biomarkers.

What’s the bottom line for doctors deciding on AI versus tamoxifen?

Go back to the aromatase inhibitor overview group’s evidence on the trials. Most of those suggest that up front AIs would appear to be more beneficial. The p-value for the switch versus up front AIs is 0.45, it’s a very modest change at present. I would say watch this space for further evidence emerging in that particular setting but certainly every patient probably requires an AI at some point in their endocrine therapy.