Share Print

PEPI: A New Predictive Measure for Breast Cancer Recurrence


Sept. 29, 2008 – Medical oncologist Matthew Ellis, MB, BChir, PhD, discusses research that has resulted in the development of a new predictive tool for breast cancer treatment. The preoperative endocrine prognostic index (PEPI) estimates the risk of cancer recurrence after treatment. PEPI scores could help physicians determine whether chemotherapy is necessary after surgery to reduce the chance of relapse.

Other Siteman web resources.


On this edition of Cancer Connection, we’ll talk about the Preoperative Endocrine Prognostic Index and what it means to women diagnosed with breast cancer.

Host: Thanks for downloading this podcast from the Siteman Cancer Center at Barnes-Jewish Hospital and Washington University School of Medicine in St Louis. I'm Jason Merrill. Breast cancer specialist Matthew Ellis of the Siteman Cancer Center has developed a new predictive measurement for breast cancer patients called the PEPI score, or Preoperative Endocrine Prognostic Index. In essence, the score looks at how tumors respond to anti-estrogen therapy in the short term to predict how well patients will do in the long term. A study on the topic was published in the Sept. 23 Journal of the National Cancer Institute. To tell us more about it is Dr. Matthew Ellis. He is head of the Section of Medical Oncology at Washington University School of Medicine. Dr. Ellis, thank you for joining us.

Ellis: Happy to be here once again.

Host: So let’s start at the beginning. What is the PEPI score?

Ellis: So this is a simple, pragmatic and easy-to-apply approach to individualizing the care of patients with hormone-receptor-positive breast cancer. So here’s the bottom line: Normally when patients are treated for breast cancer, the first instinct is to remove the cancer by surgery as the first step. Over the years, we've begun to recognize that doing that may mean we lose our best opportunity to find out what the patient actually is suffering from because once you remove the cancer, you have no way of knowing what it responded to. So this has been a principle that has been developed in part through the application of chemotherapy before surgery. My interest was applying endocrine drugs before surgery.

We've known for some years, perhaps as long as the last 10 years, that you could shrink breast cancers that were hormone-receptor-positive with simple low-toxicity drugs such as tamoxifen or estrogen-lowering drugs called aromatase inhibitors. When we did these studies, we noticed that some patients had great responses, and the tumor shrank before surgery, often leading to lesser degrees of surgeries, like avoiding mastectomy and getting a lumpectomy. We also found that some patients have tumors that should have responded – they had the hormone receptors, and they should have shrunk – but they didn't. So we followed these women over several years. (The average follow-up is over five years.) Some of them had relapsed and died, and some of them hadn’t. And we asked the question: Could we predict those relapses not based on the tumor samples taken before the hormonal therapy but the tumor sample taken at surgery after four months of treatment, when the hormonal therapy had an opportunity to shrink the cancer or not?

We then put all the data we had into a big statistical meat grinder called a Cox Proportional Hazards Analysis and found that four factors independently predicted relapse-free survival. One, the pathological tumor size. Now that’s not a surprise. If the tumor didn't shrink or was big to begin with and didn't get smaller, the patient didn't do well. If the tumor shrank, then the patient did better, particularly if the tumor size dropped below 2 centimeters. The second thing is whether the cancer involved the lymph nodes in the armpit or not. That’s also not a surprise. We’ve known for many years that breast cancer that spreads to the armpit tends to do badly. But there were two new components to the score that were interesting. One is a measurement of tumor growth. Now this sounds very straightforward, and it is. What matters for the rate patients relapse is not what the rate of growth is in the beginning so much as what’s the rate of growth when the patient is on the hormonal therapy. So we found there were some tumors that were growing quite fast at the beginning, but when we measured them at surgery, they’d completely stopped growing – there were no growing cells in the tumor. The measurement is called Ki-67. It’s a measure of tumor growth, and if the tumor had stopped growing, patients obviously did better than if the tumor continued to grow on the drug. And then the last thing that was interesting is I told you all these patients were hormone-receptor-positive at the beginning – they had estrogen receptors. But some tumors appeared to lose their estrogen receptors with treatment, and when they lost the hormone receptors, they did badly. Well, their tumors did badly.

So what we did is put all these factors together in a model where we weighted them by points. The bottom line is there was a group of patients whose tumors shrank below 2 centimeters, weren’t growing at surgery and still expressed those hormone receptors. And in that group, there were no relapses. Now in one study you would have said, well, that's a fluke result. But we did a second study in London in a different cohort of patients and found exactly the same thing. So there's a group of patients who have no relapses who couldn’t possibly benefit from getting chemotherapy. And then there's another group with a PEPI score of four or more. These are big cancers that continue to grow despite the hormone therapy, and some of them have lost the estrogen receptor. They behave like hormone-receptor-negative disease with a very aggressive disease course. So those patients clearly need everything available. And as with all these prognostics tests, there’s a middle group that’s either patients with small cancers but bad biomarkers or patients with big cancers but good biomarkers. Further research will be required for that group.

Host: So if a woman as a score of zero or a score of four what does that mean?

Ellis: Well, if she has a score of zero, she'll do very well and will not need chemotherapy. That's the implication of the research: The hormone therapy alone is sufficient to ensure a good survival. If she has a score of four, she definitely needs all available treatment and should probably go into a clinical trial because many of these patients in the study got chemotherapy and still did badly. So it's a group of tumors that are really aggressive. And only more research will tell what we’ll do with the middle 40 percent, where we’ve got small cancers that may have aggressive biomarkers or big cancers with good biomarkers. We’ll have to study those patients further. But for now, I'd recommend they all get standard-of-care treatment.

Host: How important is this approach to individualizing breast cancer treatment?

Ellis: I think it's very important because it's establishing a new approach to the problem. You see, hitherto we had just taken the tumor out and tried to guess how that tumor was going to respond to drugs. That’s been very difficult because you know some tests work a little bit overall but not very well. This is a devastatingly simple and therefore easy-to-apply approach, where instead of taking the tumor out, you just start the most important drug class and see if it works. This is the oldest trick in the medical textbook. If you want to treat a disease, you start the drug, and you see if the disease gets better. Whether it's blood pressure, infections or diabetes, you start the drug and see if they get better. And if they’re not getting better, you adjust what you do, right? Well in cancer, because it's a chronic disease, we’ve all known that if you take out the cancer immediately, you have no way of knowing whether that's a responsive tumor or not. And all we’re arguing for is giving more patients systemic therapy before surgery to work out which ones are going to be cured by these drugs and which ones need the clinical-trial approach. So it’s really, I think, another piece of the puzzle in changing clinical practice from “take it out immediately” to “let’s start a drug and see if the tumor responds before we take it out.”

Host: This is in the Journal of the National Cancer Institute, so a lot of physicians nationally will be seeing this. How soon do you hope this approach catches on?

Ellis: Well, it’s interesting. I’m running a large trial right now. It’s sponsored by the National Cancer Institute and the American College of Surgeons Oncology Group, and accrual is going incredibly well. So I think that this is something that physicians are very interested in. Now with all new approaches, there are groups of physicians who are willing to take on the advance immediately, and others who will want to see further data. But I guess the thing is, you know, measuring hormone receptors and measuring Ki-67 is something that everybody can do in their local laboratories. So I suspect this approach will be applied to patients with breast cancer who for whatever reason choose to undergo neoadjuvant endocrine therapy, perhaps to shrink cancer and promote breast conservation. And then at the point when they’ve had that done, I think physicians are going to be measuring proliferation index (Ki -67) and estrogen receptors and looking at the stage and the node status and applying a PEPI score and deciding whether they want to give them chemotherapy or not. I think it's going to happen immediately.

Host: Is there anything else we haven’t talked about?

Ellis: Having said all that and said I think it’s going to happen immediately because the data is all there and anyone can run these tests, I’d say a few things. First of all, I think it would be best if we recorded the outcomes and put these patients on clinical trial where possible. Secondly, I think we have to recognize the difficulties with immunohistory chemistry tests. I think that you should work very carefully with your local laboratory because the way the Ki-67 analysis is done requires a bit more effort than just the standard “is it high or low.” You actually have to count and give an absolute percentage. I also think that in the end, we might be able to replace the immunohistory chemistry with even more accurate diagnostic tests, and that's we’re in the process of doing right now.

Host: Dr. Ellis, thank you for joining us.

Ellis: Well, happy to have explained my research to you.

Host: For more information about breast cancer, you can visit the Siteman Cancer Center online at, or you can check out your risk for breast cancer at Thanks for downloading. Until next time, I’m Jason Merrill.