Posted on Nov 07, 2023

Professor Sunil Badve, MD, FRCPath, is Vice Chair for Pathology Cancer Programs and Professor at Emory University in Atlanta, Georgia, USA. He is a surgical pathologist and translational researcher with expertise in breast cancer and thymic pathology. He was also a member of the International Ki67 in Breast Cancer Working Group. We spoke to Sunil about the role of Ki67 in modern breast cancer diagnosis and treatment, as well as the challenges of testing with immunohistochemistry (IHC), and the opportunities and challenges for RT-PCR testing of Ki67.

Thanks for speaking to us Sunil. Firstly, how important do you think Ki67 is as a marker? Ki67 is a good marker for proliferation – it’s a fairly stable marker that is expressed throughout the entire cell cycle. It's a cheaper alternative to the existing molecular assays. And one could argue that a lot of the gene expression assays are measuring proliferation, but in a slightly more sophisticated manner. There was a paper published in 2008 Wirapati et al. which basically showed that if we remove proliferation from any of the gene signatures, all of them show significant loss of their predictive value. But the question ultimately is, how much additional benefit are you going to get from testing? And that’s a little difficult to answer. It depends on the clinicians using it, how they use it, what is the ultimate goal that they’re trying to achieve.

What are the issues with the reliability of IHC for Ki67? There are three inherent problems in Ki67 testing. First, there’s no tissue that is negative for Ki67. Everything in the human body proliferates to a lesser or greater extent. So, the question basically is, what’s a good negative control? Where is the lower limit? These have been fairly controversial. The second part of it is that cells tend to show a continuous expression of Ki67, so how much brown is brown enough? The [International Ki67 in Breast Cancer] Working Group basically said anything that has the slightest brown speck should be considered positive, while Dako and Eli Lilly Ki67 guidelines had a slightly higher threshold for calling something positive. These kinds of differences could lead to a significant difference in percentages of positivity between tumours. The third thing that causes problems is that there’s a significant amount of heterogeneity in Ki67 expression within different parts of the tumour. So how many fields should we count? How many cells should we count? Should we count hotspots? Should we count the entire slide? All of these things are significant variables when people are looking at slides. And so, there is significant variability between pathologists and labs when they’re assessing Ki67.

Do you think there’s good evidence for Ki67’s use as a predictor of chemotherapy benefit? “Evidence” is always a difficult question to answer, because we believe in personalised medicine and yet we do a 1,000 patient clinical trial to prove that something works or not - that’s the paradox of personalised medicine. At the end of the day, when you have a given patient in your office or you’re trying to plan their therapy, there’s no evidence. A given patient is going to have a binary outcome – they’re not going to be cured 86% of the time, they’re either cured or not cured. If your probability of having a good prognosis is in the high 90s, you have a choice that you may not treat a patient [with chemotherapy]. But if your probability is in the 50s or 20s, then what choices do you have? You’re using a hammer and a nail approach, where it doesn’t matter the size of the nail, you’re using the same hammer. Better approaches at individualizing chemotherapy are necessary. So, if you test for Ki67 and it comes back as 5%, you’re basically going to say, “forget about chemotherapy, I’ll just try endocrine therapy”. If it comes back at 30% or 40%, then you’re going to say, “well, there seems to be a decent chance that the chemotherapy will work, or I can give some newer agents such as CDK4/6 inhibitors and see if we can keep the cancer under control”. But within that grey area between 5% and 30%, we are playing the probability game, because we don’t have a marker of response. We have markers for selection of therapy.

What about neoadjuvant chemotherapy – does Ki67 have a role to play there? Let’s figure out what the goals are. For me, the main goal of neoadjuvant chemotherapy is measuring biological response. Then, you know that a drug that you’re giving is actually working on the tumour – as opposed to taking the tumour out, you have nothing to measure, so you’re just giving the chemotherapy and then hoping that it’s working against the tumor. Somehow that is not stated as strongly as it should be, in my opinion. Biological in vivo response is a phenomenal indicator of long-term outcome. Theoretically, if a patient has to receive any chemotherapy, that patient should receive neoadjuvant chemotherapy. Coming back to the Ki67 question, what are we additionally providing? If the patient is node-positive, whether you should prescribe chemo or not has always been an important question, particularly after the RxPONDER trial that showed that high Oncotype is associated with chemotherapy response. Low Oncotype patients, irrespective of the node positivity, may not get that much benefit from chemotherapy, particularly if they are post-menopausal. So, with all of that data, how we use Ki67 again becomes subjective. And some people say, well, if there’s relatively low Ki67, I might try neoadjuvant endocrine therapy, look at Ki67 before and after, see if the Ki67 drops. If it drops, wonderful. If it doesn’t drop, then we can use additional drugs including chemotherapy. That was the design of Mitch Dowsett’s POETIC trial. A fair number of oncologists have started using that approach anyway, because by the time the patient gets scheduled for surgery, you might as well give the patient something in the meantime and then use that as a parameter to predict outcomes. But we know heterogeneity exists, we know that biopsy and excision specimens may not correlate. So how you interpret the data is always problematic. If Ki67 drops, that’s good. If it doesn’t drop, what does it mean? We have no idea and further work is needed in this area.

What’s your opinion on RT-PCR versus IHC for Ki67? The advantage of IHC is you’re looking at single cells, but the disadvantage of IHC is you’re looking at single cells. You can flip it around too: The advantage of RT-PCR is you’re looking at a population of cells. But we don’t know whether they are tumour cells, luminal cells, stromal cells, granulation tissue, or anything else. They’re never going to be congruent because fundamentally there’s a difference in technology. When that kind of difference exists, it’s very difficult to say which is better. Obviously, when you have a population of pure tumour cells, the genomic assays are going to be better. If you have a mixed population of cells where we have normal cells, some proliferative disease, some lymphoid clusters, and this, that, and the other… it’s much more difficult for the genomic assays to differentiate. So, in some ways, they’re apples and oranges. And most of the time it’s not a problem as they are congruent, but understanding the superiority or inferiority is going to be more challenging.

What are your thoughts on Ki67 testing for breast cancer? What role do you think IHC and RT-PCR have to play? Join the conversation – follow Cerca Biotech on LinkedIn, or email us at