102 Issue 32 Interview Transcript
Kristine Crane: Welcome to the JNCI podcast. I’m Kristine Crane. As genomics research expands, scientists are coming up with models of disease risk based on genetic information. I talked with Dr. Matthew Mealiffe, the founder of Clinical Utility Consulting, a biotech, pharma and digital health consultancy company. Dr. Mealiffe, along with his colleagues from the Fred Hutchinson Cancer Center conducted a cohort study that combines two genetic risk models. They looked at whether combining the Gail Model and seven SNPs or single nucleotide polymorphisms associated with breast cancer risks improved overall breast cancer risk prediction. Welcome, Dr. Mealiffe.
Matthew Mealiffe: Thanks.
Kristine Crane: Why don’t you start by explaining the differences between the Gail Model and SNPs?
Matthew Mealiffe: So I think it’s best to start with the Gail Model, since it’s been around longer and is more established. Basically, the Gail Model was developed to provide a better way to estimate breast cancer risk, and in part to identify parts of women for clinical trials who potentially could be most likely to benefit in terms of preventative therapy with tamoxifen and other drugs. And the Gail Model includes a few different components like personal reproductive history, age at first period, whether a woman has had children, and if so age at first live birth, family history of breast cancer in first degree relatives, in other words, mothers or sisters. And then it also includes some personal medical history components like whether a woman has had previous breast biopsies and if so, whether there have been a particular conditions called atypical hyperplasia that can be seen on breast biopsies. And essentially, the Gail Model allows risk over the next five years and life time risk for invasive breast cancer to be calculated. And then moving on to SNPs or single nucleotide polymorphisms, these are basically areas of our genome where different individuals quite commonly differ in which DNA base pair is present at a particular site in our genome. And so, in other words, if our genomes are like a book and the DNA bases are like letters, different individuals co
Kristine Crane: Okay, so what did your study look at specifically?
Matthew Mealiffe: So the Gail Model includes an element of family history, and obviously genetics plays a role in that. We knew that there were these seven SNPs that seem to be unequivocally associated with breast cancer in white, non-Hispanic women. And so we decided to combine those under the assumption that they acted multiplicatively. So we multiplied the risk conferred by those different SNPs, and then basically combine it with the Gail risk scores. So in other words, we estimated the Gail risk scores and multiplied those by the risk scores that came from the SNPs.
Kristine Crane: Okay. So what women were you looking at?
Matthew Mealiffe: In this study, we collaborated with investigators from the Women’s Health Initiative Clinical Trial. The women who were studied were a subset of those in the WHI clinical trial. And in total, it was a little bit over 1,600 women from the clinical trial who developed breast cancer during the period of trial follow up. And little bit more than 1,600 women who did not develop breast cancer during that time. And because we needed to know whether those women had the SNPs that we were studying, the women were genotyped. In addition, we had quite a bit of clinical data from that study that allowed us estimate the Gail Model risk scores for these women.
Kristine Crane: And so, what were the main findings of your study?
Matthew Mealiffe: I think there were a few key findings. First, when we looked at the risk scores that came from the SNPs, and compared those that came from the Gail Model estimation, there was a slight overlap. But really, when you looked at the overlap, for all practical purposes, it wasn’t clinically significant. So you can sort of infer that these were providing distinct types of information that could potentially be combined. And then secondly, when we divided women into groups based on the risk scores coming from the SNPs, there was pretty good agreement between the predicted number of cancers expected during the follow-up period and the observed number. And then additionally, we compared the Gail and SNP and combined models in a couple of different ways. The first way used a technique called receiver operating characteristic curve analysis. And this is a way that clinicians use to compare the performance of diagnostic tests, and more specifically, to learn how good they are at discriminating individuals who will have a particular outcome, which in this case was those who would develop breast cancer and those who would not have that particular outcome. And so the bottom line was that the addition of SNPs improved, in a statistical way, significantly on the Gail model, but I also think its actually quite important to note that the improvement was modest. So this wasn’t a dramatic improvement, but it was definitely a statistically significant improvement.
Kristine Crane: So what do you think are the clinical implications of your study?
Matthew Mealiffe: I think that if you talked to several different people, you’d probably find some strongly differing opinions about that. From my perspective personally, I think that the major and near term implications are more in the clinical research setting than in the clinic. And to give an example, I think that undoubtedly there will be more clinical trials in the future that focus on other drugs in the context of breast cancer prevention. So trial design for those studies will need to consider the fact that genetic risk information can add something to the Gail Model. And so hopefully, in the future by more optimally estimating breast cancer risks for groups of women in clinical trials like that, the efficiency with which they can be conducted can improve. With respect to the clinic, personally I think we really need to spend some more time examining the clinical validity of models like this one, and also in populat-- excuse me, in population-based studies. And then there are really two other things that are important. One is that since starting this study, further GWAS studies have been conducted, and a significant number of additional SNPs that impact breast cancer risk have been discovered. In the future, those will need to be incorporated into models like this one. And then secondly, as I think I mentioned, because the Gail Model can not be applied to women of all ethnicities, and also because these particular SNPs were discovered in white non-His
Kristine Crane: So how does the kind of genetic information from your study and other studies like it differ from the information, for example on the BRCA genes?
Matthew Mealiffe: Sure. That’s a good question. There’s a couple of key points there. One is that what we’re talking about here is really quite different than hereditary breast and ovarian cancer risk that conferred by mutations in the BRCA1 or BRCA2 genes. And it’s that type of breast cancer risk that people are often hearing about when they hear about breast cancer risk in the lay press. For a small but significant number of women and families, mutations in the BRCA1 and BRCA2 genes really impact breast and ovarian cancer risk. And in general, in a clinical setting, we identify women who need testing for mutations in those genes based on family history. And although those genes are important, and really impact a small subset of individuals, we also know that risk of breast cancer can be impacted, albeit much more modestly, in a larger percentage of women by SNPs like those we studied. And it may be that in the future by utilizing combined models perhaps something like what we studied here, but more realistically, probably something that includes more SNPs in the future. We may be better identify groups of women who should perhaps be screened more aggressively or less aggressively, whether that be MRI or by mammography so that we can optimally identify those who need screening resources. And perhaps there’s a group of women who doesn’t need to be screened or doesn’t need to be screened as often. But it certainly remains to be seen whether that’s the case.
Kristine Crane: So just to sort of sum up, what do you think your study says about where we are in general, with genomics research?
Matthew Mealiffe: I think it’s important to stay tuned to this space. Expectations were really high initially in the context of the Human Genome Project. But this is all still really new. There’s been quite a bit of hype surround genomics. And I think we’ve reached a point where some of that hype has started to subside. And I would argue that there’s going to be a really long, slow slog through the steps of translational and clinical research that are ultimately necessary in order to show that something is useful in the clinic or not, and to get to the point that genomic medicine really can impact care in a lot of areas. So you know, I think I would sum up by saying that this and other studies are interesting initial steps, but there’s a long ways to go, and I think people need to be patient, but also cautiously optimistic.
Kristine Crane: Okay. Thank you very much.
Matthew Mealiffe: My pleasure.
Kristine Crane: That was Dr. Matthew Mealiffe, the founder of Clinical Utility Consulting. For more information on this podcast, a transcript of the interview or to listen to other JNCI podcasts, please visit our Web site at www.jnci.org. For general news updates from JNCI, you can follow us on Twitter at JNCI_now. I’m Kristine Crane. Thank you for listening.
#### End of jnci_102.21.meali ffe.mp3 ####
- About this journal
- Contact Us
- Rights & Permissions
- Dispatch date of next issue
- This journal is a member of the Committee on Publication Ethics (COPE)
- We are mobile – find out more
Carmen J. Allegra
Impact factor: 14.336
5-Yr impact factor: 14.794
For the Media
Open access options for authors - visit Oxford Open