 |
|
Christina Bandera
Assistant professor of obstetrics and gynecology Brown Medical School |
At A Glance
Name: Christina Bandera
Position: Assistant professor of obstetrics and gynecology, Brown Medical School, since 2005.
Background: Assistant professor obstetrics, gynecology and reproductive biology, Harvard Medical School, 2003-2004. Instructor of obstetrics, 2002-2003.
Assistant professor of obstetrics and gynecology, University of Pennsylvania School of Medicine, 1998-2001. Instructor, 1995-1998.
Internship and residency, department of obstetrics, gynecology and reproductive biology, Brigham and Women's Hospital and Massachusetts General Hospital, 1991-1995.
MD, Johns Hopkins University School of Medicine, 1991.
Recently, ProteoMonitor has reported on a variety of ovarian cancer diagnostic tests that are being developed, including Ciphergen's test, the OvaCheck test by Correlogic, and a test developed by the Nevada Cancer Institute (see PM 7/29/2005; PM 5/13/1005).
This week, ProteoMonitor spoke with Christina Bandera, a gynecological oncologist involved in ovarian cancer biomarker research to find out her opinion on the various diagnostic tests being developed, and about her own work.
What do you think of the ovarian cancer diagnostic tests that are out there? How far away are they from the clinic?
I think that when it comes to the Quest and Correlogic assay — OvaCheck — it was a very interesting article that Liotta published in Lancet, and an interesting approach to finding ovarian cancer biomarkers. I'm sure you know that he was looking at patterns, instead of trying to identify individual proteins. The problem is that they took that approach and immediately tried to bring it to clinical use, and that's really not appropriate. These things really need to be tested prospectively. And we don't know based on his one little retrospective study if the woman has diabetes, or if she has a different kind of cancer, or if she's getting treatment for her cancer. It's really dangerous to just sort of put it on the market. I think that's why the Society of Gynecological Oncologists really came out very strongly against the possibility of making that test widely available. What really needs to happen is we need to do prospective studies. We need to enroll patients in studies that will look and see which biomarker tests are most appropriate.
In terms of how far away these types of studies are that would involve patients prospectively, that's happening on a small scale now, and I think it's going to be happening on a larger scale within the next couple of years. But I think it's going to be several years before it's something that a physician can just check off a box and say I'm sending you for this test. And my own bias is that it's not going to be a single protein test. I think it's going to be a panel of markers.
A panel of ovarian cancer biomarkers being developed by Ciphergen were reported to have a sensitivity and specificity of about .90. Is that too low?
It's too low in terms of specificity because you'd end up operating on a lot of patients who don't actually have ovarian cancer. It's encouraging, but I think it's going to take more than three markers in order to screen for ovarian cancer. Part of the problem is unlike some other cancers, like prostate cancer, ovarian cancer is very heterogeneous. We use one term, 'ovarian cancer', but actually when you look into a microscope at the cancer, there are many different cell types — there's endometrial, papillary, clear cells, transitional cells, and other more rare types, and within each type you have different grades. So you're coming up with, visually, close to a dozen different types of categories of ovarian cancer. And probably each type of cancer produces a different protein profile. So I think that's why we're not going to find one test, or even three proteins to screen for ovarian cancer. I think it's going to have to be a panel, and whether there are seven proteins on the panel, or 10 proteins or 20 proteins, I don't know.
If you add more proteins to the panel, do you think you'll improve the sensitivity and specificity?
Absolutely. I think you're going to improve the sensitivity that you can get with a specific specificity. The trick is figuring out which markers to put in the panel, and that's where the panel lies for the future. I think there are lots of markers that have already been identified that are going to be part of that test. And there are others yet to be identified.
A marker that I found is called M2-PK. In a small, retrospective study, combined with [the standard ovarian cancer biomarker] CA125, it had a pretty good sensitivity and a high specificity in a small number of patients. But what was really interesting to me was that this marker was very sensitive in the endometrioid types of cancers and the clear cell types of cancers which tend not to make the CA125 marker. So I'm not saying it'll necessarily be one of the proteins that are part of the panel, but those are the types of things we're looking for — proteins that aren't necessarily produced by most ovarian cancers, but are produced by cancers that don't produce anything else.
For the most common type of ovarian cancer, CA125 is a pretty good marker. But for other cell types, we don't have good markers.
How did you get into doing proteomic work related to ovarian cancer?
I'm a physician, and I was fortunate enough to get a couple of grants to do research — a women's reproductive health sciences grant from NIH, and I also got a grant from a private foundation called the Ovarian Cancer Research Foundation.
What I was really interested in is understanding how these different subtypes of cancer behave. And along with that came the question of how can we pick up these different types of cancer. There are a lot of ways to think about early detection, but clearly proteomics is the best way to non-invasively look for markers for cancer. So that's where my research led me to.
What kind of instrumentation did you use to do your work?
We're using liquid chromatography and MS/MS to ultimately identify the proteins.
What proteins did you find?
We found several candidate proteins — some of them we're still testing. The M2-PK is the first protein I started studying because an antibody was available for it. For the other proteins that I identified, I need to make antibodies.
I think what was really novel in the study [in which] I identified M2-PK was the samples I used. Most people compare ovarian cancer patients with normal controls. There's a problem with that because protein profiles vary so much from patient to patient. So instead of using normal controls, I sort of made every patient their own control. I took serum from their time of diagnosis and compared it to serum when they had finished treatment.
Ovarian cancer is a very treatable disease — 80 percent of patients will respond to surgery and chemotherapy and go into remission. Unfortunately, the cancer comes back. But at least when they're in remission, they presumably should have a pretty normal protein profile.
I think I found different proteins from what other people have reported, so I'm hopeful that it is a valid way of approaching the question. I really just had samples from eight patients. So it wasn't a large study, but on the basis of this I'm going on to do a larger study using samples from a clinical trial.
Besides M2-PK, what other proteins did you find?
I found lots. I'd say that for proteins that really seem to be of interest, so far I found five that I think are worth studying more.
They were of interest for various reasons. One was called Apolipoprotein D — it's a hormonally-associated protein that has been associated in the past with ovarian cancer and a good prognosis of breast cancer. Another protein was called Sp-alpha. It's known to be an inhibitor of apoptosis. Alpha-enolase was other protein that's involved in glycolysis. Transketolase is known to support tumor cell survival and proliferation. And M2-PK is involved in glycolysis in early fetal tissues, adult stem cells and some neoplasms including colorectal cancer.
What are you looking to do with these proteins now that you've identified them?
I need to figure out an ELISA test to determine whether they're elevated in women with ovarian cancer, and then see if it might be helpful in screening. Then of course it would be interesting to do functional studies to see if these proteins are produced by ovarian cancer cells or determine whether they're some sort of response by the body to ovarian cancer cells.
Do you have a cohort of women that you're planning on testing?
Our lab actually banks ovarian cancer patients, so those are the samples that we're using. We have hundreds of samples banked, and what we do to do the validation is we pull out at random a mixture of different ovarian cancers and controls, and see what we find.
Do you have any preliminary numbers for sensitivity and specificity of your biomarkers?
I do for the M2-PK. I think for the preliminary study I looked at about 48 cancer patients, and I found that combining M2-PK with CA125 when I set the specificity at 98 percent, I got a sensitivity of about 95 percent.
Would you look to commercialize a test that you developed?
It's certainly not my main goal, but ultimately I think a test would need to be commercialized. It's not my personal goal, but the ultimate goal is to have the test be commercially available.
How good would the test have to be in order for it to be used in the clinic?
That's a good question. That's a question for a statistician really, not for me. There's a lot a debate as to how good it has to be. Clearly, we need a very high specificity — approaching 100 percent, but at least 98 or 99 percent, with a good sensitivity. I don't really want to give you numbers because I don't think that we really know what the numbers are. We have to better understand what the test means — in other words, are we looking for a test to screen everybody, or only women that are post-menopausal, or a test to screen people who have a family history of cancer? Based on your population, I think your willingness to accept a certain sensitivity and specificity is going to move.
