Posted 21 April 2005
Interviewed by Tom Fagan
Walter (Bud) Kukull is the principal investigator of The National Alzheimer's Coordinating Center (NACC), which facilitates collaborative research among the approximately 29 Alzheimer's Disease Centers funded by the National Institute on Aging (NIA). He is professor of epidemiology at Washington University, Seattle.
NACC also maintains a database of information collected by the Alzheimer's Disease Centers regarding patients with Alzheimer disease and controls.
ARF: Tell me a little bit about the NACC and how it came into existence.
BK: In 1998 the NIA issued a Request For Applications for some group to take over data management of a common data set from all the NIA-funded centers, officially 29, but there are a few others that were on no-cost extension. In the year or so prior to this, an interim data center at Rush University Medical School ADRC had been established, and it housed a minimum data set of about 60 different elements. The minimum data set was agreed on by the ADC directors in the year prior to its establishment at Rush, and was to be a retrospective accumulation of data from the beginning of the AD Center's program in 1984. We competed for that RFA and won the privilege of maintaining and expanding that data set. Then we set up the database based on those original elements and made it a little bit more accessible. We then went to the work of cleaning up the database because by the time we took over in mid-1999, the Center's program had been going for 15 years, but the centers had joined in different waves over that period of time, so they each had different amounts of data. Originally there were a lot of missing data, so we had to do extensive cleanup in working with the centers, filling in [missing data] retrospectively.
ARF: So when you started this, what was the biggest challenge you had in assembling all of this data?
BK: I think going back and getting additional participation from the actual clinical people and the directors. When it was originally set up, it was sort of an agreement between data managers and the data center, computing people, and so on, to just put together files on a set of variables and send them off to the repository. It turned out, a lot of times, that if the data managers didn't necessarily know whether something in their files equated to one of these elements that was in the new database, they might just fill it in as missing if they didn't have easy access to a clinical core with whom to verify that the elements were equivalent. There wasn't very much clinical validation of what was going on in a lot of the places, so we tried to go back a little bit more and tried to get the clinical people more involved to get rid of a lot of the missing data. We did pretty well with that, with everyone's help.
ARF: What about standardization? So you're dealing with 29 or 30 different centers; had they already developed some kind of standards?
BK: That was probably the stickiest point, because the ADCs had been going for years, collecting data however their individual protocols directed collecting it, and so what each ADC was essentially doing was kind of mapping in their data, approximately, into whatever data elements were asked for. It was retrospectively done, primarily, so that was kind of the stickiest point.
Let me move on to another little bit of a level on that. In the initial set, there were only one or two items that dealt with neuropathologic diagnosis, a sort of general overall diagnosis, and we worked with the ADC Neuropathology Core Leaders' committee to establish a parallel, expanded neuropathology database. So, with their content input, we constructed the database and got the neuropathologists involved really very well. ADC neuropathologists went back and gathered or recoded the data from their individual centers for almost all the cases that they had entered at their center. So that made another particularly nice chunk of data; it really added to the research capabilities of the database.
During the last couple of years, the NIA decided that we should also be doing an expansion and standardization of the clinical data in the same way we expanded the neuropathology data. So NIA formed a Clinical Task Force of various ADC directors and Clinical Core leaders, chaired by John Morris of Washington University. We have been working closely with them, and we're at a point now where we're going to be, well, we're almost at the point of getting final approval of a more detailed set of clinical data that will be standard across the centers and will be collected prospectively. So that will yield a huge improvement in both the data and the kind of studies that we can do. Hopefully we'll get that going. We just did another draft with some forms for the task force and they're going to be going over them shortly, and then we're hoping that we can ultimately get this started some time around the fall of this year.
ARF: So what's the scope with the 30 centers? How much data on how many patients are we talking about?
BK: Overall, right now the cumulative number of subjects (patients and controls) is close to 70,000. For prospective annual enrollment the numbers vary from center to center. Some might enroll between 50 and 100, maybe up to 200 patients a year, plus some number of controls. Control description and definition is another thing that we'd like to work on with them. It would be nice to have a little bit more uniform and careful definition of "control" subjects. It isn't always appropriate to assume that anyone without AD (or the disease in question) will be an adequate comparison subject.
ARF: In the database itself, how many different parameters are you recording? How much information is in there?
BK: In the minimum data set itself, in the clinical side, there's around 60 elements. It's pretty basic. I'm not sure exactly how many in the neuropathology side; there's probably about 50 elements in that data set.
ARF: Right. So in the clinical set, what kind of parameters are you talking about?
BK: Demographics like, age, race, language, and education; clinical information like when they were evaluated first; their Mini-Mental score; initial clinical diagnosis; if they were not demented, what kind of other diagnosis they had; what was their age of onset; whether they had some other kinds of conditions, like if they had a non-Alzheimer dementia, what that might be, or whether they had stroke or Parkinson's or depression, delirium; whether they had relatives with dementia; family history; whether they've died and had an autopsy.
ARF: So they are the sort of things people would use to do an epidemiology study?
BK: No, it is not well-suited for classical epidemiological risk factor studies because these are largely samples of convenience, and exposure data prior to disease onset is not often available. Some clinical epidemiology studies looking for prognostic factors for disease course can be addressed rather well, however. Probably more often the data would be used to get an idea of what kinds of cases and biological specimens are available for research. We would be able to link data with whether they also had an autopsy and have a brain specimen stored, or whether they have CSF or DNA, or something like that stored so that biomarker studies could be possible. So people are interested in that because then they can link back, or ultimately go and contact the center and see if they can have access to specimens. We don't have information on how the specimens are stored, or how much DNA each center has left from what was stored originally, however.
ARF: In the best-case scenario, how do you think this is going to facilitate and help drive the research?
BK: Well, I think the biggest payoff will come from the new initiative of this uniform data set, this new standardization, because up until now, even though the centers have been around since 1984, they haven't had a standard protocol for diagnosing or collecting data. They've been doing things roughly the same way, but they haven't been following a protocol, basically. So I think that will be the biggest contribution: to ensure that standard data is being collected and collected prospectively. This will be more like a longitudinal assessment of all the people enrolled, because there will be regular follow-ups. So that's going to be very important in studying differential disease progression and potentially also progression from normal to mild cognitive impairment (MCI) to dementia, the real thing. What we've been doing for the last five years is pretty much trying to re-gather or re-collate the data that was collected in the previous 15 years, and bring us up-to-date with the new cases that are coming in. Now, we move forward.
ARF: So, how do people use the data?
BK: A couple of different ways. One is to identify throughout the centers where there might be cases of a particular diagnosis, for example, if researchers wanted to find out where they might collaborate with someone who has additional, say, FTD [fronto-temporal dementia] cases or some other particular kind of case. Or someone might want to take a look generally at what the age of diagnosis was across the centers just to get a picture of how the people who are enrolled in the centers might fit in the overall distribution of what we know about the age of onset distribution of Alzheimer disease. However, because the people who are enrolled in centers aren't strictly any kind of statistical sample of the universe of dementia cases—they're just people who either present or are referred to ADC—researchers need to be a little careful about inferences they draw. I think, on the research side, there's a little bit more than just looking at the data and getting an idea of where people might collaborate, or using it as a gateway for generating other studies. The way we have this database set up now is so that online, people can get data from both the neuropathology and what we call the MDS, or the clinical side, to form files and look at the data. Because we have a substantial amount of neuropathological data, clinical-pathological correlations can be examined with much larger samples than usually available at a single center. We have about 6,000 autopsies, but there are differing amounts of data associated with each of those, because again, everything is retrospectively filled out, so sometimes, for example, all the vascular information isn't totally complete on all of them. If you do the intersection of 12 different variables, you might end up with a small number of subjects that fit your desired criteria.
ARF: So, to avail of the data, do people have to register or be vetted somehow?
BK: Yes; anyone who's an investigator at one of the NIA-funded AD centers can have access to the data. The investigators have to obtain approval from their directors, call in to us and get a password so they can manipulate data and form spreadsheets, or have us form analysis files for them. We also have a public access set of data that people can get at online. It doesn't cover everything that's in the center-oriented data, but it's almost everything. It's just de-identified. The center-accessible data fits the definition of what's called a research limited data set under HIPAA (Health Insurance Portability and Accountability Act), so the public access data is de-identified to ensure they can't identify subjects.
ARF: Do you find people use it as a means of, say, tracking down particular samples that they might be interested in obtaining?
BK: To some extent; within the centers that happens. We usually are sort of a middleman for that. But within the centers they can identify which centers might have particular groups of cases, but they don't have access to individual patient numbers or identifiers at a center. Researchers must make individual contact with centers to obtain permission to gather any additional data held at the center itself.
ARF: So, in addition to the database, are you in the business of developing tools to analyze the information?
BK: We give consulting help to anybody from the ADCs who asks for it, basically, as far as statistical help or design help, or that sort of thing. We can help them understand how the data has been collected and that helps them decide what kind of questions they can ask.
ARF: And what are the goals regarding prospective studies in the future? Do you have funding for this?
BK: We just recompeted and were funded for another five years beginning this last July, so we'll be doing that.
ARF: And this is funded by the NIA?
BK: Right. It's a cooperative agreement, a U01, so NIA has a lot of say in what we do. It's not strictly like a contract, though, so we have a little bit more leeway, I guess.
ARF: So did you have your own in-house database going at the time that you applied for the initial RFA?
BK: Well, we set it up specifically for this particular project. We got additional hardware and so on, we had people here, including myself, who have been working in AD and dementia studies since the beginning of the centers program in 1984-1985 and could understand and do the kinds of things necessary to mount the operation. As part of this award, we're also supposed to be giving about half of the award amount back to the centers to help them do the work of getting the data to us.
ARF: So you're channeling funds, as well?
BK: Yes, we provide each center a flat amount that they can use to help with their data efforts. We also decided to form a pool of funds for competitive collaborative projects so we could fund collaborative studies among the centers. They could decide to gather whatever data they wanted or could use data from our database and address particular problems, then we would fund them. In the initial years we funded two or three of these at about $300,000 for a one-year shot. We built a scientific review committee, comprising people from the centers, to review these projects so that we were kind of separated from it. The NACC steering committee ultimately decides on funding, not us, so we have no conflict there, either.
ARF: Now, is this something for which people have to submit a proposal?
BK: Right; it is similar to a regular NIH-type application and then it gets reviewed. It's just like a regular peer-reviewed process.
ARF: So are these ongoing? Are they still available?
BK: Yes, they are. Because of budget cuts and everything, we have less money this year than in the past. We're going to issue another one pretty soon—the applications will be due probably in February or March. We may fund one or two this time; it depends on what sort of money we have.
ARF: And what kind of proposals are you looking for? Are there particular research goals in mind?
BK: Not specific topics. We did have, in previous times. We do have a steering committee made up of center directors, too, and sometimes they'll indicate particular areas that they would like to see some proposals in, but so far they haven't done that for this cycle.
ARF: So if I were a researcher and I wanted to submit a proposal, what kind of leeway do I have? Could it be basic, clinical…?
BK: Primarily it must be collaborative (involve at least three ADCs) to accomplish something that could not be easily accomplished at one center alone. Content could be pretty much anything having to do with AD and related disorders, and the review committee will look at it for basic scientific merit, how it might fit with where the field's going, and what's been done before, and those sorts of things; they'll score it, and then, ultimately, the NACC steering committee will choose the best, the most well-scored projects, but they could reach down and grab a different one if it was a particularly hot item or something that was very closely scored.
ARF: How much are these grants worth?
BK: They're for either one or two years, but for a total amount of about $300,000. It's meant to be a feasibility study, pilot kind of thing. They can use this to then develop an R01 or something that they can move on with, so it's not necessarily meant to just answer a question. Usually people stretch the money over a couple of years because it generally takes them longer than they think, anyway. It seems to be working well. We're also thinking about maybe some small ones this time around, like some $20,000 to $50,000 ones for people who just want to identify a particular question primarily from the data we already have in the database and address that—specifically secondary data analysis targeted more for junior faculty.
ARF: So, did you say that you had awarded some of these already?
BK: During the last five years, we've done about 12 of them (see NACC website).
ARF: What is your own background?
BK: I'm an epidemiologist, and I've been working with Alzheimer disease since about 1984 or so, and with the center here, and I also had a couple of other grants, one called the Alzheimer's Disease Patient Registry, with Eric Larson, and then I wrote another case-control study based on that and kept that going for a long time. The Alzheimer's Disease Patient Registry is still being renewed. It first came online in 1987, and it changed from a registry into a cohort study that we got started about 1992, and it's still going as a cohort study, but I'm not very much involved with that right now. I'm just a little bit too busy with NACC and other things.
ARF: Do you have your own pet theories about Alzheimer's?
BK: We've made a lot of progress on the genetics side in the last so many years. I think that's really good. I think right now we haven't made a lot of really good jumps in terms of environmental causes. I guess, if you want to call homocysteine maybe environmentally oriented, I suppose that's possible, or diabetes or some of the potential drug associations, but we haven't been very good at identifying other environmental factors that probably play a strong role. I think there could be a lot more work done on that, but it's been difficult. From time to time there have been little bits of things about occupational exposures, or there was some indication that some metals like copper or zinc might be involved, but usually the genetics gets the most press. It's also harder to do big environmentally oriented studies. Identifying population groups to study, diagnosing cases accurately and completely, measuring "exposures" or risk factors (including genes) carefully and accurately, and linking those exposures to critical periods in disease pathogenesis, is how, in a nut shell, it should be tackled. That's analytic epidemiology, too, I guess.
ARF: Is this something you think the database could cope with?
BK: Well, this one will be more oriented toward, I guess, progression of the disease because people will enter when they're already getting diagnosed, although we'll probably have some control people who will be followed and then convert. That side might help us with some risk factors. I guess, the risk factor side has got to come from the big cohort studies that are out there. I've always thought that it would be really nice to figure a way to combine those, or launch one from across the centers, where, say, every center sampled several hundred people without dementia and then followed them so we'd have a built-in risk factor study, but right now, that's kind of pie-in-the-sky. Ultimately, someone will see the wisdom of it.
ARF: Is that difficult because it's just too expensive?
BK: A little bit of that, a little bit because people have their own agendas of what they want to do and so on, and don't necessarily want to launch a big thing like that unless…maybe if it could be launched as part of what the centers are now doing. I think if it was separately funded, then people would probably get on the bandwagon, but to just do it as an extra project, to try to work it in with their ordinary funding, it's not so appealing.
ARF: Right. Is there anything we haven't covered that you'd want people to know about the centers?
BK: We do a lot of public service. On our Web page, we have a lot of links to different sites, and so on, and that's probably helpful. We do try to coordinate and help with that end of things, and we do other kinds of consulting with people at the centers for design and analysis on different projects they're thinking about doing, and so on, just as part of our job.
ARF: Now, presumably all the researchers at the centers are probably well aware of your role and what you do. What about those people who may be working independently?
BK: They can get at the public access data easily enough, and I guess they could contact us if they wanted us to be a collaborator just like anyone else would do with an R01. That hasn't strictly happened very much except for the collaborative project RFAs that the NIA has put out, but they wrote us in there as a required participant, basically, to provide data management and statistical consulting.