Скачать 1.21 Mb.
DR. MICHALEK: Right now.
DR. STOTO: So I think the question is not whether or not they should do a matched analysis; the question is whether or not the matched analysis should be incorporated as Model No. 5, something like that.
DR. MICHALEK: In some clinical areas, maybe not others.
DR. STOTO: See, that's one of the drawbacks. I don't like the fact that they sort of have this canned approach of doing exactly the same analysis for every outcome. It would be funny to kind of do it in some and not in the others in these blue books.
So I guess I would be in favor of doing, having these guys having the Air Force continue to do that analysis in their own publications, but not necessarily making it part of the .
DR. CAMACHO: Where are you now, 22.214.171.124?
DR. STOTO: No.
DR. HARRISON: What we're talking about --
DR. STOTO: No, it's 3.6.1.
DR. HARRISON: What we're talking about, though, is whether or not it would be desirable in some cases to do a matched analysis or whether or not it would be desirable to add to the scope of work a matched analysis in some of the categories.
So you can just kind of pick any section.
DR. GRUBBS: To paraphrase a little bit of what Dr. Stoto is saying, there's benefits and deficits to it; of course I think the question is, we go by the Statement of Work as to what should be in our voluminous 1800-page report. Is that the appropriate place for it, or is it end papers, or supplemental documents?
DR. MICHALEK: Personally, I tend to feel that's really not the place for it. Because we generally are motivated to do a matched analysis based on a referee comment to a submitted article to a journal. "Okay, you did the logistic regression, that's fine, but please do it this way in addition, and summarize the results in a sentence." The main body of the article will have the logistic regression modeling results, and a sentence in the text will be the results of a matched analysis.
To cover all of the concerns people would have about a possible loss of adjustment that may have crept in to the multivariate modeling.
Another drawback to doing that matched analysis in the big report is that if we do that, we'll have no frame of reference relative to the previous reports, because we've never done that previously in a big fat report. We've done it in articles, but not in the reports. So you wouldn't be able to track that backwards to other findings from previous physicals.
DR. GOUGH: Why are we talking about this?
DR. MICHALEK: Well, because it's an issue.
DR. HARRISON: It's a --
DR. GOUGH: Well, I brought it up because it's written here, but it seems to me--
DR. HARRISON: It's in red because it's a red herring.
DR. STOTO: No, I think I was good to have discussed it and decided not to add it.
DR. GOUGH: All right, to continue as before.
DR. HARRISON: Okay.
DR. MICHALEK: That's exactly right.
DR. HARRISON: So the next item then is multivariate analysis.
What's the question there?
DR. MICHALEK: Well, some people, and I am included, have the feeling that we may be missing an effect. Because the reports that we have been doing are concentrated one clinical area at a time. And that we know that many of these areas are related, such as general health, cardiovascular and diabetes.
Individuals in general health report fair or poor health, and we'll find out through subsequent analyses, "Well, they're really telling us about their diabetes." And individuals who are having heart attacks are also having diabetes and so that the outcome in our cardiovascular chapter is related to the outcome in our endocrine chapter; and that's related to our outcome in general health.
We've never made an attempt to put it altogether into a multivariate analysis, and that's actually, we've worked with Michael and Bonnie LaFleur, and we gave Michael a dataset, and Bonnie's going to be looking at it from that point of view.
So the issue is where we should include that kind of multivariate approach in the next report.
DR. HARRISON: And again, that suffers from the same problem as before; it lacks the continuity that the study has now in terms of how it's analyzed this data.
DR. STOTO: And it's not the kind of thing that you want to just crank out.
DR. MICHALEK: No.
DR. STOTO: You need to think very carefully about how to structure that.
DR. MICHALEK: Now you're putting your finger on a kind of quandary we're in, in that we only have a few number of years left; that we who have been looking many years at the data have the feeling we may be missing something; and this is sort of the thinking we've had.
This kind of modeling is fairly complicated. It's not a simple presentation of the results even when you're finished. It's a complicated presentation of results, because you've got inter-correlations between the endpoints, you've got correlations between the endpoints over time, and you've got interrelations between the risk factors and the endpoints. It's a difficult analysis to summarize; and on top of that, it requires a certain amount of exploration.
This company will not be in the position of doing data exploration; we can't afford it. They have a fixed, 13 month time period to write this report, from the middle of September of the year 2003 until 13 months later. That's the typical period, because we have a time frame for release of the report to the public. Those are the issues.
DR. HARRISON: Joel, I wonder, though, if you haven't already solved the problem to some extent. A lot of the Ranch Hand data has been published by you and your colleagues; accessibility of the Ranch Hand data is more or less open --
DR. MICHALEK: Complete.
DR. HARRISON: and if there's anything that academicians love is being able to use somebody else's data to write a paper.
DR. MICHALEK: Good point.
DR. HARRISON: So I really think the kinds of analytical analyses that you're talking about are the kinds of things that people are already interacting with you about, and that may be the answer to the question that you continue to you know, the Statement of Work continues to crank out this sort of --
DR. STOTO: "Cookie cutter approach."
DR. HARRISON: general approach. That's difficult enough to do.
DR. STOTO: Right. And it's important to do it in a consistent way all the way through.
DR. HARRISON: Exactly. And then these other innovative approaches can be some of them you'll have the wherewithal to pull off yourself, and some you'll find partners to pull off with.
DR. MICHALEK: For the record and you already mentioned that, Dr. Harrison all of our data is released to the public. There is nothing left to release; it is available on our web page.
DR. HARRISON: I really apologize for even trying to temporize it in any way.
DR. MICHALEK: I want to make that crystal clear to anyone who is in this room, that all of it is on the web page and all of it is with the Government Printing Office.
DR. HARRISON: But there's always some issue in terms of releasing this data in a way that would allow identification of subjects.
So as I recall you've looked at it in the past and been careful about how that data is available.
DR. MICHALEK: The names are not there we've protected privacy. But all of the end points and clinical measures and laboratory results, all of that are on the web page.
` DR. HARRISON: All right, what about interaction models?
DR. MICHALEK: Well, again, this is part of our suspicion that we may be missing something. We run what are called main effects models in all these reports: Is dioxin related to health adjusted for body fat, age, and family history?
Well, a piece of the model that we don't put in is, is there a change in the relation between dioxin and health with age? In other words, is there an age by dioxin interaction? That analysis has not been done, and that's the kind of analysis that could be done by any other researcher because all the data is on the web page. To require SAIC to do that would enormously increase the cost of the effort and the timeline, too.
DR. HARRISON: You know, another thing that just crossed my mind, Joel, is that there's no rational physiological basis for what you're looking for here, so you're trying to design statistical strategies to reveal truths that you're not certain are even there.
DR. MICHALEK: Exactly right.
DR. HARRISON: So once again, I think you have to decide that you're limited at some point.
DR. MICHALEK: Exactly right. In fact, the attitude of many journal referees and editors is that you're not to invoke an interaction model unless you have good a priori biological reason to do so.
I only can identify one place where there might be a biological rationale; and that is that in the nondiabetic Ranch Handers, we see a strong dioxin effect on insulin. We don't see it in the diabetic Ranch Handers. There could be a biological basis for that, but that's maybe the only place in the whole study where I've ever seen anything that may have a biological rationale for an interaction.
DR. STOTO: I think the bottom line is the same; is that this is the same as the two previous ones.
DR. HARRISON: That's really the bottom line with all this stuff.
LTC BURNHAM: The reason we're coming up with these ideas and presenting them is because we've had the attitude of "this might be the last chance." "How can we make this the best exam ever?" And so we're just brainstorming and if you think they're all not worth it, that's fine.
DR. HARRISON: No, The issue is not whether they're not worth it; you misunderstood me if that's what you thought I said. It's simply that it's different objectives as I see it, and I've already said that I'm not sure that I really understand any of this is that the health study is a longitudinal health survey; and as such, without what I would see as a physiologically-based hypothesis to start with, you really are constrained in what you can do with that kind of a study.
It looks like there might be some relationships, and it would be nice to try to figure out a way now to design one last go-around that you could prove this. But I don't think that's feasible within this context.
I think the question should be how can you finish not finish this off, but how can you maintain the continuity and strengthen what you have?
DR. STOTO: I would answer it differently; I would say that the kind of question you bring up is an important one with respect to doing laboratory tests and so on, because it presumably is the last time we're going to see these men. But these analyses can be done later.
The fact that we don't include them in this Statement of Work doesn't mean they can't be done. And I think they probably can be done better with a different arrangement.
DR. HARRISON: You retire from the Air Force, write a grant, and .
LTC BURNHAM: But that's a big assumption, too, though. A lot of times, if you don't do it now, it never gets done, and that's the plain truth of it.
DR. HARRISON: Well, that's true.
LTC BURNHAM: Because we have funding through '06, but that may never happen again.
DR. HARRISON: The other part, I think, I'll bet you that it makes a whole lot of sense, is that for some of this you're going to be a lot more efficient doing it yourself than trying to do it within the scope of work and trying to do it in all of the categories that's, that would be my .
Do you have anything to add, Paul?
DR. CAMACHO: No, I'm getting a little weary. I think the argument comes down to the continuity with the past studies to match, and taking some new tests that you could analyze in the future or you could preserve that might be analyzed in the future.
DR. HARRISON: I agree.
What about administering the spousal--
DR. GOUGH: Can I ask a question.
Joel, have there been requests for the data?
DR. MICHALEK: Not since we put it on the web page.
DR. GOUGH: Were there requests before that?
DR. MICHALEK: In the past we have received Freedom of Information Act requests from the Veterans of Foreign Wars, for example, from other Vietnam veteran organizations. And we've always handed them diskettes.
Now since everything is on the web page, we haven't --
DR. HARRISON: Do you know how many hits you got?
DR. MICHALEK: We keep track of the hits, but I didn't see a big shoot-up in the number of hits since we put the data up.
DR. HARRISON: But has the data been downloaded?
DR. MICHALEK: We can tell that, but we haven't checked our webmaster to see how many times a day it has been downloaded.
DR. HARRISON: Has it been downloaded?
DR. MICHALEK: I don't know. We'll have to check and see. It's only been up there for two weeks now.
LTC BURNHAM: Well, some parts of it have been up --
DR. MICHALEK: Right. Some of it has been up there for about six months.
DR. HARRISON: Did it get archived in the various search engines? Was that a part of your deal?
DR. MICHALEK: In other words, can you reach our web page from Yahoo and the others?
DR. HARRISON: No, I mean, normally when you have somebody make a web page for you, they actually submit your web site with the meta tags to the various search engines; Google, Yahoo Yahoo is Google now Northern Light, and so on, so that if somebody plugs dioxin in the day after your web site goes up, your web site should hit.
DR. MICHALEK: I don't think that's been done.
LTC BURNHAM: No, I believe so. I don't know how extensively, but it has been.
MS. YEAGER: It does hit.
DR. MICHALEK: It has been done.
DR. GOUGH: I've been somewhat disappointed since we began, the Air Force began making these data available in 1990, that they haven't been exploited more by outside academics.
And I think part of the reason is simply people don't know about it. In terms of knowing about what's going on with middle-aged, largely white guys, there's nothing compares to this. And I think it's under-used.
I don't know what we can do or what the Air Force can do, but I certainly do think that by the year 2006, I would hope a lot of people are using it or are at least aware of it.
DR. STOTO: We discussed this, I think, two meetings back before you were on the committee, and discussed the idea of trying to get the word out to people who get support from NIH about the availability of this resource, presumably if people knew about it they would be able to put in RO1 applications that NIH would be willing to fund out of its own resources if people can put together a compelling argument, which I think that they certainly can.
DR. HARRISON: Well, remember we wrote the letter, and --
DR. STOTO: We wrote the letter, but that got misinterpreted as trying to get NIH to set aside money for this analysis; and I don't think that's likely to happen.
DR. HARRISON: But it was hoped that NIEHS would use its review capabilities to decide who should be helped. The problem is that the Air Force has no way to review proposals; so anybody with a --
DR. GOUGH: I wasn't thinking about Air Force review now; I was thinking more of the general PHS review process.
|Advisory committee on immunization practices||Medical Devices Advisory Committee|
|Veterinary medicine advisory committee||National Vaccine Advisory Committee (nvac)|
|External Advisory Committee on Cities and Communities||Wildlife Diversity Policy Advisory Committee|
|Peer reviewed by the Arizona Department of Commerce Economic Research Advisory Committee||Food and drug administration national institutes of health advisory Committee on: transmissible spongiform|
|Advisory Committee, Cuyahoga Valley School-to-Career Consortium, Broadview Heights, Ohio 1996-2002||Jane D. Siegel, md; Emily Rhinehart, rn mph cic; Marguerite Jackson, PhD; Linda Chiarello, rn ms; the Healthcare Infection Control Practices Advisory Committee|