Study Of AIDS Statistics Hinges On Debate Over Methods, Politics

WASHINGTON—At the National Center for Health Statistics in Hyattsville, Md., James Massey and his colleagues are holding their breath and waiting with fingers crossed for word from Dallas. Starting Sept. 26, researchers will be knocking on doors to collect blood samples and information about possible risk behaviors for AIDS from 1,600 Texans. The results of this pilot study will determine the future of the National Household Seroprevalence Survey, a planned $25 million project to canvass U.S. neighborhoods in an effort to provide the nation’s best estimate of who may develop AIDS. “If we have a failure in Dallas, it will mean the end of the survey,” says Massey, chief of survey design at the center. He is not the only one keeping a close eye on this project. During the past five years, statistical and mathematical experts used to working backstage are finding themselves blinded by the AIDS spotlight. They are designing surveys, making projections, and evaluating treatments. Policymakers are depending on them to forecast the course of this deadly epidemic. “The starting point for any plan has to be the number of cases,” says John M. Karon, AIDS statistician at the federal Centers for Disease Control in Atlanta. As with so many AIDS-related projects, the survey has already drawn fire from the public and raised eyebrows in the statistical community. Statisticians worry about its multi-million-dollar cost and fear that changes imposed on the project’s original design by political considerations will undermine the usefulness of the results. And they are troubled by recent congressional action that effectively killed another survey— of sexual practices—deemed by many to be crucial to planning an effective anti-AIDS strategy. The two surveys are meant to help statisticians and mathematicians fill in the blanks in their equations for predicting the course of the epidemic. But more than numbers are at stake. Many observers of the science and medicine communities see this survey not only as providing data essential to effective anti-AIDS strategies, but also as a proving ground for a new breed of statisticians who are much more outspoken and attuned to social and political issues than their predecessors. Scientists among this new breed have begun to forge alliances with other segments of the health community: with physicians, who want to know as quickly as possible whether a drug, vaccine, or educational effort is effective; with AIDS activists, who want to speed up the testing of potentially effective drugs, forcing analysts and designers to find ways to obtain useful results more quickly; and with medical researchers, who depend on statisticians to determine when the results from a clinical trial are adequate to make a decision. “We could easily go wrong,” says John P. Phair, professor of medicine at Northwestern University Medical School, who has found that the long duration and deadly outcome of AIDS has made him much more aware of the importance of statistically valid study designs. “If you take volunteers and ask then to commit for a long time, it’s very important that the results be meaningful.” So much is riding on those projections that researchers wind up in the public’s eye. “Statisticians have moved into the political forefront,” says Reynolds Farley, professor of sociology at the University of Mich igan Population Studies Center. “Before, such [statistical] issues might have been argued in scholarly journals. Now results are as likely to be on the front page of the Washington Post or the New York Times.” Also, pressure from government agencies to try to understand the course of the epidemic has revved up the research. This year’s big research papers will be gathering dust by year’s end because the field is changing so fast. Some statisticians see these changes as long overdue. “I think our opinion is as valid as any doctor’s or lawyer’s,” says Philip Smith, associate professor of statistics at the University of Minnesota in Minneapolis. “Statisticians should play a greater role in science policy. We should be more than people who just know the right formulas.” Recent past presidents of the American Statistical Association (ASA) has urged members to stop taking a back seat in science policy issues. In 1987, Barbara Bailar pushed to set up a public affairs of fice for the society. And there are those on the association’s board of directors who would like to-see the organization become even more active in shaping governmental agendas. But not everyone agrees. “There are a lot of people who are very worried about becoming too activist,” says Bailar, now ASA executive director. “We have the image of a learned society, and some do not want to get caught up in the issues of the day.” Many statisticians are not comfortable with this new requirement of outspokenness—their “mathspeak” being a protectively arcane language that sets them off from other disciplines. Statisticians themselves admit that they are more savvy about numbers than about working out compromises on emotionally charged issues. As well as being deadly, AIDS is a particularly baffling and headline-grabbing disease. It thus presents statisticians with a variety of new challenges, requiring them not only to find new ways to probe the sensitive subject, but also to accustom themselves to seeing the results to their investigations immediately publicized. “In this environment, we do not have the luxury of designing the perfect study or the perfect clinical trial,” says Ron Brookmeyer, associate professor of biostatistics at Johns Hopkins School of Hygiene and Public Health. Recent events highlight the role politics can play in the statistician’s science. This year Congress refused to finance a $15 million survey that would have asked 20,000 Americans about their sexual behavior, even though that information is crucial predicting the course of AIDS and developing ways to stem the spread of the virus. Also, a pilot seroprevalence study set up for the District of Columbia was canceled after the media portrayed pollsters preparing to show up on doorsteps, needles in hand, to draw blood and to ask residents some very personal questions. New York city officials, for their part, refused to participate in a pilot survey. ‘We’ve been really surprised at the reaction to what we consider standard statistical procedures,” says Massey. “This has been a real experience for a lot of us who usually go about our business routinely. This has been anything but routine.” Daniel G. Horwitz, whose Research Triangle Institute in North Carolina is coordinating the national seroprevalence survey, has found surveys about AIDS to be much more challenging. For example, community representatives have insisted that people who initially refuse to answer questions be called back only once, rather than repeatedly. In addition, community leaders objected to the idea of focusing on areas where one is more likely to find people infected, saying that the survey was then discriminating against certain groups of people. The seroprevalence survey now going through the pilot phase in Dallas is designed eventually to collect blood samples and questionnaires from 50,000 adult volunteers, ages 18 to 54, living throughout the country. However, because it attempts to find a relatively rare disease in the general population, the project may in fact turn Out to be too expensive to carry out as originally planned. In addition, some observers contend that results may be biased because of the controversy surrounding the disease, which has increased the probability that people potentially at risk will refuse to participate. Despite these obstacles, Horwitz isn’t discouraged. “We would be making a mistake if we blindly fought to do things the way we did [things] for nonsensitive issues,” he says. “We can design our studies recognizing those constraints and still be effective.” Some of his colleagues, however, are not so sure. “It’s a very controversial idea,” says Susan Ellenberg, chief of the biostatistics research branch of the AIDS division of the National Institute of Allergy and Infectious Diseases. Speaking about the attempt to survey a representative sample of the public, Ellenberg says that “some statisticians think it’s hopeless.” Skeptics say that this house-by-house count will miss such important groups as the homeless and intravenous drug users. The people most apt to be infected, they argue, are least likely to want to participate for fear of being stigmatized. Minnesota’s Smith worries that not enough is being done to deal with bias or problems created when people refuse to participate. “That’s a failure of this study,” he says. “They [the survey designers] have not convinced the communities that they need to cooperate.” The survey is so anonymous, he notes, that it will be difficult to link a blood sample to a respondent’s questionnaire and impossible to do a follow-up survey at a later date. “I think they have really compromised their science in the interest of confidentiality,” he complains.. “It will not be as rich a source of data as it could be.” The results of the Dallas survey due next March, will help health officials decide whether to take the giant step into a nationwide poll. But whether this giant step is indeed taken, the fact is clear that the nature of the statistician’s discipline has already undergone a profound and permanent change. “The statistician cannot work in a vacuum,” says Horwitz. “We must put our minds on a way to address the political constraints. Then better designs will come out.”

Study Of AIDS Statistics Hinges On Debate Over Methods, Politics

Interested in reading more?

Become a Member of