The last patient on the last day. Mid-40s. In pain. A pelvic exam showed a lot of blood, a large mass, all the indications of late-stage invasive cervical cancer. "It was like a sign," says Goldie. "The last night I was there, Paul said, 'Okay, Sue, do you still think we can do this?' I think he thought I was going to give up. And I told him, I am so much more committed. I have no doubt in my mind." Even before she got home, Goldie was making plans to conduct a pilot program in August. Eventually, she envisions a massive, once-a-year outreach program that would screen 100,000 women annually. "I am literally working on four different grants, and I am going to get $3 million to do this program, come hell or high water," she says. "I've even figured out how many friends I could ask for $1,000 donation each. That's my backup plan!"
Public health attracts more than its fair share of idealists who are good with numbers, but few have the indomitable will and sheer force of personality of Sue Goldie. The 40-year-old, Yale-trained doctor is one of the stars of decision science and risk analysis at the Harvard School of Public Health. She publishes widely, gets rave reviews from her students, and is the principal investigator on several grants, including an ambitious five-country policy analysis of cervical cancer prevention programs funded by the Bill and Melinda Gates Foundation through the Alliance of Cervical Cancer Prevention. Her work is arcane even by public health standards--all about quantifying tradeoffs and hard, health care choices. The toolbox is computer code and research data; the finished product, dense research papers in medical journals and dry policy recommendations. But, paradoxically, the person producing all of this cool-headed abstraction is labeled passionate by the people who know her. Goldie even describes herself as someone who is hit with epiphanies (like that "last patient") and who isn't afraid of barreling ahead when others would say it's impossible. "The things that my heart is in, I don't let them sit," she says.
fabulous. Your article ought to say somewhere that she is a star,"
says Thomas C. Wright, associate professor of pathology at Columbia, who
has frequently collaborated with Goldie. "You meet her--and you know,"
says Serena Koenig, medical director of Partners in Health's Haitian program,
who was in Haiti with Goldie. "She is one of the most remarkable
human beings I've ever met because she is that unusual combination of
being utterly competent and brilliant and focused on doing exceptional
research, while at the same valuing patient care above all else."
She credits her difficult childhood with making her dogged. In another life, Goldie might have been a hard-charging CEO or a political star. "I have never taken no, or that something can't be done, as an acceptable answer," Goldie says. "So many things that seem undoable to other people seem so doable to me, because compared to my childhood and just trying to get through normal life, they seem relatively easy." Wright says good-naturedly, "Only Sue could have gotten me to go to Haiti in August. It is one of those things about her. She is very convincing."
After graduating from the Albany Medical College in Albany, N.Y., in 1988, Goldie did her residency at Yale. She was drawn to the study of infectious disease ("that is where you learn public health in medical school") and adored both education ("teaching is really my first love") and delivering care ("I was passionate about not becoming a specialist--I just wanted to take care of patients"). But in 1995, a month at Stanford led to an epiphany, and an unpremeditated swerve in her career path. The faculty development program she was enrolled in was designed to hone teaching skills; what you were assigned to teach wasn't supposed to matter very much. As it turned out, the only open section was decision science, which promotes reasoned decision-making by quantifying all the available data on causes and consequences, risks and rewards. A definite second choice, Goldie thought at the time.
But the frog turned into a prince. She found herself enthralled by the scope and edginess of decision science--"that you have to take concrete actions in the face of slippery information." On the plane ride back from Palo Alto, Goldie said to herself, This is it, my entire career has just changed. Within days, she was in Boston, talking her way into Prof-essor Milton Weinstein's summer clinical effectiveness program and paying for it with a Visa card. Weinstein remembers Goldie and Scott Gazelle, who is now a radio-logist at Massachusetts General Hospital, perched in the front row. "They would ask the best questions and would get 100s on the exams. At the end of the course, Sue told me, 'I'm hooked. I want to do this,' namely decision science." Goldie smiles and says, "I tortured him, poor Milt."
Red Backpack Years
Her first decision science paper came out in 1997, when she was one of the co-authors of a study about the cost-effectiveness of airbags by seating position. Goldie said she was encouraged to get involved in the project by Professor John Graham, now a high-ranking official in the Bush administration's Office of Management and Budget (OMB). They made an odd pair: "I said, 'John, you do traffic safety, and I do sexually transmitted diseases. You're a Republican, and I'm a complete liberal!'" The study, which was published in the Journal of the American Medical Association, garnered some press attention. Her young boys got confused. "They went to school and told the other kids that their mom invented the airbag," says Goldie, chuckling.
For most of us, models are toy airplanes or, perhaps, unnaturally slim young women wearing obscenely expensive clothes. For Goldie, models mean the disease models that lie at the heart of her research. With passion and an intellectual drive that keeps pace with the rest of her personality, she argues that disease models are the best, and perhaps the only, way to pull everything that is known about an illness--it's biology, prevention, and treatment--and render all of it into usable information on which policies can be based. No clinical trial or single longitudinal cohort study is able to consider all these components simultaneously and assess all possible strategies for all potential populations. "It is almost like you're creating a lab in a computer," she says, "and then you match that lab to the disease and a real population."
Construction of a model usually begins with a rough sketch of all the various stages of a disease--or, as modelers call them, "disease states." For example, in very simplified terms, a model for AIDS might include HIV infection, then several categories of AIDS based on CD4 cell counts and opportunistic infections, and finally death. For Goldie's students, it's literally a rough sketch because she encourages them to start with white poster board and magic markers. Of course, people often have or are at risk of having many diseases. It can take months to complete this first step of mapping out all the various disease states. Then modelers trawl through the published medical literature for any and all studies that might have addressed the "movement" of people from one disease state to another and sometimes back--or, as Goldie puts it, "the probability of progress in a given unit of time." All those disparate results must be melded and standardized so they can be plugged into the model. One of the final steps is to "calibrate" it against a real population to make sure it bears some resemblance to reality.
Goldie's lab analogy works. Once a model has been constructed then researchers can test different interventions--for example, alternative screening programs--and see which one will save more lives and prevent more suffering. If costs are added into the model, then it can churn out cost-effectiveness data: how much money was spent to save that life or prevent that suffering. A thorough disease model may have scores, even hundreds, of different disease states. The routes between them look like the imaginings of a mad traffic engineer. The poster board is kept for old time's sake, and the model exists as a computer code.
her medical training is a major help in conceiving disease models: "One
of the problems my health policy students have when they first conceptualize
a clinical situation is that it is very difficult for them to keep all
the biological details in their heads. They're like, Wait a minute, where
is that high-grade lesion, stratified in one of three ways, going? Whereas
for me it is second nature." On the other hand, one of the keys to
building a good disease model is keeping it simple. "There is a real
art to making sure that your model has enough detail to be realistic,
but you can't be modeling that a woman's grandmother died yesterday,"
she says. "A model has to be focused on core issues."
That's My Baby
In 1999 and 2000, Goldie published two influential papers on HPV-associated anal cancer in men and what would amount to a Pap smear for anal cancer for gay men. Before the AIDS epidemic, studies estimated that 35 out of every 100,000 gay men would get anal cancer every year. The incidence might even be higher among HIV-infected gay and bisexual men. Persistent HPV infection in the anus seems to play a role, because it leads to abnormal tissue growths that, in some cases, turn into anal cancer.
In women, the Pap smear involves a health care provider collecting a small sample of cervical tissue and having it examined under a microscope for any suspicious abnormalities. Similarly, a Pap smear for anal cancer would be checked for abnormal, possibly precancerous tissue, although, of course, it would come from the anus, not the cervix. In her studies, Goldie showed that the benefits--deaths and illness prevented--of anal Pap smears for gay men who are not infected with HIV are greater than the proven benefits of Pap smears for cervical cancer for women. The optimum strategy was to screen every two to three years starting at age 30. The case for screening HIV-positive gay and bisexual men was even stronger. Acceptance of any new screening test comes slowly, but Goldie said some health centers and health plans in the US have started to offer anal Pap smears.
A new screening test for anal cancer--it was a concept bound to get some press. But in the long run, Goldie's less conspicuous research on cervical cancer screening in poor countries will likely influence--and save--the most lives. Cervical cancer is the leading cause of death from cancer among women in the developing world. Women infected with HIV have an increased risk of HPV infection, so in places like sub-Saharan Africa, where the prevalence of AIDS is so high, the rate is likely to go up. Regular Pap smear screening has dramatically reduced cervical cancer mortality in countries where it is available, but in poor countries, where very few women are screened, it's problematic. There, women have to come back for multiple visits if there is an abnormal finding, and a health clinic can be a day's walk away. The infrastructure, equipment, and trained personnel to administer and interpret Pap smears are also relatively expensive for countries who spend just a few dollars per capita on health care each year.
Working with Wright and other collaborators, Goldie modeled three different screening strategies for women in South Africa. The simplest and cheapest is called direct visual inspection (DVI). Precancerous cervical tissue turns white when it's exposed to a vinegar solution. Nurses and mid-level health care workers are trained to apply the vinegar and identify any white tissue. Treatment consists of cryosurgery, which involves using a super-cold probe to freeze and kill abnormal cells. A second choice is testing women for HPV infection, which involves taking a tissue sample and testing it for the virus with a chemical probe that "finds" the DNA of HPV. The third choice is Pap smears.
Goldie's analysis showed that a single lifetime screen, using either a one-visit "screen-and-treat" strategy with DVI or a two-visit strategy with DVI or HPV testing, was much more effective than Pap smears, both in terms of cost and saving lives. But her study, published in the June 27, 2001, Journal of the American Medical Association, has kicked up a controversy, because direct visual inspection and HPV testing produce more false positives than Pap smears. Treating all screen-positive women would, therefore, result in some women getting treated unnecessarily. First, Goldie notes that the treatment is a minor, relatively uncomplicated procedure; obviously, you'd rather not treat false positives, but in this case the treatment risks are low. Second, the alternative is missing cancers altogether, because Pap smear screening has simply not been feasible in most poor countries. In fact, despite two decades of recommendations, less than 5 percent of women in developing countries are ever screened for cervical cancer. "Every preventive intervention has false positive and false negatives," she says. "The question is how do we quantify those and make the necessary tradeoffs. From a public health perspective, we need to balance the consequences of treating some women unnecessarily with the consequences of continuing to do nothing. Each year we wait, another 200,000 women die from invasive cervical cancer."
For all of the sophisticated computer work she does, Goldie is "old school" when it comes to teaching: the lights stay on as she uses an overhead projector and peppers her lectures with down- to-earth examples. "Something she says all the time is 'In plain English words, this means that '," says Kim. There are no prerequisites for enrollment. Students range from "completely geeky modelers" to future policymakers who are, uh-hum, a bit quantitatively challenged, smiles Goldie. She likes it that way: "Because in the real world, they have to speak to each other. And that has been one of the problems with this field: the methodologists have not been able to translate what they are trying to say into a language that can be used."
Goldie says her class is popular "because, presented the right way, these research methods are very compelling for these students because they are all driven by how do you take what we know and use it to form a bridge to doing something. Very rarely is there a subject like this, one that holds the hands of both research and action. I think that is what gets them fired up."
Meanwhile, her work on cervical cancer screening in poor countries is being batted around in the pages of the Journal of the National Cancer Institute. "Intense argument--that's awesome. At least we're getting them to talk about it," she says. A paper on abnormal Pap smears that she wrote with Kim is in galleys at the Journal of the American Medical Association. She's been asked to weigh in on the debate over changing cervical cancer screening guidelines in the United States.
A bit busy, you might say. But rather than being overwhelmed, Goldie's circuits seem to need all charge they can get. The demand for disease models is only going to grow, she believes, as health and other officials try to pick their way through the welter of disease, competing ways of dealing with it, and budgetary constraints. "This is the perfect time for someone who wants use the tools of decision science to make a difference," she says. "AIDS, tuberculosis, antimicrobial resistance, the use of genomics--all of these major global problems are going to need modeling."
Goldie sees herself as the ultimate go-between. "I want to be the person who helps develop a language between the policy world, the basic science world, the application-research world, and the public health practice world. And I want to train enough people to speak that language fluently. I think that can facilitate the implementation of so much that we already know and also develop new ways of solving the world's health problems. That is my mission."
Photos: top left, Kent Dayton; center, courtesy of Sue Goldie
page is maintained by Development Communications in the Office of Resource
To contact us with suggestions, comments, and questions, please e-mail: firstname.lastname@example.org
Copyright, 2005, President and Fellows of Harvard College