From mother-to-child transmission of HIV to coronary artery lesions, Dr. Elizabeth Brown loves a statistical challenge. And the more uncertain, the better—that’s where she believes good statistics can really make a difference.
As a research assistant professor of biostatistics at the University of Washington, Brown started working at the Hutchinson Center in 2002 on an HIV Prevention Trials Network trial with SCHARP, which examined mother-to-child transmission of HIV. She was appointed to the Center faculty in 2004. Now, as a researcher in the Vaccine and Infectious Disease Institute, she continues to address HIV transmission and has started a new venture as co-principal investigator of a data-coordinating center for an HIV and cardiovascular disease group.
Brown’s biostatistical work focuses on using probability to assess how a given marker, for example, CD4 cell counts, will predict one or more events over time, such as progression to AIDS or death.
As much of Brown’s studies and methods tie to HIV, especially in resource-poor countries, she is especially interested in applying her statistical knowledge to real world problems.
Challenges of HIV transmission
“One of my interests in HIV/AIDS and mother-to-child transmission is research on how to target treatments to people who don’t have access to a lot of treatments or prevention strategies,” Brown said. “There’s something fulfilling about contributing in that area, that even as a statistician you might be making a difference in peoples’ lives. But I also find the statistical problems really interesting. Mother-to-child transmission has a lot of unique issues, like not being able to determine exactly when transmission took place.”
If the mother is HIV positive and taking antiretroviral therapy or gets a single dose of the antiretroviral drug nevirapine during delivery, chances of transmission are low. However, there are many who can’t or don’t take these drugs, especially in low-income areas. In those cases, there’s a higher chance of transmission either in utero or through breast-feeding. But tests for HIV aren’t sensitive enough to detect the virus immediately after infection.
Mother-to-child transmission of HIV is a challenging problem because you can never get the full picture of what happens when, Brown said.
“When you add breast-feeding into the mix, if you test at birth and at six weeks, those infants who were negative at birth and positive at six weeks, you don’t know if infection happened through breast-feeding or in utero or during delivery,” she said.
Typical HIV tests for adults look for the presence of HIV antibodies in the blood plasma, however, those tests don’t work on young infants because they still have their mothers’ antibodies. So tests for infants look for HIV DNA or RNA instead, Brown said. These tests are currently done at birth and at six weeks.
Brown is using probability-based statistical models to tease out the “sensitivity profile” of transmission in utero and through breast-feeding. These models will accomplish three goals, Brown said. They will help elucidate the sensitivity of the HIV tests, determine how long after breast-feeding stops a second test should be administered, and shed some light on the mechanism of transmission through breast-feeding.
New research on HIV and cardiovascular disease
Brown came to UW and the Hutchinson Center after she completed her doctorate in biostatistics at Harvard University. She became interested in statistics while working in the Colorado Department of Public Health and the Environment after college, and later completed a master’s degree in biometrics at the University of Colorado.
She’s currently charting exciting and unknown territory as co-PI of the data-coordinating center for the HIV-CVD collaborative with UW biostatistician Dr. Richard Kronmal. This group was initiated last September to look at the relationship between HIV and cardiovascular disease through a variety of studies, as both HIV infection and antiretroviral therapy can increase the risk of heart disease. Brown is applying her statistical methods to imaging studies in this area. Some studies are using CT scans to look at coronary artery lesions, which are predictive of vulnerability to a heart attack. The scientists want to compare lesions between HIV-positive and HIV-negative patients, but this presents a statistical challenge, as there can be a wide range of variability between any two people in lesion number and characteristics.
Brown is also interested in new collaborations with other biostatisticians, especially in the area of multistate models—modeling for the risk of multiple events, such as HIV infection and death.
For more on Brown’s research, visit her faculty profile at http://myprofile.cos.com/eliz_brown.