Stay Ahead of the Curve
Engineering Researcher Draws a Broader Picture from Data in AIDS Studies, Drug Trials
To help make sense of the raw data they collected, scientists in each of these studies called in Colin Wu, a biostatistics expert from The Johns Hopkins University's Whiting School of Engineering. Wu, an associate professor of mathematical sciences, uses a relatively new form of data analysis called curve estimation to produce clearer pictures of how health changes over time.
Standard statistical methods provide just a snapshot of health conditions at a single point in time. In the growth study, for example, standard methods could have told researchers the average weight, exactly two years after birth, of children born to HIV-positive, vitamin A-deficient mothers. But Wu's technique yielded a curve that spanned the length of a project. It let researchers compare average growth figures for the children two months after birth, 23 months after or anywhere in between.
The curve produced by Wu and his partners is a very useful research tool, he says. "The shape of growth over time cannot be told by conventional methods," he explains. "Children's growth over time may not look like a straight line. Our method will show what it does look like."
Wu and colleagues Donald R. Hoover, John A. Rice and Li-Ping Yang reported on the mathematical foundation for this study in the December 1998 issue of Biometrika. Working with data compiled in a Johns Hopkins School of Public Health study in Central Africa, the researchers looked at weight figures collected from 328 infants born to HIV-positive mothers. The mothers had been tested for vitamin A levels during their third trimesters. The biostatisticians developed a growth curve showing that children whose mothers were deficient in vitamin A during pregnancy grew significantly slower than their counterparts, regardless of gender, throughout their first year of life. Not surprisingly, the analysis also showed that children who contracted HIV from their mothers at birth grew more slowly than those who did not.
Wu has successfully used curve estimation in other projects. In one clinical study funded by the National Institute on Drug Abuse, Wu and his former Ph.D. student, Chin-Tsang Chiang, were able to track the day-to-day comfort levels of two groups receiving in-patient treatment for opiate addiction. One group received a medication that researchers hoped would reduce painful withdrawal symptoms. The other group received a placebo. Again, traditional statistical methods could not do the job. "How do you know if one group is doing better on a day-to-day basis?" Wu asks. "There may be no difference between the two groups for the first day or two. Then the patients in one group may feel better. You have to be able to make a valid comparison."
Using curve estimation, Wu produced a picture of how the addicts felt throughout the treatment period. It showed that the medication did indeed ease the pain of withdrawal.
The technique has also proved helpful in cohort studies, which involve long-term tracking of health conditions in a particular group of people. Wu used data from the Multicenter AIDS Cohort Study to determine how the level of CD4 cells -- a critical measure of the strength of the body's immune system -- declined among 400 gay men who were infected by HIV between 1984 and 1991.
Wu was able produce charts that will enable physicians to predict how quickly these cells will decline. This information will help doctors and patients plan an appropriate course of treatment. "If you want to say where, on average, a person's CD4 cell count will be perhaps four years after HIV infection, this curve will tell you," he says.
A native of China, Wu came to the United States in 1982 to study at UCLA, where he earned a bachelor's degree in mathematics. In 1990, he received a doctorate in statistics at the University of California, Berkeley. He has been a faculty member in the School of Engineering at Johns Hopkins since 1991.
Wu's long-term goal is to develop computer software that will allow public health researchers to insert their data and generate their own curve estimations without training in advanced mathematics.
Color slide of Colin Wu available; Contact Phil Sneiderman
Go to Headlines@HopkinsHome Page