University of Chicago Department of Pediatrics
Rebecca Lipton spends her time investigating a disease for which there is no cause--no single cause anyway. Since 1991 she's worked on the Chicago Childhood Diabetes Registry, studying factors linked to the disease in urban populations. After asthma, diabetes is the most common childhood disease, but black and Latino children, among whom there's a high incidence, have been underrepresented in studies.
Harold Henderson: You used to be a nurse. What drew you to epidemiology?
Rebecca Lipton: As a kid, I read lots of books about archaeology. I really wanted to be an archaeologist. Epidemiology is a lot like archaeology. You're looking at what you have now, and you want to figure out what happened in the past to lead up to it. The important diseases now are chronic--heart disease, cancer, diabetes. It's very difficult to single out just one cause for them, so there's a lot of work to be done.
HH: You can't just knock out one bacterium and it's gone.
RL: You really have to do the detective work. One person smokes all their life and dies of old age at 100; someone else smokes five years and they're done. There's no doubt smoking causes lung cancer, but other risk factors are involved too. Chronic diseases usually come from processes and exposures that start years before the disease ever shows up. Figuring those out is what makes it fun. AIDS isn't like that--it's pretty much syphilis with a different germ. There's nothing new to learn there.
HH: Isn't diabetes really two diseases?
RL: Type 1 diabetes (what used to be called juvenile-onset diabetes) is basically an autoimmune disease in which your immune system attacks your own insulin-producing cells. Type 2 diabetes happens when your body can't use the insulin it produces correctly, often because you've been overweight for years. So Type 1 is more common in children and teens, while type 2 is most likely to affect middle-aged people.
HH: Type 1 is like being allergic to yourself?
RL: Right. And it has a genetic component. There are certain genes that code for the proteins that sit on the surface of white blood cells and recognize what in the bloodstream is foreign and what belongs to your own body. People with certain genes are at much higher risk of having proteins that can't always tell the difference, leading to Type 1 diabetes. But not all of them actually get it.
The Chicago Childhood Diabetes Registry is a typical population-based epidemiology study where you go out and count the people that have the disease. I had a professor who didn't have a real high opinion of his colleagues, and he used to call epidemiologists "body counters." That's what we do--we try to count everyone who was a Chicago resident and under 18 when they were diagnosed with diabetes. We get funding from the National Institutes of Health, mainly because Chicago has a lot of black and Latino children, and there isn't much information about diabetes among them.
HH: Why is it important to have a large number?
RL: Because you have to have a big enough denominator. Epidemiologists are all the time talking about fractions, numerators and denominators. In this case the numerator is the number of children with diabetes in Chicago; the denominator is the total number of children in Chicago. That fraction is the rate of diabetes occurrence. Of course we can do the same thing for subgroups, say black females, and we can compare the rates of different subgroups and get rate ratios between them.
If you think about it, the fraction of black females under 18 with diabetes is also a good estimate of the risk--it's an incidence rate. This is an incidence study. And because kids mostly don't get sick, you need a big population in order to get the statistical power to learn anything.
HH: What are the numerators and denominators here, overall?
RL: Every year we're seeing about 17 new cases of diabetes for every 100,000 African-American kids and about 12 per 100,000 Latino kids. In total numbers in the city of Chicago we see about 50 new cases a year among African-Americans, about 25 among Latinos, and about 25 among whites and other groups.
HH: I see that you published one paper where you were able to break down the incidence of childhood diabetes in Chicago according to what month the children were born in, and it was somewhat better to be born in October.
RL: That's not as weird as it sounds. We figure that has to do with seasonal exposure either in the womb or shortly after birth--maybe something in the food or a virus.
HH: When we read about election polls, we generally have some idea of how to judge them--a poll should be based on a random sample and tell its margin of error, for instance. How should we read epidemiological studies in the media?
RL: If you see the same association turning up in different studies, with different populations, in different places, then it's more likely to be true. If month of birth has a statistically significant effect on childhood diabetes in Chicago but doesn't in Detroit, there might not be anything in it.
About 25 years ago a brilliant epidemiologist named Bradford Hill put a list together of ways to tell when an apparent association is likely to be real: Is there a large association, three or five or ten times the risk of another group? Is there a dose-response relationship, so that more of whatever you're investigating as a possible cause leads to more of the disease? Is it biologically plausible? You have to look at all these things together.
HH: You mean I can't just write that you've found the cause of Type 1 diabetes?
RL: If you talk about causation to epidemiologists, steam will come out of our ears. As I was saying, it's almost always too simple to talk about the cause of a chronic disease. Of course, it would put our registry out of business if they did find a single cause of childhood diabetes. I'd have to go back to emptying bedpans.
Art accompanying story in printed newspaper (not available in this archive): photo/Lloyd DeGrane.