Statistics researcher examines the design and influence of meta-analyses
Everything is bad for you. Or maybe it’s good for you? The headlines on health news can be conflicting — as can the studies behind the headlines.
To help sort through the mass of information, scientists use meta-analyses: reviews that comprehensively examine studies on the same topic and use statistics to combine the results. The aim is to provide the best evidence to guide practice in fields such as ecology, education and medicine.
Meta-analyses are especially popular in health research, and have an influence on things like health care policy and which medications and supplements physicians prescribe.
Wiest conducts and studies meta-analyses and advocates for methods leading to stronger, more accurate results.
A good meta-analysis requires statistical expertise to make sure data are compared and interpreted accurately.
“There’s so much information now about different topics, so many research projects being published,” Wiest says. “The research runs the gamut from poorly designed studies to really tightly run mega-trials. How do you compare a mega-trial to a tiny little study?”
However, Wiest says, meta-analyses are often conducted by under-resourced individuals, including busy clinician-researchers and graduate students, who may be unable to carry out the analysis in a rigorous and reproducible way without assistance.
Wiest, her colleagues and students in the Department of Statistics are stepping in to provide this expertise. Through the Statistical Consulting Center, they help faculty, staff and graduate students design their studies, define their research questions, structure their databases and figure out the best way to interpret their results.
The Statistical Consulting Center provides statistical collaboration and support to approximately 150-200 researchers a year and is committed to supporting high quality and reproducible research at UI, Wiest says.
Wiest also provides expertise outside the university.
In spring 2016, she gave a presentation about meta-analyses to the global organization for the fish oil industry — a group that cares about meta-analyses because studies examining the effects of Omega-3 fatty acids influence their product sales.
Wiest’s goal for presentations like this is to help people understand how to interpret meta-analyses and understand their potential pitfalls.
“It comes down to being an informed consumer,” she says. “Just because they’re taking a bunch of studies and doing this synthesis doesn’t mean they have the best meta-analysis going on.”
This is important for people outside specialized industries, too. For example, Wiest cited the headline-grabbing October 2015 announcement by a World Health Organization-convened panel that links eating processed meats to an increased risk of colorectal cancer. The panel reviewed 800 studies on cancer in humans to help reach their conclusion.
But, Wiest says, people who read this news should know that while the increase was statistically notable, the type of cancer studied is already very rare.
“Even double the percentage is still a really small percentage,” she says. “It doesn’t mean everyone in the world has to stop eating bacon.”
It’s important to consider that meta-analyses are considered observations, not direct evidence, Wiest says.
“When you start doing a meta-analysis, you’re doing an observational study, which are known to be subject to things we can’t control,” she says. “You’re not controlling the populations of these studies, you’re not controlling who’s doing the studies. Publication bias is also a major issue meta-analysts need to contend with.”
Still, no study is ever expected to answer all the questions, Wiest says. Meta-analyses, when conducted and interpreted wisely, are useful tools for synthesizing information. They’re part of an open discussion about what research findings mean, and they allow scientists to continue asking questions and seeking answers.
“The role of meta-analyses is bringing in that quantitative piece and trying to quantify some of the subjectivity in order to make things clear, in order to make things repeatable,” Wiest says.