From Our 2010 Archives
New Method for Predicting IVF Success
Researchers Look at 52 Factors to Estimate Chances of Success From in Vitro Fertlization
Reviewed By Laura J. Martin, MD
July 19, 2010 -- A new method for predicting IVF success that takes 52 variables into account works better overall than the commonly used age model, according to a new report.
In IVF (in vitro fertilization), eggs and sperm are brought together in a lab dish to fertilize an egg.
"Our model is more than 1,000 times more predictive than the age-based model," researcher Mylene Yao, MD, an assistant professor of obstetrics and gynecology at Stanford University School of Medicine, tells WebMD.
"We are pinpointing patients more specifically, using [more than] 50 variables rather than one," she says.
The model estimates the likelihood of a live birth with future IVF cycles for women who have already gone through one cycle.
Eventually, Yao hopes the method will be available for commercial use. The report is published online in the Proceedings of the National Academy of Sciences.
Predicting IVF Success: The Model
Nearly 75% of IVF treatments don't produce a live birth, Yao writes. When deciding whether to try another IVF cycle, patients are guided mostly by age considerations, with estimates based on that.
Yao and her colleagues wanted to develop a more personalized way to estimate future success. So they evaluated data from 1,676 first IVF cycles done at Stanford Hospital & Clinics between 2003 and 2006. They found 52 factors, including age but also hormone levels, quality of eggs, and embryo characteristics, that had an effect on the chances of having a live baby.
"It's basic information," Yao tells WebMD. "We stuck to information that is freely available in people's medical records."
Next, they put together a computer model that classified patients into subgroups based on their clinical characteristics, a method called "deep phenotyping."
They validated the model by testing it on 634 first IVF cycles and 230 second IVF cycles done at the facility from 2007-2008.
The findings? The new model wasn't perfect, but was often more accurate than using age alone, Yao says. "For every patient for whom the age test was more predictive, there were more than 1,000 for whom our test was more predictive."
Yao and another co-author have founded a company, Univfy Inc., a start-up that will focus on refining the model and bringing it to market. Stanford holds the patent on the test.
Predicting IVF Success: Second Opinion
The new model is termed an improvement by Andrew R. La Barbera, PhD, scientific director of the American Society for Reproductive Medicine in Birmingham, Ala., who reviewed the report for WebMD.
"This model certainly improves the ability to inform patients of the likelihood of them conceiving after IVF," he says. Further research is needed, he says, to make the predictions more reliable.
And he has this caveat about the new model: "It does not provide an accurate prediction in all cases. This model correctly predicts outcome more frequently than the test based on age."
The new model, La Barbera says, takes variables and data that experts often discuss as playing a role in IVF success and plugs them into the prediction.
IVF in the U.S.
About 1% of newborns born in the U.S. annually are IVF babies, according to Yao.
If fertility problems exist, she writes, IVF treatment "offers the highest live birth rate per treatment cycle."
Even so, she says, the decision to pursue more treatment after the initial attempt fails is a difficult one due to high costs and an "uncertain prognosis."
Her model, she says, aims to provide a tool that can provide both an evidence-based and personalized prediction of the chances of a live birth from an IVF cycle.
SOURCES: Banerjee, P. Proceedings of the National Academy of Sciences Early Edition, published online July 19, 2010.