Alarm Over Autism Test
A hypothesis that maternal antibodies can impair fetal brains stirred plenty of doubts-— and that was before the researchers set out to turn it into a test for predicting autism – Science
In July, immunologist Judy Van de Water and her team at UC Davis, which includes Bauman and Daniel Braunschweig, bolstered the hypothesis that maternal antibodies cause some autism with two studies, including one showing autismlike symptoms in monkeys injected with such antibodies. And women may soon be able to check whether they have the suspect antibodies: California company Pediatric Bioscience announced that it is moving forward with a new diagnostic test, based on patented antibody screening techniques licensed from Van de Water and UC Davis.
Van de Water and colleagues propose that antibodies in a mother’s bloodstream travel through the placenta and cross the fetal blood brain barrier, interfering with proteins key to multiple steps of brain development, including maturation of neurons.
From the press release [doc]:
The study was published online today in the Nature journal, Translational Psychiatry, and was led by immunologist Judy Van de Water, Ph.D., who is a faculty member in the Department of Internal Medicine, and the UC Davis MIND Institute, a leading center for research and clinical services on autism and other neurodevelopmental disorders. Dr. Van de Water is also the Chief Scientific Advisor of PBI.
“Based on these findings, PBI is developing a simple, quantitative blood test for MAR autism, which would be made available through physicians to the mothers of young children between the ages of 12 – 24 months who may be showing some signs of developmental delay,” said Jan D’Alvise, Chief Executive officer of PBI. “If the test is positive, the child would be an immediate candidate for early behavioral intervention, which studies show can dramatically improve outcomes and quality of life for the child.”
The study was the largest to date and analyzed blood samples from several hundred mothers of children with autism and control mothers of children without autism to examine the reactivity of their blood sample with the candidate proteins. Seven proteins were found to be significantly more reactive to the blood of mothers of children with autism than that of the control mothers. Nearly 23 percent of mothers of children with autism had one of the specific combinations of autoantibodies against the target proteins, compared with less than 1 percent of mothers of children without the disorder.
What’s the problem?
There are concerns that the conclusion is premature and the test is overblown. Why is this test being commercialized?
“This is very, very premature—this research has come out of one group, and basically one study. I’m amazed that they’re going ahead at this point and trying to commercialize a test,” says autism researcher George Anderson of Yale University. He and others say that Van de Water’s data are too preliminary, and her statistics too weak, to support such clinical uses. They are skeptical of the mechanism she has proposed for how maternal antibodies could damage the fetal brain. And before any antibody test for autism is launched, they say, her results need to be extensively replicated.
Van de Water has three large prospective trials under way to validate her results but is still pushing ahead with marketing of the test. (Shades of Andrew Wakefield are cast. Conflict of interest?) She has people asking for it already. There is a concern that some already pregnant women will get a positive result and abort their babies based on these results. Van de Water says the company plans to limit use to those not currently pregnant and testing goes in hand with counseling.
If she is right, this is huge and will help lower the cases of autism and provide early intervention for those babies whose mothers are diagnosed. But there are so many potential confounding factors with this.
One researcher notes autism is diagnosed in the general population in about one in 88 births. Even a few false positives will make the test of limited value as a diagnostic too. His calculations show that the “positive predictive value” is only 16.5%, not good, meaning only one in six of the positive tests would be correct. That’s a not ready for prime time test.