Columnists Dispatches from experts on various facets of autism.
Opinion / Columnists / Connecting matters

Why we must be patient when searching for an autism biomarker

by  /  7 November 2014

Ivan Canu Connecting matters Helen Tager-Flusberg links autism science to society.Read more columns »

We have found differences in brain functioning in children with autism at an age that would be ideal for offering an early diagnosis1. So why do we not want to pitch our findings as evidence of a new biomarker?

Autism is a behavioral syndrome that emerges sometime in the second or third year of a child’s life, but the search is now on for ‘biomarkers’ that could serve as a diagnostic test for autism at a much earlier point in time, providing the opportunity for intervention when the greatest impact may be possible.

One approach for finding biomarkers is to study infants who have an older sibling with autism, and who begin life with increased familial risk (the estimated odds for these infants developing autism are one in five). The infants can be followed from shortly after birth until the time when a firm behavioral diagnosis can be made, usually by age 3. Several groups, including my lab in collaboration with Chuck Nelson’s lab at Harvard Medical School, are using this research design to explore measures of brain development and see whether these might yield neural biomarkers.

In a paper published in August in PLoS One, we found significant differences on a measure of functional brain connectivity in the first year of life between infants with high familial risk and those with no family history of autism1. The infants listened to speech sounds (such as ‘da’ and ‘ta’) while we recorded their brain activity using electroencephalography (EEG), a noninvasive measure of brain functioning.

Coherent results:

From the raw signal elicited soon after the onset of each sound, we computed the linear coherence (a measure of synchrony) in brain regions that are important for processing speech and language. Higher linear coherence is taken as a measure of greater connectivity between these language network regions.

Between 6 and 12 months of age, linear coherence increased in the infants at low risk, as we had expected. In contrast, in the high-risk infants, coherence decreased over this time period, and by 12 months was significantly lower than in the group of low-risk infants, suggesting lower regional connectivity. The infants who were later diagnosed with autism had the lowest coherence scores compared with all others.

Although our sample size is still relatively small, the results are exciting, adding to other studies that have found differences in brain functioning in young infants later diagnosed with autism.

Have we found a biomarker that could eventually be useful in diagnosing autism in infants? On the face of it, this question is obviously far too premature and simplistic, but this is the message that is frequently presented in press releases and picked up by the media. Often, when a study shows a difference between the brains of people who have autism and the brains of those who do not, the press tells the public that we are on the cusp of having a new, and presumably better, approach to diagnosing autism early.

These stories leave out all the complexities of the science. Our study is small, so the most important next step would be to look for replication in an expanded sample. We replicated our finding using another paradigm and technology in the same group of infants2. But independent replication is the key to evaluating whether a finding will hold up over time.

Is what we found unique to infants at risk for autism, an important criterion if this biomarker is to be used in diagnosis? If not, then we may have identified a non-specific biomarker that extends to other neurodevelopmental disorders, such as language disorders or dyslexia, which might also produce early differences in the neural circuitry for language.

Fuzzy boundaries:

In our study, we did not find sharp boundaries: The distributions of individual scores for infants in each group overlap, so as a biomarker it will not be sensitive enough to be used for diagnosis on its own.

At best, then, this measure signals risk rather than autism itself. The finding also raises the question of whether this ‘biomarker’ would be found in infants with autism who don’t have an older sibling with the disorder. So, for now, we have only found a measure that potentially can predict autism for infants with familial risk.

Clearly, there are several important follow-up studies that are needed before we can reach any conclusions. And this applies to all studies that identify neural or cognitive differences between infant siblings and controls.

We need to remind ourselves that this line of work has just begun; scientific progress involves successive studies that address the questions raised by findings from earlier ones. The challenge, however, is that prospective developmental research designs are expensive and take a minimum of about five years for a reasonable size sample. But it doesn’t seem than anyone — scientists, funders, the media, the public — has enough patience.

The evidence emerging from many neurodevelopmental, cognitive and behavioral studies of infant siblings points in one direction: Autism is a disorder that emerges over time. In our study, for example, it is the difference in the developmental trajectory between 6 and 12 months that is most significant, not the measure taken at any single time point. All this suggests that studies of older children or adults with autism are not going to yield diagnostic biomarkers, though they may provide biomarkers that could be useful for other purposes, such as sensitive measures of response to treatment.

It is time for us to reevaluate how we communicate the findings from our research on biomarkers. We must stop saying that every study showing differences between people with autism and controls has found a new approach to diagnosis. There are dangers in repeatedly offering a false promise of having found the kind of objective, biological test for autism that the public is hungering for; instead we should take the opportunity to present our work in a realistic light that conveys not only what we have found in each study, but also what still needs to be done.

Helen Tager-Flusberg is professor of psychological and brain sciences at Boston University, where she directs the Center for Autism Research Excellence.


1. Righi G. et al. PLoS One 9, e105176 (2014) PubMed

2. Keehn B. et al. Front. Hum. Neurosci. 7, 444 (2013) PubMed

  • Seth Bittker

    I appreciate that we are trying to move beyond behavior alone to define autism. This is a step in the right direction.

    However, it appears to me that much too much funding and research attention is now going toward this search for new so called “biomarkers” which to date appear to be just neurological markers dressed up with a more sophisticated name. Let’s say you find some neurologocal marker such as eye-tracking, reaction time, EEG measure, or some other neurological marker which tends to pick up at risk children. What do you do? I would imagine you would setup huge testing programs and will put all at risk children in intensive ABA and hope for the best.

    The problem is that in the vast majority of cases autism is a biochemical disease. So in many cases even if you give hours of ABA a day you will still have a child with significant neurological as well as health issues because the child’s biochemistry is unbalanced. In addition as these neurological markers give little information about underlying biochemistry from a true biochemical treatment perspective they are not very useful.

    There are already a number of existing biomarkers (real ones: not just neurological markers) available in the literature that could be used to pick up a at risk kids but they don’t seem to be used. For example, many with autism have very high levels of sulfate loss in urine and low cysteine levels in blood. So a high sulfate to cysteine ratio could serve as an excellent marker for at risk children. See: In addition often those with autism have high ratios of oxidized to reduced glutathione. So this too could serve as an excellent marker for at risk children. See: Frequently in autism one also sees high levels of vascular damage through lipid peroxidation markers such as: 2,3-dinor-thromboxane B(2). See So this too would make a good marker for at risk children.

    I wish that some of the money and effort that is now being funneled into a search for new neurological markers for autism could instead be put toward projects that experiment with treatments for those with biochemical markers that have already been documented in the literature as being significant in autism. For example, take a population with autism or a group at risk for it. Check markers for lipid peroxidation. Divide those with the marker into two groups. Give one group herb X which has been shown to decrease lipid peroxidation; give the other group a placebo. What is the effect on lipid peroxidation measures in the two groups after 4 weeks of treatment? What is the effect on behavioral measures?

    It seems to me if such an approach were taken we could get much more useful information about the underlying causes and treatments for autism. If somebody reading this makes funding decisions, I would be grateful if you would consider this.

    • Rene Anand

      Seth Bittker,

      I concur 100% with your thinking and suggestions. I am sorry, but I don’t have any say in funding decisions.

      I also think its futile to think heritability is all in the genes, when we know genes are only one part of the equation.

      Investing insane amounts of funds in probability theory algorithms including Bayesian analysis to find genes using “big” and “bigger” data analysis of gene expression data alone will reach an asymtotic limit and has led to the “missing” heritability paradox.

      And no, I DONT mean epigenetics either! I am afraid, we are looking for the lost key under the light, to quote an old saying. You are suggesting an empirical approach that I think will yield great knowledge and better understanding if done well.

      The reason we study genes is because its easier and machine accessible. However, biochemical pathways take a much higher skill level and this skill has been neglected by funding agencies chasing genomics alone.

      My best wishes to you, stay the course!

      Rene Anand

      • Seth Bittker

        Thank you so much for your reply. I agree with you that there is a misallocation of resources toward research focused on genetics. From our off line discussions I have learned that your own research effort is truly innovative.

        Here are some worthy research projects that I hope some ambitious researchers pursue:
        1) Placebo controlled trial of oral MB12. Some literature suggests MB12 injections are effective for those with autism. See: Other literature suggests that MB12 is well absorbed orally. See: As oral MB12 is available OTC, it would be useful to know whether anecdotal reports of its efficacy in autism stand up to rigorous research. The trial could be structured with an initial blood test of oxidized and reduced glutathione levels and methylmalonic acid levels. Only for the subset of patients where oxidized to reduced glutathione is high or alternatively where methylmalonic acid is high would the patients proceed to be randomized to receive the treatment or placebo. Measure same metabolites after the trial. Also do behavioral assessments before and after the trial.

        2) Trial of oral methylfolate vs oral folinic acid vs placebo. There is literature suggesting that folinic acid can help some of those with autism. See: In addition much of the same literature points to methylation deficits. Therefore it would seem that a methylated form of folate could be a good potential therapy. This trial could also be conducted based on an initial blood test and only for patients where oxidized to reduced glutathione was high would it proceed to the next stage. Measure metabolites before and after the trial to test biochemical effectiveness. Also do behavioral assessments before and after the trial.

        3) Placebo controlled trial of Epsom salt baths. There is a lot of literature supporting the notion that sulfur deficits are common in autism. See Epsom salts are magnesium sulfate and anecdotal reports suggest they are helpful in autism. A placebo controlled trial could be run where the placebo is sodium chloride salt. In other words, those who receive the treatment would be bathing in Epsom salts and those who receive the placebo would be bathing in sodium chloride salt. The trial could be structured to measure cysteine in plasma and sulfate excretion before and after to see if there are differences between the two groups. Behavior should also be assessed.

        4) Placebo controlled trial of Methylsulfonylmethane (MSM). As mentioned previously there is a lot of literature supporting the notion that sulfur deficits are common in autism. MSM is a common sulfur supplement that has been shown to have promise in other inflammatory conditions. See: Would it help in autism? A trial would be useful. Look at cysteine and sulfate excretion before and after the trial as well as behavior.

        5) Placebo controlled trial of Melatonin for cytokine imbalance. Trials of melatonin for sleep in autism have already been conducted with positive results. See: Melatonin also has been shown to help regulate cytokines in some conditions featuring cytokine dysfunction. See: Autism often features cytokine dysregulation. See: It seems that melatonin might help some of the population with autism due to its affect on cytokines even those who do not have significant issues with sleep. Measure cytokines before the trial and measure cytokines after the trial. Also look at affects on behavior.

        6) Placebo controlled trial of Silymarin. Autism often feature cytokine abnormalities: Silymarin modulates cytokines. See: Relatedly vascular damage is often present in autism. See: It would be interesting to know if silymarin would help normalize cytokine or vascular dysfunction in autism. Measures of both could be taken before and after a trial. Measures of behavior should be looked at as well.

        7) Detailed parent survey on conditions, feeding practices, diet, and especially supplement use before and first 3 years after birth: 2 groups: parents of kids with autism and parents of controls. Some have hypothesized that vitamins provided to infants and toddlers may increase the risk of autism: Others have suggested that vitamins are a protective factor. See: It would be useful to get some information on whether common practices such as providing vitamin A and D drops to infants are decreasing risk of autism, increasing risk of autism, or have no effect on risk. Similar contradictory findings apply to breastfeeding and a well designed parent survey could answer whether it increases risk, decreases risk, or has not effect.

        8) Placebo controlled trial of B. Fragilis probiotic. Evidently this probiotic reduces symptoms of autism in mice. See: Are we ready for human trials yet? If not, I guess we can wait on this one until we are. If we are, this should be done soon.

        9) Is it a microbe? There are indications in the literature that sometimes autism or autism like symptoms are induced by microbes. For example Lyme disease ( or streptococcus ( have induced autism. It would seem an interesting study could be conducted by taking a number of children with regressive autism and obtaining levels of metabolites that are commonly associated with infection and comparing the levels with controls. For example, one could look at markers such as CRP, prolactin, ferritin, and cytokines, and various titers in those with regressive autism, lyme, pandas, and age matched controls. Perhaps the group with autism could be further partitioned into those whose markers appear to suggest some kind of infection and those whose markers do not and then some analysis could be done on those in the group that have markers that suggest infection.

        • ASD Dad

          I would add – Rapamycin

          In the online journal PLoS ONE, Frances Jensen, MD, in the Department of Neurology and the F.M. Kirby Neurobiology Center at Boston Children’s, and lab members Delia Talos, PhD, Hongyu Sun, MD, PhD, and Xiangping Zhou, MD, PhD, showed in a rat model that early-life seizures not only lead to epilepsy later in life, but also produce autistic-like behaviors.

          Drilling deeper, they showed that early seizures hyper-activate a group of signaling molecules collectively known as the mTOR pathway. This increased signaling – above and beyond the normal surge that happens early in life – disrupts the normal balance of connections (synapses) in the rats’ developing brains. The rats go on to develop epilepsy and altered social behavior, and Jensen believes something parallel happens in humans.

          But here’s what’s exciting. They did other experiments where they gave the rats the drug rapamycin, which disables the mTOR pathway, before and after seizures. The mice did not show abnormal synapse or circuit development, and were less likely to have seizures later in life. Autistic-like symptoms appeared less often.

          “Our findings show one of probably many pathways that are involved in the overlap between epilepsy and autism,” says Jensen. “Importantly, it’s one that is already a therapeutic target and one where treatment can reverse the later outcome.”

  • ASD Dad

    I must admit the thought of children undergoing intensive ABA because fundamentally they are sick could be something that we reflect on in 10 – 20 years time. I advocate for a multi-disciplned understanding of autism.

    Am I asking to much ?

  • Autism Services

    Autism spectrum disorders (ASD) encompass a range of neurodevelopmental conditions that are clinically and etiologically very heterogeneous. ASD is currently diagnosed entirely on behavioral criteria, but intensive research efforts are focused on identifying biological markers for disease risk and early diagnosis. Here, we discuss recent progress toward identifying biological markers for ASD and highlight specific challenges as well as ethical aspects of translating ASD biomarker research into the clinic.


Log in to your Spectrum Wiki account

Email Address:



Request your Spectrum Wiki account

Spectrum Wiki is a community of researchers affiliated with an academic or research institutions. To be considered for participation, please fill out this form and a member of our team will respond to your request.


Email Address:

Title and Lab:

Area of Expertise: