Does Filtered-Music Therapy Work? The Evidence Is Too Thin

What the review found

Auditory integration training is ten hours of listening. The child wears headphones for two half-hour sessions a day across ten consecutive days and hears music that a device has altered, damping the frequencies the child is judged to be hypersensitive to and randomly modulating high and low sounds as the recording plays. Guy Berard developed the method in France in 1982, on the premise that unusual sensitivity to particular frequencies underlies a range of behavioural and learning difficulties and that hearing can be re-educated. A 1991 book describing one child’s recovery made it popular, and it is still sold to families worldwide.

Whether it works was the question that Sinha, Silove, Wheeler and Williams, at The Children’s Hospital at Westmead in New South Wales, took to eight databases with no restriction on language. Their systematic review appeared in Archives of Disease in Childhood in 2006. Six randomised controlled trials met the criteria, with 171 participants in total, aged between 3 and 39. Five compared the training against unmodified music through the same headphones; the sixth played unmodified music into the room.

Across those six trials the reviewers counted 17 different outcome measures, only two of which were used by as many as three studies, and twelve of which appeared in a single study each. Five trials used the Aberrant Behaviour Checklist, a rating scale of problem behaviour completed by adults who know the child. All five reported a total score. The author of the checklist has stated that a total score is incorrect and inconsistent with the instrument’s design, which calls for five separate subscale scores to be read on their own.

The intended meta-analysis was abandoned. Only one trial had published data in a form that could be pooled at all, and when the reviewers attempted to combine subscale scores from two trials at three months, the disagreement between them was severe, with an I² statistic ranging from 55 percent for hyperactivity to 88 percent for stereotypic behaviour. That statistic estimates how much of the variation between studies exceeds what sampling error alone would produce.

Three of the six trials found no benefit over the control condition. Three reported improvement at three months on the total checklist score of questionable validity, and one of those also reported improvement on the subscales the instrument was built for. The two largest trials, of 80 and 30 participants, found no difference between training and control, and the trial of 80 was the only one in which anyone had worked out in advance how many children would be needed to detect an effect. All six described themselves as randomised, five gave no account of how, and after the reviewers wrote to the authors, allocation concealment was judged inadequate in every one.

What this means for you and your child

Three trials finding nothing, both of the largest among them, is suggestive and stops short of proof. A trial can miss a real effect because there is nothing there, or because it enrolled sixteen children and assessed them on an instrument no other trial used. Six studies of this quality are unable to show that ten hours of filtered music does nothing at all, and the reviewers’ conclusion was correspondingly narrow: at present there is not sufficient evidence to support its use.

Nothing in the review suggests the sessions themselves are harmful. Two trials set out to record adverse effects and found minor physical complaints reported by parents in both the treatment and the control groups, with no significant difference between them, while three trials did not record adverse events at all. A 1998 analysis concluded that lower rather than maximal equipment settings put no listener at risk of hearing loss, and four trials described keeping their machines within the manufacturer’s recommendations. No study reported a child deteriorating on a standardised measure.

Unusual sensitivity to sound is common in autism and it is distressing, so an intervention aimed straight at it is a reasonable thing for a parent to want. The trials record no harm from the sessions themselves, and no evidence that any calm a child finds in them belongs to the filtering.

The trial capable of testing the claim has never arrived. The American Academy of Pediatrics recommended in 1998 that the therapy be confined to research protocols, citing concerns about its validity and its theoretical basis, and the evidence has not moved since. The reviewers observed that it continues to be practised worldwide at what may be considerable cost to the family.

A course of auditory integration training buys ten hours of altered music, a device, and a practitioner’s time. It does not yet buy the change in behaviour or in sound tolerance that the sales material describes, because no trial has been run well enough to establish that the change occurs. If you have already paid for one, you bought a plausible theory and ten hours of your child’s time, and nothing in the brochure was ever going to tell you that six trials had been run and not one of them could settle the question.