Two Autism Studies Fuel Hope -- and Skepticism

Two autism studies fuel hope -- and skepticism

Is this treatment for autism too good to be true?

By Hugh McIntosh

Two surprising research reports from behavioral psychologist Ivar Lovaas and colleagues have sparked hope and controversy in the autism community over the past 12 years--and a third one may again roil debate.

The initial study found that nine of the 19 autistic children who went through Lovaas's intensive Young Autism Project reached "normal functioning"--measuring in the normal range of IQ and being able to complete the first grade unassisted.

A follow-up study six years later reported that these children were still functioning normally. And preliminary findings from another follow-up, with completion expected this summer, indicate that these nine--now 20 to 30 years old--have maintained their developmental gains.

"Eight of them look indistinguishable from typical individuals, other adults, on the tests," says Lovaas from his office at the University of California, Los Angeles. The ninth one "definitely has some personality problems...but he wouldn't be diagnosed as autistic anymore."

Lovaas's dramatic results have raised high hopes in parents of children with autism, the often devastating developmental disorder of the brain that causes communication difficulties, social interaction problems and unusual or repetitive behaviors and interests. Most children with autism--from 61 to 74 percent--are unable to function independently when they reach adulthood, and only about 5 percent lead a fairly normal life. Lovaas, however, reported remarkable success treating autistic children using teams of trainers working one-on-one, 40 hours a week, for two to three years to teach them thousands of individual behaviors one at a time--a technique known as discrete trial training.

Parents now frequently ask school districts to provide intensive discrete trial training for their autistic children. When districts balk, parents often take them to court and usually win. Such treatment costs about $60,000 a year per child. Lifelong protective care of a person with autism, however, can cost $2 million, Lovaas reports, so investment in early, intensive treatment could be a bargain.

But the Lovaas study has raised caution flags for some in the scientific community. While calling the results promising, critics question the study's methodology. In particular, they suggest that the children Lovaas selected for his treatment may have already been high-functioning autistic, in which case the findings would not be applicable to the general population of children with autism.

"There is a need for healthy skepticism regarding claims as dramatic as those of Lovaas," conclude Frank Gresham, PhD, and Donald MacMillan, EdD, of the University of California-Riverside, in reporting their investigation of the Lovaas outcomes (Journal of Autism and Developmental Disorders, Vol. 28, No. 1, p. 5-13).

Strongest data to date

Gresham, a psychologist, says the Lovaas study tended to select autistic children with mild mental retardation--IQ's around 63 to 65--as opposed to those with severe or profound mental retardation. The children were also high-functioning in echolalia, the ability to imitate language, which is a good predictor of positive treatment outcome, he says.

"The kids he selected were kids that really had the best probability of doing well to begin with, probably regardless of what you did with them," he says.

Lovaas's research, however, is methodologically the strongest to date of any research on the efficacy of behavioral treatments for autism, says developmental psychologist Sally J. Rogers, PhD, of the University of Colorado in Denver.

In an article published in the Journal of Clinical Child Psychology (Vol. 27, No. 2, p. 168-179), she evaluated the methodology of eight outcome studies published over the previous 10 years, including the one by Lovaas, and found that none met rigorous methodological criteria. Among the studies' shortcomings were a failure to randomly assign subjects to intervention and control groups, lack of independent replications, and failure (though not in the Lovaas study) to have raters of outcomes blind to whether a child got an intervention or not.

"The science is soft right now," Rogers says. "There are so few studies published with a decent design that it's just hard to draw conclusions."

The Lovaas study, which has not yet been replicated, relied on a quasi-experimental design that did not use strict randomization in assigning subjects to treatment and control groups. Nevertheless, Rogers says, "even laying aside the issue of random assignment, there isn't another rigorous study at the level he did of control groups, regardless of what [outcomes] they're getting."

Lovaas says the subjects in his study were, indeed, a representative sample of autistic children. They were independently diagnosed with autism by licensed clinicians and assessed on 20 pre-treatment variables considered descriptive of autism or related to outcome. Their mean IQ was 60 at intake and, according to his 1987 report in the Journal of Consulting and Clinical Psychology (Vol. 55, No. 1, p. 3-9), their IQ scores indicate normal intellectual functioning in two of the 19, moderate retardation in seven, and severe retardation in 10.

Fifteen percent of the children referred to the study were rejected because of profound retardation, which was one of the criteria for exclusion from the Lovaas study. Children with profound mental retardation were not included because of the difficulty in diagnosing autism at that level. As for vocal ability, one child had minimally appropriate speech, seven were echolalic, and 11 were mute.

Strict randomization of the children was impossible because parents objected, Lovaas says. Instead, with approval from the National Institute of Mental Health (NIMH), which funded the study, a child was assigned to the experimental group if therapists were available. If not, the child entered a control group. Two children were assigned to a control group because they lived more than an hour's drive from UCLA, a situation that prevented sufficient staffing from treating those children.

This "staff availability" assignment procedure ensured that experimental and control groups were similar, Lovaas says. In fact, they showed no differences on 19 of the 20 pre-treatment variables, including echolalic speech (nine in control and seven in experimental). Echolalic speech did not predict normal functioning. The variable on which they differed, chronological age at treatment onset, was not related to outcome, Lovaas says. The poor outcome in the similarity constituted control groups one and two would seem to eliminate spontaneous recovery as a contributing factor to the favorable outcome in the experimental group.

Replication in the works

Nonetheless, to resolve the controversy, Lovaas is coordinating a multisite study funded by NIMH to replicate his original findings. Although originally planned for just three sites, the replication study has been expanded to 14 locations in the United States and Europe. Researchers are comparing about 150 children receiving the treatment with control subjects matched by age and IQ who are getting alternative forms of treatment.

"They are really a...representative sample of children with autism," Lovaas says. The study will take another five years to complete, but preliminary data from two or three sites may be submitted for publication in December.

Hugh McIntosh is a writer in Chicago.

Read our privacy statement and Terms of Use

Cover Page for this Issue

APA Home Page . Search . Site Map