For nearly four decades, study after study showed that for many young children diagnosed with autism, more hours of treatment using applied behavior analysis (ABA) methods yielded better outcomes than less-intensive treatment. Studies conducted by multiple research groups showed that many children who received at least thirty hours a week of comprehensive ABA intervention for at least two years made larger improvements than similar children in control or comparison groups who received other interventions. Those findings led several parent and professional organizations to endorse early intensive ABA intervention. But some have criticized the approach for being demanding and expensive, and some recent reports appear to call into question the benefits of intensive intervention for young children with autism.
One such study from Catalight Research Institute, published recently in the Journal of Autism and Developmental Disorders, examined the healthcare records of 725 children with autism ages 2 – 12 who had received ABA intervention through community-based providers. The authors reported finding a weak association between treatment hours and outcomes reported by parents and therapists. For adaptive behavior and reductions in dangerous behavior, the correlation was statistically negligible. Only for goal attainment did the researchers find a modest link. “More isn’t always better,” said Doreen Samelson, Catalight’s Chief Clinical Officer and a co-author of the study.
Questions of Method
The finding has not gone unchallenged. Researchers have long recognized that only limited conclusions can be drawn from retrospective reviews of records on a selected group of treatment recipients that were collected for billing and clinical purposes rather than research. That is especially true when there is no comparable group that did not receive the treatment, like the study by Samelson and colleagues. Retrospective studies can identify factors that may be associated (correlated), but they cannot establish causation. The investigators in such cases are, by definition, looking backward at data that were never meant to answer a research question, such as whether apparent outcomes were due to a treatment or something else.
Dr. Gina Green, a former president of the Association for Behavior Analysis International whose own research helped establish the evidence base for intensive ABA intervention, raised concerns about retrospective chart reviews like the Catalight study in an interview with Acuity Media Network. “They’re not controlled studies,” she said. “They’re not done from a point in time forward to address a specific research question, which would be a prospective study.”
Green also noted that Catalight manages a provider network for a major health plan in Northern California, and the records used in the study apparently came from some of those providers. That raises additional questions about the Catalight study methods, she said, because the report does not indicate whether standard procedures to limit investigator bias in retrospective studies were used in that study. Additionally, there were no direct, objective measures of child outcomes, only indirect assessments completed by parents and ratings of “goal attainment” by the therapists who treated the children.
The Scientific Evidence Base
A 2026 meta-analysis by Sigmund Eldevik and colleagues, published in Autism Research, tells a different story. The authors pooled individual participant data from fifteen controlled, prospective studies involving 621 children with autism ages 2 – 6, 341 of whom received early intensive ABA intervention and 280 in comparison groups who did not receive that intervention. The studies were conducted by several different sets of investigators in nine different countries and were published in peer-reviewed journals between 1987 and 2023. Unlike the Catalight study, these were controlled studies designed from the outset to evaluate the effects of ABA intervention and compare them to the effects of either typical services or eclectic interventions. Most of the researchers used direct, objective measures of child functioning as well as caregiver assessments of adaptive skills. Eldevik and colleagues found that treatment intensity—the term preferred by many in the field over “dosage,” which Green called “not great”—was significantly associated with better outcomes across every measure examined.
The numbers were striking. Children who received high-intensity ABA intervention, defined as twenty-six to forty hours weekly, gained an average of 22.4 IQ points, compared with 14.6 for the moderate-intensity and 11.1 for low-intensity groups. Nearly half of the high-intensity group reached the non-clinical range of intellectual functioning; for low intensity, the figure was twenty-three percent. The pattern held for adaptive behavior and autism severity. “The Eldevik meta-analyses—the one from 2010 and now the new one—provide the strongest evidence available about the effects of early intensive ABA,” Green said.
Subject matter experts who wrote a Council of Autism Service Providers (CASP) white paper on intensity of early ABA intervention pointed out that no single study can answer broad questions like whether a treatment is generally effective and if so, how effective and for whom. To answer those kinds of questions, it’s necessary to evaluate the entire body of relevant research and the rigor of the methods used in those studies, as Eldevik and colleagues did. Strong evidence comes from prospective studies with well-defined treatment and control or comparison conditions and valid, reliable measures of participant outcomes selected to address specific research questions. The white paper notes that in retrospective studies like the recent one by Samelson and colleagues, “…the researcher can only sort through whatever data already exists. Often those data were collected for non-research purposes…[in comparison to prospective, controlled studies], The researcher is less able to control the quality of measures, values of the independent variable [treatment], or biases that could impact the data… Because adherence to important characteristics of scientific inquiry is often limited, retrospective studies are more likely to provide incomplete information or conclusions that are more vulnerable to bias.”
CASP’s practice guidelines—the generally accepted standards of ABA health care services for individuals with autism, to which Green was a contributor—draw a distinction between “comprehensive” and “focused” ABA. Comprehensive treatment, which is typically delivered to young children, targets multiple skill domains. The best available evidence from controlled studies like those analyzed by Eldevik and colleagues shows that thirty to forty hours of direct treatment weekly for 1-2 years is more likely to produce clinically important improvements in most domains than less intensive treatment. Focused ABA treatments typically address small numbers of specific behaviors and may require ten to twenty-five hours of direct treatment per week. The guidelines emphasize that treatment intensity should be individualized: “When there is uncertainty regarding the appropriate level of service intensity,” they state, “the practitioner should err on the side of caution by providing a higher level of service intensity.”
The Stakes
The debate extends beyond the academic. High-intensity treatment models require substantial staffing and carry significant costs for insurers and families. Catalight operates one of the nation’s largest behavioral health networks, serving twenty thousand clients daily. In 2024 the company released its own practice guidelines suggesting that research did not support more than fifteen hours weekly for most cases. The recent Catalight retrospective study appears to be consistent with that position, but contrasts with the scientific evidence base and professional standards. Green, for her part, noted that most generally accepted standards of care are based on the best available scientific research. She worries that reports like those from Catalight might be used to justify low-intensity treatment for financial rather than clinical reasons. Cost-effectiveness studies, she noted, consistently show that intensive early ABA intervention saves money over the long term—across healthcare, education, and adult services—by helping children develop skills that reduce their need for support later in life.
On one point, at least, both sides agree: blanket prescriptions are problematic. Green drew an analogy to medicine. A good physician prescribes a medication for a patient’s condition that scientific research shows is generally effective, then individualizes the dosage based on the patient’s age, weight, and other characteristics. The same principle applies to ABA. “Behavior occurs at the level of the individual,” she said. “Behavior analysis by its very nature is highly individualized.” That is emphasized repeatedly in the CASP guidelines. A child who thrives with moderate-intensity intervention at age three might need more intensive services at thirteen if challenging behaviors emerge or new skills are needed for successful participation in school or activities with peers. The 2026 Eldevik meta-analysis found that children’s characteristics at intake—IQ and adaptive behavior—predicted their outcomes. The Eldevik analysis provides benchmarks that parents and clinicians can use to make decisions about treatment intensity for each individual, not mandates.
Samelson has called for “respectful open minded discussion of conflicting study results.” Neither her study nor the Eldevik meta-analysis, she wrote, represents the final word. Both point to the need for more research—particularly on outcome measures for individuals with more severe disabilities, a population she believes current tools may not adequately capture.
What remains clear is that evidence from multiple controlled studies and meta-analyses indicate that in general, more hours of ABA intervention tend to produce better outcomes for many young children with autism. Researchers are still working out all the variables that might predict outcomes from ABA as well as other approaches to autism intervention.
