http://207.21.243.44/feat/research3.html
Tristram Smith, Washington State University
Since 1980, 12 peer-reviewed outcome studies (nine on behavior analytic programs, one on Project TEACCH, and two on Colorado Health Sciences) have focused on early intervention for children with autism. Mean IQ gains of 7-28 points were reported in studies of behavior analytic programs, and 3-9 in studies on TEACCH and Colorado. Gains were also reported on other measures in some studies. Individual children varied greatly in treatment responsivity. No fully randomized studies have been published, and studies with features such as large sample sizes and blind examiners have been rare. Further progress is likely to require more rigorous clinical trials, more comprehensive pretreatment and follow-up assessments, and greater attention to brain-behavior relationships.
DISCUSSION
A number of investigators have reported that children with autism make major gains with early intervention. The present review included nine such reports on behavior analytic treatment, one on Project TEACCH. and two on Colorado Health Sciences. Unfortunately, close inspection of these reports reveals that the results have been less favorable than reviewers have claimed (Dawson & Osterling, 1997; Rogers. 1996). Some studies have yielded highly favorable results on every outcome measure given (notably McEachin et al., 1993). However, others have produced favorable results on some measures but not all (e.g., Harris et al., 1990, 1991). Four others, including one on behavior analytic treatment (Birnbrauer & Leach, 1993) and all three studies on other interventions (Lord & Schopler, 1989: Rogers & DiLalla, 1991; Rogers et al., 1986), have found little change. Two of these four studies (Birnbrauer & Leach, 1993; Lord & Schopler, 1988) indicated that a subgroup of participants fared well, but each study also contained a larger subgroup whose standardized test scores tended to decline or remain stable.
Substantial individual differences in treatment responsivity have emerged. In the six investigations that examined such differences (Anderson et al., 1987; Birnbrauer & Leach, 1993; Hoyson et al., 1984; Lord & Schopler, 1988; Lovaas, 1987), all indicated that some children made great progress while others did not. In studies on behavior analytic treatment, relatively high-functioning children at pretreatment were apt to make the largest gains at follow-up, but in the study on TEACCH, the lowest functioning children reportedly benefited most. Whether these trends are reliable remains to be determined.
Of great concern is that most studies have lacked even the most basic features of scientifically sound studies, such as an experimental or quasi-experimental design, reliable assessments, and replicable descriptions of the treatments that children received. Most studies also suffer from other substantial weaknesses such as small sample sizes and employment of therapists who apparently had minimal training and experience. Further, 11 of 12 did not provide data on children's progress following the termination of treatment (the exception being McEachin et al., 1993). This is a crucial omission because, even if children acquire skills in treatment, they may not continue doing so after the treatment ends, when they may no longer be receiving specialized services. Consequently, children may fail to derive long-term benefits-an outcome that would defeat the purpose of early intervention. In the present review, it was often necessary to reanalyze data to determine precisely what was found because investigators at each of the nine sites with outcome data have presented test scores in a different way, some have selectively reported positive but not negative aspects of their data in subsequent presentations of their work (e.g., Handleman & Harris, 1994: Lord & Schopler, 1994), and many have combined peer-reviewed and non-peer-reviewed data (e.g., McClannahan & Krantz, 1994; Strain & Cordisco, 1994) or presented only the latter (e.g., Koegel, 1995: McGee, Daly, & Jacobs, 1994).
The foregoing problems hinder drawing firm conclusions from existing early intervention studies. However, it appears that, as indicated in Table 2, the UCLA study (Lovaas, 1987; McEachin et al., 1993) had both the most favorable results and the strongest methodology. Though this study has generated much controversy, commentators have agreed on two points (Baer, 1993; Foxx, 1993; Kazdin, 1993: Mesibov, 1993; Mundy, 1993): The study makes a strong case that children with autism in the study made major, long-lasting improvements as a result of the treatment they underwent. However, the study needs to be replicated by independent investigators using improved methodologies. Two of three studies that have conducted partial replications of this study (employing the same treatment manual but providing fewer hours of treatment and altering some treatment procedures) have also reported favorable results (Anderson et al., 1987; Sheinkopf & Siegel, 1998). However, the third reported mixed results, with one subgroup making gains and another showing declines on standardized tests (Birnbrauer & Leach, 1993). The extent to which more exact replications would obtain results comparable to those of Lovaas (1987) remains an open question of considerable importance. At this time, though, the long-term effects of the UCLA model developed by Lovaas and colleagues have been the most rigorously documented of any treatment model for children with autism. Other behavior analytic treatment programs almost certainly Yield short-term benefits (Matson et al., 1996), and preliminary evidence indicates that they may also produce long-term benefits, while interventions based on different theoretical orientations, such as TEACCH and Colorado Health Sciences, appear essentially invalidated at this time.
