A Meta-Analysis of Pivotal Response Treatment As an Early Naturalistic Developmental Behavior Intervention for Autism

Background: In contrast to Early Intensive Behavioral Intervention (EIBI), there are far fewer systematic reviews and meta-analyses evaluating Naturalistic Developmental Behavioral Interventions (NDBIs), a group of behavioral interventions that (a) share critical treatment components and (b) are often deemed evidence-based (Schreibman et al., 2015). For instance, as an NDBI, stakeholders routinely consider Pivotal Response Treatment (PRT) empirically supported in treating core social-communication deficits; however, the literature lacks sophisticated meta-analytical studies aimed at ascertaining cumulative PRT evidence and effect sizes (ES) across early intervention outcomes. Indeed, stakeholders need clarity regarding PRT’s effect on outcomes beyond social-communication. Furthermore, meta-analyses should dually investigate single-case (SCD) and group designs (GD), but researchers typically do not employ the necessary methodology (i.e., hierarchical linear modeling, HLM; Shadish d-statistic). Critical to understanding PRT and individualization, such methods also allow for analyses of key moderators (e.g., child characteristics, outcomes, study quality).

Objectives: We answer the following research questions: (a) For both GD and SCD, what are PRT ESs across outcome areas and ES types?; (b) What key variables moderate PRT effects?; (c) What components best constitute PRT?; and (d) What descriptive variables contextualize the findings?

Methods: We systematically selected articles that compared PRT to no-treatment conditions for children with autism under 6 years old; 36 studies (30 SCD, 6 GD) were included. We gathered adequate inter-coder agreement at every stage of search, selection, and variable coding. We coded numerous variables from each study, and extracted SCD data point coordinates and GD statistics. For SCD, we employed HLM, Tau-U, and the Shadish d-statistic analog of GD standardized mean differences. For GD, we aggregated Cohen’s drandom effects by inverse variances.

Results: Accounting for study quality, we found medium-to-high SCD ESs across social-communication and language, play, and affect outcomes (omnibus HLM ES: γ₁₀₀  = 0.87 standardized units, p< .001), with larger estimates for social-communication and language. The GD meta-analysis indicated a moderate-sized effect across outcomes, with larger estimates for measures of communication and language. Observed utterances showed slightly larger comparable effect sizes for SCD versus GD studies. For SCD, higher study quality significantly penalized PRT effects. The HLM analysis of SCD also revealed PRT effects significantly varied between participants. We gleaned inconsistent PRT procedures across studies, but key components (i.e., choice and natural reinforcement) were ubiquitous.

Conclusions: We found early PRT was generally effective across studies and outcomes, evinced by moderate to large ESs. The social-communication and language area, however, generated the largest effect sizes (functional speech, in particular) and greatest confidence (representing 73.6% of the weighted data). PRT produced favorable play and affect outcomes; however, with slightly smaller and/or far fewer effect sizes, claims of positive PRT effects beyond social-communication and language are tenuous. Evidence for PRT’s effect on receptive language and cognitive functioning is also weak, while its effects on adaptive behavior is emerging – making it difficult to draw comparisons with EIBI meta-analyses, or understand PRT as a comprehensive early intervention. Based on our findings, we outline a model for advancing PRT and NDBI research.