There is little evidence regarding the effects of individual and group intervention for children with autism spectrum disorder (ASD) on important outcomes. We performed meta-analyses using a random effects model to investigate the effectiveness of the individual and group intervention studies and to compare the effectiveness of these two types if possible. The main analysis which excluded studies at a high risk of bias (Analysis I) included 14 randomised controlled trials targeting children with ASD≤6 years of age (594 children). The results suggested that both individual and group interventions showed significant effects compared to the control condition on "reciprocity of social interaction towards others" (standard mean difference[SMD] [95%confidence interval{CI}] = 0.59[0.25, 0.93], p = 0.16; 0.45[0.02, 0.88], p = 0.39, respectively). Only individual interventions showed significant effects compared to the control condition on "parental synchrony" (SMD [95%CI] = 0.99 [0.70, 1.29], p<0.01). Our results showed no significant differences between individual and group interventions in effects on "autism general symptoms" (no study available for group intervention), "developmental quotient" (no study available for group intervention), "expressive language" (p = 0.56), "receptive language" (p = 0.29), "reciprocity of social interaction towards others" (p = 0.62), or "adaptive behaviour" (p = 0.43). We also performed sensitivity analyses including studies that had been excluded due to being at a high risk of potential bias (Analysis II). The results suggested that "reciprocity of social interactions towards others" showed significant effects for individual intervention compared to the control condition (0.50[0.31,0.69], p<0.001) but not for group intervention (0.23[-0.33, 0.78], p = 0.42). Individual intervention also showed significant effects on "parental synchrony" (0.98[0.30,1.66], p = 0.005) in the sensitivity analysis. The results also suggested no significant difference on all the outcomes between the individual and group interventions. We also reanalysed the data using cluster-robust standard errors as sensitivity analyses (Analysis III). Analysis III showed no significant effects in the intervention condition compared to the control condition on all the outcomes for both individual and group interventions. When Analysis II was reanalysed using cluster-robust standard errors (Analysis IV), individual interventions showed significant effects compared to the control condition on "reciprocity of social interaction towards others" and "parental synchrony" (mean estimate[95%CI], robust standard error, p = 0.50[0.20, 0.81], 0.13, 0.006; and 1.06[0.08, 2.05], 0.42, 0.04, respectively), and none of the outcomes showed significant effects under the intervention condition compared to the control condition for group interventions. The discrepancies in the results between the main analysis (Analysis I) and the sensitivity analyses (Analyses II, III, and IV) may be due to the small number of included studies. Since the outcome of "reciprocity of social interaction towards others" can be a dependent variable that is usually measured in a context-bound setting with the child's parent, we cannot conclude that individual interventions for pre-school children with ASD have significant effects on generalised skills for engaging in reciprocal interactions with others, even if the interventions have significant effects on the outcome. However, the outcomes of "reciprocity of social interaction towards others" may be promising targets for both individual and group interventions involving pre-school children with ASD. "Parental synchrony" may also be a promising target for individual interventions.
Trial registration: (CRD42011001349).