A pragmatic randomised controlled trial comparing a minimally invasive approach based on atraumatic restorative treatment (ART) procedures (test) was tested against the standard-care approach (control) to treat early childhood caries (ECC) in a primary-care setting in Perth, W.A., Australia. Parent/child dyads with ECC were allocated to the test or control group using stratified block randomisation. Children were examined at baseline and follow-up by two calibrated examiners blinded to group allocation status. Dental therapists trained in ART provided treatment to the test group and dentists treated the control group. Restoration quality was evaluated at follow-up using the ART criteria. Data were analysed on an intention-to-treat basis; test of proportions, Wilcoxon rank test and logistic regression, controlling for clustering of teeth, were used. Two hundred and fifty-four children were randomised (test = 127 and control = 127). There was no statistically significant difference in age, sex and baseline caries experience between the test and control groups. At follow-up (mean interval 11.4 months, SD 3.1), 220 children were examined (test = 115 and control = 105) and 597 teeth (test = 417 and control = 180) were evaluated for restoration quality, of which 16.8% (test) and 6.7% (control) were judged to have failed (required replacement; p < 0.01). Intention-to-treat, multiple logistic regression found multisurface restorations (OR = 10.4) had significantly higher odds of failure, while referral for specialist paediatric care had significantly lower odds of restoration failure (OR = 0.2). The ART-based approach enabled more children and teeth to be treated, and multisurface restoration and treatment in a primary-care setting had higher odds of restoration failure.