Childhood caries have a high relapse rate after full mouth therapy. This study aimed to elucidate the relationship between the microbiome, sugar, and the relapse of childhood caries after therapy. A total of 24 children aged 2-4 years who underwent one caries treatment session participated in this study. Supragingival plaque was collected before therapy and 1 and 7 months after therapy, then sequenced using the 16S rRNA high-throughput approach. We found 11 phyla, 140 genera, and 444 species in 72 samples. The children were divided into relapse-free (n = 13) and relapse (n = 11) groups according to whether they relapsed 7 months after therapy. The bacterial community richness, diversity, structure, and relative abundance of bacterial taxa were significantly different between the two groups 7 months after therapy. The two groups also differed in the relative abundance of bacterial taxa, both before and 1 month after therapy. Bacterial community richness and diversity were lower in the relapse-free group 1 month after therapy. Using different operational taxonomic units between the relapse-free and relapse groups 1 month after therapy, a relapse-risk assessment model was built with 75% accuracy, 0.1905 out-of-bag error, and 66.67% validation accuracy. Patients in the relapse group had higher sugar intake frequencies than those in the relapse-free group during follow-up. Interactions between the microbiome and sugar intake frequency were found through co-occurrence networks. We conclude that the microbiome is significantly different between the relapse-free and relapse groups at the time of relapse. Supragingival plaque collected immediately after therapy can be used to predict the risk of relapse. Furthermore, the correlation between sugar intake frequency and microbiome is associated with the relapse. (C) 2015 Elsevier Inc. All rights reserved.