Objectives: Fluoride varnish (FV) is efficacious in caries prevention although its effects among different tooth surfaces are poorly understood. This study sought to determine the extent to which caries-preventive effects of a community intervention that included FV application among preschool-aged children varied according to primary tooth anatomy and baseline tooth pathology. Methods: Secondary analysis was undertaken of data from a community-randomized controlled trial among 543 3- to 5-year-old Aboriginal children in 30 Northern Territory Australian communities. Children in intervention communities received community health promotion and FV application once every 6 months. Net caries (d3mfs) risk and 95% confidence limits (CL) were estimated for the control and intervention arms, and stratified according to tooth anatomy/location and baseline pathology (sound, enamel opacity, hypoplastic defect or precavitated carious lesion). The intervention’s efficacy was quantified using generalized estimating equation modeling accounting for study design and clustering. The assumption of efficacy homogeneity was tested using a Wald χ2 test with a p < 0.2 criterion and post hoc pairwise comparisons. Results: The intervention resulted in a 25% reduction (relative risk, RR = 0.75; 95% CL = 0.71, 0.80) in the 2-year surface-level caries risk. There was substantial heterogeneity in FV efficacy by baseline surface pathology: RRs were 0.73 for sound, 0.77 for opaque, 0.90 for precavitated, and 0.92 for hypoplastic surfaces. Among sound surfaces, maxillary anterior facials received significantly more benefit (RR = 0.62) compared to pits and fissures (RR = 0.78). Conclusion: The intervention had greatest efficacy on surfaces that were sound at baseline. Among those sound surfaces, maxillary anterior facials received most caries-preventive benefit.