Background: Caries risk assessment in preschool children has been limited validated. Aim: To validate caries predictive ability of Cariogram using different combinations of factors and settings in preschool children from areas with high caries risk. Design: Two to five years old children (N = 175) were examined for caries (cavitated and non-cavitated lesions), at baseline and after 1 and 2 years. Mutans streptococci counts (MS) and saliva buffer capacity (SBC) were measured with chair side tests. Diet and oral health attitude were assessed through a parental questionnaire. Baseline caries risk was calculated using standard and high-risk group variables in Cariogram either with nine factors or excluding MS and SBC. Poisson regression models, ROC analysis and information criteria (AIC and BIC), P < 0.05, were used to investigate the predictive ability of Cariograms and to quantify the associations. Results: Accuracy of Cariogram was found limited. Cariograms with MS showed higher sensitivity but lower specificity than those without bacterial counts. Standard set Cariograms with MS counts performed slightly better than the other models, but the difference was not statistically significant. Caries prediction with standard and high set Cariograms was also found limited. Conclusions: Cariogram with various factors and settings displayed suboptimal ability to predict caries in this population.