Objective: To propose a novel, three-level (severe, moderate, mild) case definition using probing depth (PD), clinical attachment loss (CAL) and bleeding on probing (BOP) for epidemiologic studies on periodontitis. Methods: Case definitions (DEF) 1-30 with PD, CAL and BOP were made. Based on data from epidemiologic research in Chengde (Hebei Province, China) in 1992, prevalence of periodontitis by DEF1-30 was calculated and compared with a reference (definitions by Centers for Disease Control and Prevention/American Academy of Periodontology in 2012). Sensitivity, specificity, Youden Index, Cohen's kappa coefficient (CKC) and the area under the receiver operator characteristic curve (AUC) were calculated for the definitions selected. Results: DEF1 and DEF18 for periodontitis, DEF2, DEF3, DEF19 for moderate and severe periodontitis, and DEF5, DEF13, DEF14, DEF21 and DEF25 for severe periodontitis, which were similar for estimation of periodontitis prevalence compared with the reference, were selected. DEF18 for periodontitis, DEF19 for moderate and severe periodontitis, and DEF5 for severe periodontitis were selected because they showed higher values for the Youden Index, CKC and AUC, and formed a three-level definition. Conclusion: A novel three-level case classification of periodontitis using three parameters of PD, CAL and BOP was proposed. The estimated periodontitis prevalence according to the novel proposed definition is close to the prevalence according to the CDC/AAP definition. Keywords: bleeding on probing, clinical attachment loss, definition, periodontitis, probing depth