This study compares three common anthropometric measures; body mass index (BMI), waist-to-height ratio (WHtR), and waist circumference (WC) in predicting type 2 diabetes among 8 different sub-population stratified by age, sex, and site locations in Bangladesh. Community based surveys including health checkups, questionnaires, and telemedicine services were conducted for more than 16,000 individuals in 16 nationwide service sites in Bangladesh during 2012 and 2014. Participants were randomly selected factory and office workers in Dhaka and suburban of Dhaka and also voluntary recruited community residents in 6 different rural areas. These data were stratified into 8 different sub-population groups by sites (rural vs. urban), sex (male vs. female), and age (<40 years old vs. ≧40 years old). Participants who were BMI≧25 ≧24 or ≧23 were significantly more likely than others to be type 2 diabetes in 5 sub-population groups out of 8 total sub-groups. The ORs for BMI with type 2 diabetes were between 1.585 (95%CI: 1.12-2.25) in rural women (≧40 yrs) and 5.325 (95%CI: 2.19-12.95) in rural men (<40). Participants who were WHtR≧0.50 and ≧0.53 were significantly more likely than others to report type 2 diabetes in all sub-groups except for the urban women (≧40 yrs) group which has the smallest sample size (n=202). The ORs for WHtR vary between 1.975 (95%CI: 1.06-3.70) in urban men (<40 yrs) and 7.959 (1.90-33.43) in urban men (≧40 yrs). Similarly, the association of WC≧90 and ≧80 with type 2 diabetes were significant in all sub-groups, except for the urban women(≧40 yrs) group. The study suggests WHtR and WC are better indicators than BMI for prediction of type 2 diabetes in Bangladesh.