OBJECTIVE: Bimodality in blood glucose (BG) distribution has been demonstrated in several populations with a high prevalence of diabetes and obesity. However, other population studies had not found bimodality, thus casting doubt on its universality. We address this question in four ethnic populations-namely Malay, Chinese, Indian, and the indigenous people of Borneo. RESEARCH DESIGN AND METHODS: A national health survey was conducted in Malaysia in 1996. A total of 18,397 subjects aged > or =30 years had post-challenge BG measurements taken. To test whether BG was consistent with a bimodal distribution, we fitted unimodal normal and skewed distribution as well a mixture of two normal distributions to the data by age and ethnic groups. RESULTS: Age-specific prevalence of diabetes varied from 1.3 to 26.3%. In all ethnic/age groups, the bimodal model fitted the log BG data better (likelihood ratio tests, all P values <0.001). CONCLUSIONS: Bimodality in BG distribution is demonstrable even in populations with a very low prevalence of diabetes and obesity. Previous studies that found unimodality had failed to detect the second mode because of inadequate sample size, bias due to treatment of subjects with known diabetes, and inclusion of subjects with type 1 diabetes in the sample. Bimodality implies that diabetes is a distinct entity rather than an arbitrarily defined extreme end of a continuously distributed measurement.
OBJECTIVE: Bimodality in blood glucose (BG) distribution has been demonstrated in several populations with a high prevalence of diabetes and obesity. However, other population studies had not found bimodality, thus casting doubt on its universality. We address this question in four ethnic populations-namely Malay, Chinese, Indian, and the indigenous people of Borneo. RESEARCH DESIGN AND METHODS: A national health survey was conducted in Malaysia in 1996. A total of 18,397 subjects aged > or =30 years had post-challenge BG measurements taken. To test whether BG was consistent with a bimodal distribution, we fitted unimodal normal and skewed distribution as well a mixture of two normal distributions to the data by age and ethnic groups. RESULTS: Age-specific prevalence of diabetes varied from 1.3 to 26.3%. In all ethnic/age groups, the bimodal model fitted the log BG data better (likelihood ratio tests, all P values <0.001). CONCLUSIONS: Bimodality in BG distribution is demonstrable even in populations with a very low prevalence of diabetes and obesity. Previous studies that found unimodality had failed to detect the second mode because of inadequate sample size, bias due to treatment of subjects with known diabetes, and inclusion of subjects with type 1 diabetes in the sample. Bimodality implies that diabetes is a distinct entity rather than an arbitrarily defined extreme end of a continuously distributed measurement.
Authors: Sheena M Patel; Paul Cobb; Sharon Saydah; Xuanping Zhang; Janet M de Jesus; Mary E Cogswell Journal: J Nutr Date: 2015-01-21 Impact factor: 4.798
Authors: Xiuhong Li; Haitao Chu; Joel E Gallant; Donald R Hoover; Wendy J Mack; Joan S Chmiel; Alvaro Muñoz Journal: J Epidemiol Community Health Date: 2006-09 Impact factor: 3.710
Authors: Clinton C Mason; Robert L Hanson; Vicky Ossowski; Li Bian; Leslie J Baier; Jonathan Krakoff; Clifton Bogardus Journal: BMC Genomics Date: 2011-02-07 Impact factor: 3.969
Authors: Violeta L Calin; Mona Mihailescu; Eugen I Scarlat; Alexandra V Baluta; Daniel Calin; Eugenia Kovacs; Tudor Savopol; Mihaela G Moisescu Journal: FEBS Open Bio Date: 2017-09-02 Impact factor: 2.693