A Multi-level Model for Familial Aggregation of Obesity

• Main Authors: Ping-Yun Tsai, 蔡秉芸
• Other Authors: 陳秀熙
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/94933372592890664352

碩士 === 國立臺灣大學 === 預防醫學研究所 === 96 === Background and Study Purpose Obesity is increasing in prevalence worldwide and is known to be associated with morbidity and mortality in relation to cardiovascular disease. The etiology of obesity is complex and is not completely understood. A large portion of epidemiologic research put emphasis on individual-level risk factors. However, group-level or macro-level variables, so-called contextual factors, also play an important role through interaction with individual factors. There are also limited studies taking both familial aggregation and contextual factors of obesity into accounts. Therefore, the aim of the present study is to explore the association between obesity and familial aggregation and risk factors at individual-level and area-level by conducting a community-based study with multi-level model analysis. Materials and methods A total of 74,833 subjects with aged 20-69 years old are identified from Keelung community-based integrated screening program (KCIS) between 1999 and 2005. By dint of KCIS study, the study design is based on a case-control proband family sampling. A total of 4,499 cases and 16,932 controls were identified. Data of household registration, demographics including education and marital status, lifestyle, and diet were collected. Anthropometric measurements were taking and the criteria of obesity was defined as body mass index≧27 kg/m2. Area-level contextual factors including high educational rate, divorce or widowed rate, and population density separated with tertile were collected from seven administrative districts of Keelung City. We applied nonlinear mixed model and Bayesian analysis for multi-level analysis to investigate the odds ratios and 95% confidence interval of familial aggregation and different level factors for obesity. Results The prevalence rate of overweight and obesity were higher in aged, male, low educated, divorced or widowed subjects, the least tertile of high education rate, and the most tertile of population density among areas. The relative risk of familial aggregation in association with obesity among relatives in case proband families compared with control proband families was 1.29 (95% CI: 1.29-1.30). The relative risk of familial aggregation with obesity in the least tertile of high education rate, the most tertile of divorce rate, and the most tertile of population density were 1.39 (95% CI: 1.36-1.41), 1.36 (95%CI: 1.35-1.38), and 1.34 (95% CI: 1.32-1.35), respectively. The risk for obesity among relatives in case versus control proband families was 2.31 (95%CI: 1.67-3.20) after controlling for significant environmental factors, and it was modified by individual marital status and high education rate of area. The odds ratio was 1.52 (95%CI:1.10-2.11) in married subjects, 1.44 (95%CI:1.04-2.00) in divorced or widowed subjects, and 2.68 (95%CI:1.94-3.71) in the least tertile of high education rate, respectively. When Bayesian analysis for multi-level model is applied, the random effects considering unexplained heterogeneity among different families, different areas, and different area effect on familial aggregation are taken into account with better goodness of fit than others. Conclusion The present study confirmed a strong tendency to familial aggregation for obesity by using the case-control proband family study with a multi-level model approach. The risk of obesity was heterogeneous among families and areas by using multi-level analysis, and familial aggregation of obesity was also affected by contextual factors. The selection of more contextual factors from different levels and the selection of the appropriate contextual factors are needed in the future.