| ObjectiveThis study intended to use the data from China Health and Nutrition Survey(CHNS)to explore the relationship between multiple exposures and chronic metabolic diseases by path analysis,and to analyze the impact of mediating effects of BMI and moderating effects of gender on disease.MethodsThe survey data of CHNS in 2009 was obtained,there were 9474 respondents in dataset.According to the inclusion and exclusion criteria,a total of 8427 people was selected for analysis.The prevalence of diabetes mellitus,hypertension,mixed hyperlipidemia and central obesity were described by frequency.The Apriori algorithm of association rules was used to calculates the comorbidity of chronic metabolic diseases.Taking four diseases as outcome variable respectively,the suspected risk factors and risk factors reported in the previous literature were primary screened by t-test andχ~2 test.The web visual network map and C5.0 algorithm was used to screen risk factors for comorbidity diseases.Taking the chronic metabolic disease as the outcome variable,the suspicious risk factors selected by single factor analysis were used as independent variables,and the stepwise logistic regression method based on Wald statistic was used for further screening risk factors with statistical differences.The final screened risk factors and binary dependent variables were used to building the path map,in which,BMI was used as mediating variable,gender as moderator variable.The path map model was evaluated according to the fitness index.Statistical analyses were conducted with IBM SPSS Statistics 23.0,IBM SPSS Modeler18.0 and Mplus7.0.All P-values were 2-tailed,and the level of significance was set atα=0.05.Results1.Among the included subjects,44.7%had at least one chronic metabolic disease,34.3%of all patients had at least two chronic metabolic diseases,and 15.4%suffered from three or more chronic metabolic diseases.The probability of comorbid at least one chronic metabolic disease with diabetes mellitus,hypertension,mixed hyperlipidemia,and central obesity was 83.0%,61.5%,70.6%,and 60.5%,respectively.2.The results of the association rules showed the rules formed by male and female were basically the same.In the population with normal weight,there was one strong rule,and in the overweight/obese people,there were 10 strong rules,of which,the support and confidence of rule(hypertension and mixed hyperlipidemia→central obesity)were 11.78%and 82.20%,respectively.In the population of[18,40),[40,60)and≥60 years old,there were 0,4 and 12 strong rules,among which,in the population aged≥60 years,the support and confidence of rule(diabetes mellitus→diabetes mellitus with comorbidities)were 13.51%and 85.71%,respectively.3.The results of the Web visualization network diagram showed that there were two strong links,that was link 1:central obesity with one or more of other three chronic metabolic diseases(1695),and link 2:hypertension with one or more of other three chronic metabolic diseases(1571).There were two moderate links,that was link3:hypertension with central obesity(1289),link 4:mixed hyperlipidemia with one or more of other three chronic metabolic diseases(944).Suspicious risk factors had different degrees’importance for different disease links.BMI was of great importance to the target variables of links 1-3.4.Logistic regression showed the risk of four chronic metabolic diseases were higher with the increasing of age and BMI(P<0.001).the risk of diabetes mellitus(OR(95%CI)=0.662(0.544-0.805)),hypertension(OR(95%CI)=0.692(0.606-0.789))in female was lower than in male,but the risk of central obesity in female was higher than that of in male(OR(95%CI)=1.310(1.132-1.516)).Compared with extremely light physical activity,heavy physical activity might reduce the risk of diabetes mellitus(OR(95%CI)=0.718(0.550-0.939)).Moderate and heavy physical activity was a protective factor for mixed hyperlipidemia(Moderate:OR(95%CI)=0.772(0.608-0.980);Heavy:OR(95%CI)=0.627(0.518-0.759)).Compared with those who did not eat spicy flavor,those who ate spicy flavor 3-4 times a week had a lower risk of hypertension and central obesity(P<0.05).Compared with those who drank alcohol every day,those who drank alcohol≤once a month had a lower risk of mixed hyperlipidemia(OR(95%CI)=0.627(0.403-0.977).When logistic regression model did not consider the influence of BMI,the results were somewhat different from those of model including BMI.5.Path analysis showed that 3.20%and 13.16%of the total effects of age and physical activity on diabetes mellitus were related to BMI.The mediating effects of age and physical activity on hypertension through BMI accounted for 2.24%and28.57%of the total effects,respectively.The mediating effects of age,serum magnesium concentration and physical activity through BMI on mixed hyperlipidemia was statistically significant,which accounted for 6.82%,10.03%and10.53%of the total effects,respectively.The mediating effects of smoking on mixed hyperlipidemia through BMI were opposite to that of smoking on mixed hyperlipidemia directly.The mediating effects of age,serum magnesium concentration and physical activity on central obesity accounted for 14.12%,46.00%and 46.55%of the total effects,respectively.Conclusion1.Diabetes and mixed hyperlipidemia have the highest probability of comorbidity with other chronic metabolic diseases.Especially in the elderly and overweight/obese population.2.BMI is a mediating variable between exposure and disease,and age,serum magnesium concentration,smoking and physical activity can affect the prevalence of chronic metabolic diseases indirectly by changing BMI.Gender can regulate the effect of serum magnesium concentration on central obesity. |
PDF Full Text Request