| Objective: Type 2 diabetes mellitus(T2DM)has become one of the major public health problems in China,early detection and prevention are the key to diabetes health management.Diet is identified as the most important controllable factor of diabetes.Compared with single food or nutrient,dietary pattern can accurately reflect the impact of overall diet,providing dietary factors in the etiology of the disease.Nutritional metabolomics,as an alternative or supplementary objective method for epidemiological dietary surveys,can validate the relationship between dietary factors and T2 DM,understand the biological mechanism between diet and disease,and facilitate the risk assessment and precise nutritional interventions of diabetes.Diabetes risk assessment models mainly include age,gender,and body mass index and so on,and lack dietary factors and other modifiable health behavior factors to predict and prevent the development of diabetes;the models also lack visual presentation,which is not convenient for application and promotion in primary medical units.Therefore,the obejectives of this study were to:(1)explore T2 DM prevalence and risk factors for T2 DM among community residents,establish non-laboratory risk assessment models for T2 DM and develop diabetes prediction nomogram,identifying individuals at high risk of diabetes and applying in practice;(2)analyze associations of dietary patterns with T2 DM,develop and validate dietary pattern-driven prediction nomogram for diabetes so as to optimize the models and improve the discrimination effect,providing scientific basis of dietary interventions in health management;(3)identify metabolites associated with dietary pattern by conducting nutritional metabolomics profiling,analyze the relationship between the identified dietary-pattern-related metabolites and T2 DM,and evaluate whether nutritional metabolomics could optimize the model by finding new predictors from the perspective of nutritional metabolomics,further illustrating the importance of rational dietary nutrition and providing targets in terms of precision nutrition intervention.Methods: Relying on the National Key Research and Development Program of China,a survey was conducted in residents aged 35 years or older in Shenyang,Liaoning Province.Questionnaires,physical examinations and biological sample collection were completed.Descriptive analysis was used to describe the basic characteristics,such as social/economic/cultural risk factors,behavioral risk factors and metabolic risk factors,and T2 DM prevalence among study participants.Using the semi-quantitative food frequency questionnaire,exploratory factor analysis was used to construct dietary patterns,and the scores of different dietary patterns of the study subjects were calculated according to the factor load of each food item and the intake of standardized food groups;based on the extracted dietary patterns,the simplified scoring construction method was used to construct healthy dietary patterns,and logistic regression was used to analyze the relationship between dietary patterns and T2 DM.Diabetes cases and controls were selected using propensity score matching,and untargeted plasma metabolomics assays were conducted;the metabolite values were transformed using a rank-based inverse normal transformation,and then examined associations of dietary patterns with single metabolite using linear regression models,the yielded p values were corrected for false discovery rate using the Benjamini–Hochberg procedure;using principal component analysis,partial least-squares discriminant analysis and orthogonal partial least-squares discriminant analysis,the most influential metabolites were identified by variable importance in projection score≥2.5,and dietary-pattern-associated metabolites were identified by the above results.Logistic regression model was used to construct the risk assessment model for T2 DM,the subjects were randomly divided into training set and validation set according to 2:1,and the area under curve(AUC),Hosmer-Lemeshow test(H-L test),decision curve analysis(DCA),net reclassification improvement(NRI),integrated discrimination improvement(IDI)and the bootstrap method were applied to validate and evaluate the model in the training set and validation set,and the nomogram was used for visual presentation.All statistical analyses were conducted using SPSS 21.0 and Stata 16.0,P<0.05 was considered statistically significant unless otherwise noted.Results: 1.A total of 3587 participants were included in this study,1245 were males and 2342 were females.The prevalence of T2 DM was 14.13% in total,17.91% in males,and 12.13% in females,differences across gender were evident(P<0.001).The results of multivariate analysis showed that gender,age,family history of diabetes,smoking,obesity and hypertension were the influencing factors of T2 DM.The total,male,and female risk assessment models for T2 DM were conducted,separately.Risk assessment model in total population: Logit(P)=-3.10+0.44*(gender)+0.86*(age)+1.37*(family history of diabetes)+0.48*(obesity)+0.55*(hypertension);risk assessment model in males: Logit(P)=-2.69+0.67*(age)+0.95*(family history of diabetes)+0.61*(smoking)+ 0.50*(obesity)+0.52*(hypertension);risk assessment model in females: Logit(P)=-3.30+ 1.02*(age)+1.54*(family history of diabetes)+0.59*(obesity)+0.65*(hypertension),the AUC was 0.72,0.69 and 0.75,respectively,and P>0.05 in H-L test,the model curves were higher than the extreme curves by using DCA,and the internal validation showed the risk assessment had excellent prediction performance.The use of nomogram to draw visual predictive models was helpful for the application and promotion of risk assessment models.2.The study extracted two dietary patterns,animal food dietary patterns including beef and mutton,river fish,marine fish,poultry,shrimp and crab shellfish,animal offal,fried food,seaweed,pork;fruits,vegetables,eggs and milk dietary patterns including fruits,eggs,vegetables,fresh milk,dairy products,nuts,soy products,and seaweed.All dietary pattern scores were divided into T1,T2 and T3.The highest prevalence of T2 DM was found in the T3 group of animal food dietary patterns.Compared with the T1 group,the risk of disease was increased in the T3 group(OR=1.50,95% CI: 1.11-2.03)and the risk of T2 DM increased significantly with the increase of animal food dietary pattern scores(P trend<0.05).Fruits,vegetables,eggs,and milk dietary pattern of was inversely correlated with risk of T2DM(OR=0.61,95% CI: 0.45-0.83,P trend<0.05).In males and women,two similar dietary patterns were extracted.In males,animal food dietary patterns were significantly positively correlated with T2DM(OR=1.80,95% CI: 1.13-2.87,P trend <0.05),and fruit,vegetables,eggs,and milk dietary patterns were nonsignificantly associated with diabetes(P>0.05).In females,animal food dietary patterns were nonsignificantly associated with T2DM(P>0.05),and fruits,vegetables,eggs,and milk dietary patterns were protective factors for diabetes(OR=0.45,95% CI: 0.30-0.69,P trend <0.05).On this basis,a simplified scoring construction method was used to construct healthy dietary patterns,and healthy dietary patterns were found to be negatively correlated with T2DM(P<0.05)in total,males and females.Dietary pattern-driven risk assessment models for T2 DM were conducted and nomogram were established,risk assessment model in total population: Logit(P)=-3.20+0.41*(gender)+ 0.85*(age)+1.04*(family history of diabetes)+0.83*(obesity)+0.79*(hypertension)-0.35*(healthy dietary patterns);risk assessment model in males: Logit(P)=-2.70+0.80*(age)+0.98*(family history of diabetes)+0.47*(smoking)+0.57*(obesity)+0.63*(hypertension)-0.63*(healthy dietary pattern);risk assessment model in females: Logit(P)=-3.21+1.02*(age)+1.44*(family history of diabetes)+0.75*(obesity)+0.82*(hypertension)-0.59*(healthy dietary pattern),and the internal validation results of this model were good.Compared with the models of common impact factors,the discrimination of the dietary pattern-driven risk assessment model for T2 DM was improved,and NRI was significantly increased by 16.35%(P<0.05).Among the gender stratifications,the discrimination the new model in males was improved,and IDI and NRI increased by 1.16% and 21.46%,respectively(P<0.05).In females,the AUC of the dietary pattern-driven risk assessment model for T2 DM was significantly improved,and NRI was significantly increased by 18.28%(P<0.05).3.302 subjects were included by propensity score matching,including 151 diabetes cases and 151 controls,and there was no statistically significant difference in the basic characteristics of the study subjects in the case and the control groups(P>0.05).After adjusting for factors including gender,age,education and annual household income,linear regression analysis showed that 30 metabolites were significantly correlated with healthy dietary patterns(FDR P<0.05),of which 7 metabolites were positively correlated with healthy dietary patterns and 23 metabolites were negatively associated with healthy dietary patterns.According to VIP score≥2.5,31 metabolites were found.Based on the above results,21 dietary-pattern-associated metabolites were identified,including D-(+)-mannose and L-(+)-Tartaric acid.The main metabolic pathways were glycolysis/gluconeogenesis,pentose and glucuronate interconversions,starch and sucrose metabolism.The plasma composite nutrient metabolites score was negatively correlated with healthy dietary patterns and associated with increased risk of diabetes(P<0.05).Nutritional metabolomics-driven risk assessment model for T2 DM was conducted and nomogram was established: Logit(P)=-1.25+0.09*(gender)+0.11*(age)+0.06*(family history of diabetes)+0.17*(obesity)+0.29*(hypertension)+3.57*(composite nutrient metabolites score),and the AUC in the training set and validation set was 0.80(95% CI:0.74-0.86),0.79(95% CI: 0.79-0.88),respectively,P>0.05 in H-L test,and the model curves were higher than the extreme curves by using DCA,and internal validation showed the risk assessment had excellent prediction performance by using bootstrap method.Compared with the dietary patterns constructed by the questionnaire data,the discrimination of the nutritional metabolomics-driven risk assessment model for T2 DM was significantly improved(P<0.05),and NRI and IDI were significantly improved(P<0.05),indicating that the discriminant effect of the model was significantly improved after including nutritional metabolomics.Conclusion: 1.The prevalence of T2 DM was high among community residents,and prevalence was higher in males than in females.Gender,age,family history of diabetes,smoke,obesity and hypertension were risk factors for T2 DM.2.Two dietary patterns were derived,animal food dietary patterns and fruit,vegetable,egg and milk dietary patterns,animal food dietary patterns increased the risk of T2 DM,and fruit,vegetable,egg and milk dietary patterns reduced the risk of T2 DM.3.Compared with the risk assessment model for T2 DM including conventional indicators,the discrimination and calibration of the model including in dietary indicators were improved.4.Twenty-one dietary-pattern-associated metabolites were identified.Compared with dietary patterns,inclusion of nutritional metabolites impoved discriminant effect,the internal verification showed the risk assessment model had excellent performance,and the nutritional metabolites could be used as targets for future diabetes precision nutrition intervention. |
PDF Full Text Request