Metabolomic Analysis Of Diabetic Retinopathy And Metabolic Syndrome And Clinical Prediction Modeling Of Diabetic Retinopathy

Part ⅠAnalysis of the correlation between Metabolic Syndrome and Diabetic RetinopathyObjectives: To analyze the correlation between metabolic syndrome(MS)and its components and the risk of diabetic retinopathy(DR)in patients with type 2diabetes mellitus(T2DM).Methods: To retrospectively analyze the clinical data and color fundus photography results of 2441 patients with T2DM hospitalized in the Department of Ophthalmology and Endocrinology of the Second Affiliated Hospital of Kunming Medical University from January 2019 to October 2022.T2DM patients were divided into DR and NDR groups based on color fundus photography results.The consistency of the four MS diagnostic criteria in detecting MS in T2DM patients was assessed by the Fleiss kappa test.The prevalence of MS and its components in T2DM patients was calculated based on the diagnostic criteria with higher MS detection rate.Univariate analysis was performed to compare the differences in the prevalence of MS and its components between the NDR and DR groups.Multivariate logistic regression analysis was performed to adjust for possible confounding factors such as sociodemographic data,laboratory data and relevant medical history,and to calculate and analyze the odds ratio(OR)and 95% confidence interval(CI)of MS and its components with respect to the risk of DR.Further,the study subjects were stratified by gender and severity of disease to explore the correlation between MS and its components and the risk of DR in patients with different gender and different stages.Finally,the number of MS components was included in the logistic regression model as an ordered multicategorical variable,and the changes in the risk of DR with increasing number of MS components were calculated and analyzed.Results: In this study,2210 patients with T2DM were included according to the inclusion and exclusion criteria,and 619(28.01%)of them were combined with DR.The Fleiss kappa concordance test showed that the four criteria had a high degree of concordance for MS detection rate in all subgroups of T2DM patients,especially the NECP-ATP Ⅲ criteria had the highest detection rate.Therefore,based on the calculation and analysis of NECP-ATP Ⅲ criteria,the main results are as follows:(1)Among 2210 patients with T2DM,619 patients with DR,514 of whom also had MS,and the prevalence of MS was 83.00%.The prevalence of elevated blood pressure,elevated triglycerides(TG),and reduced high-density lipoprotein cholesterol(HDL-C)was statistically significantly higher in patients with DR than in those with NDR(P<0.05).After adjusting for confounders by multivariate logistic regression analysis,combined MS,hypertension,elevated TG and reduced HDL-C were all significantly and positively associated with the risk of DR,with ORs and 95% CIs of 2.025(1.462-2.806),2.879(2.316-3.578),2.259(1.803-2.829),2.500(1.993-3.136),respectively.And there was a trend of increasing the risk of DR occurrence with the increase of the number of MS components.(2)Among 1410 male patients,411 were DR patients,of which 324 were combined with MS at the same time,and the prevalence of MS was 78.80%.The prevalence of MS and elevated blood pressure and TG were significantly higher in DR patients than in NDR patients,and the difference was statistically significant(P< 0.05).After adjusting for confounders by multivariate logistic regression analysis,combined MS,hypertension,elevated TG,decreased HDL-C,and increased waist circumference were positively associated with the risk of DR,with ORs and 95% CIs of 2.413(1.611-3.614),2.928(2.233-3.902),2.768(2.037-3.700),2.435(1.830-3.242),and 2.009(1.544-2.616).The risk of DR in male patients tended to increase as the number of MS components increased.(3)Among 800 female patients,208 patients with DR,190 of whom also had MS,had a prevalence of MS of 91.30%.The prevalence of MS and elevated blood pressure,elevated TG,and decreased HDL-C were significantly higher in patients with DR than in patients with NDR,with statistically significant differences(P< 0.05).After adjusting for confounders by multivariate logistic regression analysis,combined MS,hypertension,elevated TG,and reduced HDL-C were all positively associated with the risk of DR,with ORs and 95% CIs of 1.189(1.094-2.380),2.353(1.582-3.500),1.594(1.082-2.348),and 2.842(1.809-4.466).With increasing number of MS components,the risk of DR was significantly increased in female patients only when5 components were combined at the same time compared to patients without MS(P=0.003).(4)Of the 619 DR patients,81 were PDR patients,of which 67 were also combined with MS.The prevalence of MS was 82.72%,which was lower than the prevalence of MS in NPDR patients(83.09%),and the difference was not statistically significant(P=0.934).Conclusions: There was a correlation between MS and the occurrence of DR in patients with T2DM,but no significant correlation was seen between MS and the severity of DR.Three MS components(elevated blood pressure,elevated TG,and decreased HDL-C)were correlated with the occurrence of DR.The risk of DR tended to increase with increasing number of MS components,especially in male patients.Part Ⅱ Diabetic Retinopathy and combined Metabolic Syndrome in patients with Metabolomic analysisObjectives: To analyze the changes of small molecule metabolites and metabolic pathways in peripheral plasma of patients with diabetic retinopathy(DR)and combined metabolic syndrome(MS)by metabolomics techniques,and to screen potential biomarkers that may be used to diagnose disease onset and progression.Methods: In this part,91 patients without comorbid DR(NDR group),57 patients with comorbid MS according to NECP-ATP Ⅲ criteria,and 91 patients with comorbid DR(DR group: 41 patients with PDR and 50 patients with NPDR),78 patients with comorbid MS were collected.91 age-and sex-matched controls with no history of diabetes(Control group),51 patients with comorbid MS were also selected.Sociodemographic data and related information were collected from all study subjects,and fasting venous blood was collected.Socio-demographic data and related information were collected from the study subjects,and fasting venous blood was collected for analysis of plasma small molecule metabolites by gas chromatography time-of-flight mass spectrometry(GC-TOFMS).After data preprocessing,the Metabo Analyst 5.0 platform,combined with the Kyoto Encyclopedia of Genes and Genomes(KEGG)and The Human Metabolome Database(HMDB)were used to screen for possible identify differential metabolites and significantly altered metabolic pathways in patients with DR and comorbid MS.Spearman’s correlation analysis further clarified the correlation between differential metabolites and clinical indicators.K-means clustering combined with logistic regression analysis was used to screen potential biomarkers for disease onset and progression.Results: A total of 1355 peaks were detected by GC-TOFMS technique,and163 metabolites were matched to the Fiehn Binbase database after data pre-processing.Principal component analysis and partial least squares discriminant analysis score plots showed significant sample separation in the NDR vs.Control group,DR vs.Control group,PDR vs.NDR group,PDR vs.NPDR group,and DR+MS vs.MS group,with significant differences in plasma metabolic characteristics between the groups(P<0.05).The main findings were as follows:(1)A total of 19 differential metabolites containing L-asparagine and L-glutamate were detected in the NDR group compared with the Control group.Pathway enrichment analysis showed that seven metabolic pathways,including valine,leucine and isoleucine biosynthesis,were significantly altered(P < 0.05).19 differential metabolites were detected in the DR group,and a decrease in 1,5-AG helped identify DR patients from the normal population(AUC = 0.968).Pathway enrichment analysis showed significant alterations in 8 metabolic pathways including aminoglycan and nucleotide sugar metabolism(P < 0.05).(2)A total of 20 differential metabolites were detected in the PDR group compared with the NDR group.Logistic regression analysis showed that the combination of L-glutamine,N-acetylmannosamine and oxoadipic acid was more effective in diagnosing PDR(AUC=0.956).Pathway enrichment analysis showed that five metabolic pathways,including fatty acid biosynthesis,were significantly altered(P < 0.05).(3)A total of 14 differential metabolites were detected in the PDR group compared with the NPDR group.Logistic regression analysis showed that the combination of inositol,cysteine glycine and L-glutamine had the best efficacy in diagnosing PDR(AUC=0.952).Pathway enrichment analysis showed that six metabolic pathways,including tryptophan metabolism,were significantly disrupted(P< 0.05).(4)In comparison with the MS group,a total of 16 differential metabolites were detected in the DR combined with MS group,significantly enriched in 6pathways including tyrosine and tryptophan biosynthesis(P < 0.05).Conclusions: The peripheral plasma metabolic profile is altered in patients with NDR and DR(PDR,NPDR)and combined MS.Abnormal amino acid and glucose metabolism play an important role in the development and progression of DR and combined MS patients,and some patients also have a combination of disorders of unsaturated fatty acid metabolism and phosphatidylinositol metabolism.Part Ⅲ Clinical prediction model for Diabetic RetinopathyObjectives: To analyze the risk factors associated with diabetic retinopathy(DR)and try to develop a Nomogram prediction model for the risk of DR in patients with T2DM,and conducted an internal validation and discussion of the prediction model.Methods: Data related to 1456 patients with T2DM(demographic characteristics and physical examination,laboratory test indexes,color fundus photographic images and dynamic glucose parameters)recorded in the hospital electronic case system of the Second Affiliated Hospital of Kunming Medical University from May 2020 to February 2022 were collected.According to the inclusion and exclusion criteria,the enrolled study subjects were randomly divided into the training and validation sets according to a 5:1 ratio.The study subjects were divided into DR group and NDR group according to whether the color fundus photography pictures showed DR fundus changes.(1)Establishment of DR risk prediction model: The training set data were subjected to single-factor analysis and multi-factor logistic regression analysis in order to explore the risk factors associated with the occurrence of DR and to establish the logistic regression prediction model of DR risk.To evaluate the DR prediction model: the receiver operator characteristic curve(ROC)of the predicted probability of DR risk was plotted,and the area under curve(AUC)was used to assess the discrimination of the prediction model;the Hosmer-Lemeshow goodness-of-fit test(H-L)and calibration plot were used to assess the accuracy of the prediction model;decision curve analysis(DCA)was used to assess the clinical validity of the prediction model.Then the prediction model was displayed in the form of Nomogram.(2)Internal validation of DR risk prediction model: The validation set data were brought into the established DR risk prediction model,and the ROC curve,H-L test and calibration plot,and DCA curve were further used to verify the discrimination,calibration and clinical validity of the prediction model.Results: This study finally included 1257 patients with T2DM,including 1048 cases in the training set and 209 cases in the validation set,and no significant differences(P> 0.05)were seen in the high agreement of the respective variables in both datasets.(1)Risk prediction model of DR based on the training set data:multifactorial logistic regression analysis showed that longer duration of diabetes,increased LDL-C,increased HBA1C%,treatment of diabetes,increased urinary albumin creatinine ratio level,positive urine glucose and age of onset >40 years were independent risk predictors of DR.Binary logistic regression analysis risk prediction model equation for DR: logit(p)=-3.802-0.396 × age at onset(>40)+ 0.876 ×glycated hemoglobin(≥7%)+ 0.201 × LDL cholesterol + 0.529 × urinary albumin creatinine ratio(30-300)+ 1.277 × urinary albumin creatinine ratio(>300)+ 0.638 ×diabetes duration(10-19)+ 1.533 × diabetes duration(20-29)+ 1.011 × diabetes duration(≥30)+ 0.871 × diabetes treatment(oral hypoglycemic drugs)+ 1.083 ×diabetes treatment(insulin-based)+ 1.028 × diabetes treatment(combination therapy)+ 0.366 × positive urine glucose.The evaluation results of the model showed that: the AUC value of the model predicted the probability of DR risk was 0.738 > 0.7,suggesting that the model had good discriminatory ability;the H-L test showed that P= 0.835 > 0.05,suggesting that the model fit was good;the calibration plot showed that the model curve was close to the reference line,suggesting that the model predicted the DR risk more accurately.The DCA curve is located at the upper right of the two gray extreme lines,and its net benefit is 0 ～ 0.23 when the threshold probability is 0.04 ～ 0.94,suggesting that the prediction model has some clinical validity in the training set.(2)Internal validation of the DR risk prediction model based on the validation set data: the AUC value of the Nomogram model for predicting the probability of DR risk was 0.742 > 0.7,suggesting good discrimination of the model;the H-L test showed that P=0.251 > 0.05,suggesting good model fit;the calibration plot showed that the model curve deviated insignificantly around the reference line,suggesting good calibration of the Nomogram model.The DCA curve is located at the upper right of the two extreme lines,and its net benefit is 0 ～ 0.23 when the threshold probability is in the range of 0.09 ～ 0.87,suggesting that the Nomogram model also has some clinical value in the validation set data.Conclusion: Age at onset of diabetes,glycated hemoglobin,LDL-C,duration of diabetes,diabetes treatment,urinary albumin creatinine ratio level and urine glucose are independent risk predictors of DR.The DR Nomogram prediction model based on these seven clinical indicators has a certain degree of discrimination,calibration and clinical validity.However,due to the heterogeneity of the incidence and population of DR,this prediction model may be of limited value in clinical guidance,and it is still necessary to continuously explore and optimize the prediction model to assist clinicians to quickly predict the risk probability of developing DR in patients with T2DM.