| OBJECTIVEOnce diabetic kidney disease(DKD)develops from the early stage(stage Ⅲ)into the clinical stage(stage IV),it becomes unstoppable and the percentage of ending up in end-stage renal disease(ESRD)is inevitably and relatively increased.Therefore,the assessment of patients in early stage of DKD to predict their potential to further develop into clinical stage of DKD is crucial in preventing and treating DKD.Taking consideration of the traditional Chinese medical theory of TCM syndromes and signs theory and laboratory tests,this study is conducted to reveal the hidden factors that are the keys to predicting the advancement of DKD,and then compare the different predicting models formed via logistic regression,decision tree,artificial neural nets.METHOD1.The study to identify the TCM syndrome elements that are key to the advancement into clinical stage DKD:565 patients diagnosed of DKD are included and classified into early and clinical stage DKD groups,according to the UACR(urine albumin creatinine ratio),with 301 and 264 patients in each.The TCM syndromes are collected and rated by the TCM DKD syndrome rating scale.Accordingly,the syndrome distribution could be formed and compared between these two groups.Later on,after 6 months of observation,the TCM syndromes of patients in early stage DKD group that advance into the clinical stage DKD are compared with those who hasn’t.2.The study of relevant biological marks that indicate the advancement of DKD:340 patients diagnosed of DKD are included and classified into early stage DKD groups,according to the UACR(urine albumin creatinine ratio),with 277 and 113 patients in each.Cross-sectional and Prospective cohort study are both applied to the survey.Hence at time point 0,the congenital yunqi endowment,TCM syndromes and signs serum MGO,APN,VEGF,and urine LAP,PCX,Mindin protein are measured so that the distribution difference can be analyzed between the two groups.Six months later,the same biological marks and TCM terms are measured in both groups and comparisons are made between those who develop from early stage into the clinical stage DKD so that the difference of biological marks’distribution between the developed and stable patients can be seen.3.The study to identify the relevant congenital yunqi endowment differences between early stage DKD and clinical stage DKD:920 patients diagnosed of DKD are included into the study and classified into early stage and clinical stage DKD groups,according to the Mogensen staging approach,with 440 and 480 patients in each.The accurate birth date in solar calendar of every patients’are collected and each patient’s yunqi information,the suiyun,sitian,zaiquan,zhuqi,keqi of the birth year is accessible on the recent-60-year yunqi rotation table.With all the information,yunqi tonghua and yunqi jialin situation of each patient can be deducted.Therefore,the yunqi distribution is concluded and the yunqi factors with significant difference between the early and clinical stage of DKD are analyzed as well as the patients’birth date yunqi information that develop from early stage into clinical stage of DKD after six months of observation is analyzed.4.The establishment of predicting models for the advancement of early stage into clinical stage DKD:the key indicators to the advancement of DKD are concluded from the results of the first 3 steps mentioned above through Logistic regression.These indicators are then analyzed and used to establish the predicting models via Logistic regression,decision tree and artificial neural nets,which will then be evaluated by the ROC curve.RESULTS1.As to the frequency of TCM syndromes and signs,yinxu and shire are the highest in the earl-stage DKD group,while qixu and xueyu are the highest of the clinical-stage group.Between these two groups,the percentage of yinxu is significantly higher in the early-stage group(P<0.05),however,the clinical-stage group on the other hand,surpasses the early-stage group on account of the percentage of yangxu,yinyang liangxu,xuexu,xueyu,shuiting syndromes.The patients developing into the clinical-stage DKD after 6 months have a higher percentage of yangxu,xueyu,and shuiting syndromes than the stable ones in the early-stage group at observation point 0(P<0.05).However,6 months later,there’s no significant statistically meaningful change between these patients.2.Serum MGO,VEGF and urine LAP,PCX and Mindin protein from the clinical stage patients have an apparent difference from those of the other group(P<0.05),but serum APN appears no such difference.Similarly,at observation time 0,the serum MGO and urine PCX shows great difference between those who advance into clinical-stage DKD from the early-stage group and who stay stable(P<0.05).However,among those patients who develop,the level of serum VEFG,APN and urine LAP,Mindin protein don’t vary much.6 months later,a significant rise of serum MGO and urine LAP,PCX levels are observed(P<0.05),but the serum VEGF level doesn’t show much difference than observation point 0 in both the developed and stable groups from early-stage DKD group.Serum APN level stay similar in the developed group 6 months later while the stable group presents a significant rise(P<0.05).On the other hand,on account of urine Mindin protient level,the developed group shows a great difference(P<0.05),while that of the stable group doesn’t vary much between 0 and 6 months.3.All of the 920 cases of DKD appears much higher percentage of taiyang hanshui and taiyin shitu of their birth year.The early-stage group shows a higher percentage of muyun taiguo as suiyun and jueyin fengmu as guest qi of the birth year(P<0.05),compared with the clinical-stage group while the percentage of taiyin shitu sitian-taiyang hanshui zaiquan appears to be much lower than that of the clinical-stage group(P<0.05).Of all the cases in clinical-stage group,the percentage of dominate qi as taiyang hanshui of birth years has no statistically meaningful difference than that of the early-stage group(P<0.05),but its inclined to go up.Tonghua and jialin between the groups:the percentage of tuyun tongtianfu of birth years is significantly lower in the early-stage group(P<0.05);tuyun suihui,yunqi jialin xiaoni and mushenhuoqi qitian of birth years have a higher percentage in the early-stage group(P<0.05).4.Multiple logistic regression analysis shows:serum MGO,urine PCX,xueyu syndrome,yangxu syndrome are relevant to the progress from early-stage to clinical-stage DKD independently.Hence the predicting equation for DKD progress is drown:P=exp(-6.391+0.501MGO+0.740PCX+1.066Yx+2.1Xy)/1+exp(-6.391+0.501MGO+0.740PC X+1.066Yx+2.1Xy).The AUC is 0.853 and the accuracy is 90.0%,sensitivity is 92.6%and specificity is 50.5%.Among all of the three models constructed,the decision tree model shows the largest AUC(area under curve)of 0.879,while the artificial neural net is the lower,0.854.As for accuracy,the artificial neural net is the highest,92.5%,and the decision tree model is 90.0%,the lowest amongst all of the three models.On account of sensitivity,all of three is 92.6%.However,decision tree model is 62.5%,while artificial neural net and logistic regression analysis model have the lower specificity of 50.5%.CONCLUSIONS1.Yangxu,xueyu or shuishi syndromes in the early-stage DKD might be related to the deficiency of shenjing,leading to the decline of the shenyin,shenyang and shenqi transformation.Exacerbation of kidney essence deficiency is the leading cause of early diabetic kidney disease progression potential pathogenesis.2.Increase of serum MGO and urinary PCX levels suggest the possibility of early diabetic kidney disease progress to clinical proteinuria stage.3.The birth year falling into taiyin shitu sitian and is the year of ox or sheep in the clinical-stage DKD is higher than early-stage DKD,while born in a year of whose number ends with 2 and the year is lower than early-stage,may tend to stay stable in the early-stage of DKD.4.The endowment of shihan of the birth year is an indispensable factor for the occurrence of DKD.Hence the attention to hanshi is crucial when it comes to the treatment of DKD in TCM.Fengmu is one of the most important keys to treating shi,to which the taste sour is the key.5.All of the three models built to predict the progress of DKD have shown great diagnostic ability,among which the decision tree model shows a slightly better AUC,indicating that decision tree predicting model has a better diagnostic ability,all of three show a high sensitivity but low specificity. |
