Cardiometabolic Assessment Of Short Stature And Neural Network Model Construction

Background:Research one:Short stature is a risk factor of cardiometabolic disease(CMD),which remains conflicting,needing research from a holistic point of view.Based on the application and popularization of medical digital technology and Artificial Intelligence(AI),Artificial Neural Network(ANN)model shows the remarkable superiority.We aimed to examine the association between short stature and CMD and further to establish a CMD risk model implementing ANN.Research two:Based on analyses of data with large sample size(research one),we found that short stature in adulthood increased the risk of CMD.Given the fact that CMD tends to occur in young adults and the scarcity for analyses of role regarding relevant etiological factors and biomarkers,we further attempted to investigate the role of growth hormone(GH)and recombinant human growth hormone(rhGH)therapy in cardiometabolic evaluation in children with short stature and establish a risk model for rhGH therapy in children with growth hormone deficiency(GHD)implementing ANN.Meanwhile,we also attempted to describe the characteristics of traditional Chinese medicine(TCM)syndrome in children with short stature and investigate its association with cardiometabolic changes.Research three:We found that the cardiometabolic changes in children with short stature were related to growth hormone,and rhGH therapy can result in cardiometabolic changes(research two).However,there is few report targeting whether those changes are time dependent.And present researches regarding rhGH treatment and cardiometabolic changes are inclusive,lacking quantitatively synthetic analyses.To shed the light on this yield,we further prepared a meta-analysis to explore the effects of rhGH therapy on cardiometabolic changes in children with.GHD and detect possible reasons for various results in previous study,especially the effect of length of intervention.Methods:Research one:The data was collected from the China National Health Survey,including eligible information of 5905 adults.CMD is diagnosed when a person has one or more cardiac and metabolic disorders,including hypertension,type 2 diabetes,hyperlipemia,stroke,and cardiovascular disease.Participants who were free of CMD formed the control group,and CMD patients were classified into two case groups based on the number of above disorders carried.Statistical analyses were done using the Anaconda(version 3-5.2.0 with Python version 3.6.5)and Stata software(version 14.0).Bi-directional stepwise method and all-subsets regression were employed to select contributing factors and determine the optimal model.Ordinal Logistic regression analyses were performed to examine the association of contributing factors with CMD risk.Additionally,restricted cubic splines were used to illustrate the risk distribution of CMD across whole height interval.Subgroup analyses and mediation analyses were further implemented,targeting significant confounding factors attributing to height prediction.Finally,a model for predicting the risk of CMD was created implementing ANN according to above analyses.Research two:A total of 400 prepubertal children with short stature were enrolled in the retrospective analyses.Multiple linear regression analyses were performed to assess the association between growth hormone and cardiometabolic changes Meanwhile,we investigated its association with TCM syndrome.A total of 150 prepubertal children with short stature were enrolled in the prospective analyses.Cox proportional-hazards model was employed to evaluate the association between rhGH therapy and cardiometabolic changes,and the increase or decrease of biomarkers were defined as increased or decreased above 10%from baseline.All analyses were further performed subsidiary analyses according to age and sex.Finally,an ANN model for predicting the risk of rhGH was created based on above analyses.Research three:Two investigators independently completed literature search,quality assessment,and data extraction.Random-effects model was adopted irrespective of between-trial heterogeneity.Both subsidiary analyses and meta-regression analyses were done to seek potential causes of between-trial heterogeneity.Cumulative analyses and influential analyses were used to inspect the stability of the results.Results:Research one:1.The baseline information showed that a total of 1151(19.5%)adults suffered CMD,and there were 895(15.1%)patients with 1 disorder and 256(4.4%)patients with?2 disorders.Compared to adults free of CMD,the body height of those suffered CMD was significantly shorter and the proportion of short stature was significantly higher(P<0.05).2.Using the bi-directional stepwise method and all-subsets regression,5 factors,including height,age,BMI(body mass index),family histories of CMD,and exercise,were found to be in significant association with CMD.3.The ordinal Logistic regression analyses were used after adjusting for confounding factors except exercise and education.Body height was analyzed on both continuous and categorical scales.Per 10 cm increment in body height was associated with 36%reduced risk of CMD(OR=0.64,95%CI:0.57 to 0.73,P<0.001).Relative to normal stature,short stature was associated with a statistically increased risk of having CMD(OR=1.42,95%CI:1.21 to 1.66,P<0.001).There was a reduced tendency towards CMD with the increase of body height under the restricted cubic spline regression model.Subgroup analyses showed that significance was only noted in subgroups with ?3 times of exercise per week and with education of secondary school degree or below.Further mediation test indicated that the proportion of total impact conferred by exercise was estimated to be 65.4%.By contrast,the proportion of total effect conferred by education was estimated to be 24.6%.5.The ANN model for predicting the risk of CMD was created with accuracy of 82.5%and 81.2%±1.25%in 10-fold cross-validation.Research two:1.The baseline information shows that LDL,LDL/HDL and TC/HDL were significantly higher in children with GHD than those with ISS(P<0.05).Though distribution of TCM syndrome was not significantly different between children with GHD and ISS(P>0.05)2.Retrospective analyses revealed that peak GH was inversely related to TG,LDL,LDL/HDL,AIP(atherogenic index of plasma)and TyG(triglyceride glucose),and positively related to HDL(P<0.05).Subsidiary analyses indicated that those relations were more significant in children with GHD,especially in school-aged and male children with GHD.Furthermore,syndrome of Liver-Qi stagnation in TCM syndrome were related to higher level of TG and AIP(P<0.05).3.Prospective analyses showed that increment of rhGH(per increment of 0.01IU/kg)were protectively related to HDL decrease(HR=0.92,95%CI:0.85 to 0.99,P=0.028)and AIP increase(HR=0.93,95%CI:0.87 to 0.99,P=0.049),but raised the risk of Hbalc increase(HR=1.14,95%CI:1.01 to 1.29,P=0.043).The results were consistent in subgroup analyses for children with GHD,while inconsistent in that of ISS(P>0.05).Further subgroup analyses for children with GHD revealed that increment of rhGH(per increment of 0.01IU/kg)were protectively related to HDL decrease(school-aged children HR=0.89,95%CI:0.79 to 0.99,P=0.041;male children HR=0.89,95%CI:0.81 to 0.99,P=0.032)and AIP increase(school-aged children HR=0.89,95%CI:0.79 to 0.99,P=0.028;male children HR=0.89,95%CI:0.81 to 0.97,P=0.009)in school-aged or male children,but raised the risk of Hbalc increase(preschool children HR=1.24,95%CI:1.03 to 1.48,P=0.022;female children HR=1.25,95%CI:1.09 to 1.56,P=0.050)in preschool and female children.4.The effect of height improvement in children with GHD was inversely related to age,bone age,height(SD),BMI and height of the pituitary gland at initial therapy.Combined with the results of prospective analyses,increment of rhGH was positive for improvement of height,HDL and AIP,but was negative for Hbalc.5.The ANN model for predicting the risk of rhGH therapy was created with accuracy of 80.4%and 79.2%±2.21%in 10-fold cross-validation.Research three:1.A total of 16 publications were qualified,including 16 independent clinical trials and 1319 children in the final analysis.2.Data from eight cardiometabolic markers were summarized.Overall analyses showed that TG was significantly decreased after rhGH therapy(WMD:-0.20 mmol/L;95%CI:-0.30 to-0.10;P<0.001),and HDL was significantly increased(WMD:0.29 mmol/L;95%CI:0.24 to 0.33;P<0.001).Marginal increase was noted for LDL(WMD:-0.22 mmol/L;95%CI:-0.47 to 0.22;P=0.092).3.Subsidiary analyses revealed that TG was increased in the early intervention(less than 12 months),and it was decreased afterwards.4.Meta-regression analyses revealed that FPG trended towards less significantly with the increase of total sample size,and progression exhibited a declining trend.5.Cumulative analyses and influential analyses indicated that three was no significant impact from first published trial on subsequently published trials and the impact of individual trials on overall effect-size estimates was not significant.Conclusions:Research one:Short stature increase the risk of CMD,but the association is confounded by exercise and education.ANN model for predicting the risk of CMD is more accurate than Logistic model.Given the fact that CMD tends to occur in young adults,more investigations targeting children with short stature are necessary.Research two:Growth hormone play a significant role in cardiometabolic evaluation in children with short stature and growth hormone is significantly related with cardiometabolic changes,whose performance is various in children with different age and sex.The TCM syndrome for prepubertal children with GHD is typical as syndrome of weakness of the Spleen and Stomach and its distribution was not significantly different between children with GHD and ISS.While syndrome of Liver-Qi stagnation in TCM syndrome is related to higher level of TG and AIP in children with short stature.rhGH therapy can improve lipid metabolism,but raise the risk of Hbalc increase.The model for predicting the risk of rhGH therapy was created implementing ANN,providing supports not only for clinical strategy,but also for personalized treatment of children with short stature.The results are helpful for judgment before treatment and reduce its potential risk,improving the quality of treatment.Research three:There was a time dependent effect of rhGH therapy in cardiometabolic changes in children with GHD,and this role might be the reason for various results in previous study.Besides,our results support the obviously favorable role of rhGH therapy in lipid metabolism in children with GHD.For practical purpose,it is still necessary to monitor cardiometabolic changes when administrating rhGH to children with GHD.