The Study On The Accurate Prediction Of Physical Activity And Dose-response Effect On Improvement Of Insulin Resistance And Risk In Obese Adolescents

PurposesThe purpose of this study was determined the characteristics of physical activity energy expenditure,and contructed assessment models of physical activity in obese insulin resistance(IR)and non-insulin resistance(NIR)adolescents.To provided accurate assessment model of physical activity in obese IR and NIR adolescents,explored the dose-response relationship between physical activity and IR or IR-risk improvement,and maximize the health effects of physical activity in prevented and improved IR in obese adolescents.MethodsA total of 249 obese IR and NIR adolescents aged 13-17 years were recruited to performed rest energy expenditure tests and 120 obese adolescents performed 3-7km/h exercise energy expenditure test,while Acti Graph GT3X+triaxial accelerometers and Polar heart rate moniters were used to collect accelerometer counts(VM3)and heart rate data.General linear models(GLM),linear mixed models(LMM)and back propagation artificial neural networks(BP-ANNs)models were constructed for metabolic equivalents(METs)and physical activity energy expenditure(PAEE)of obese IR and NIR adolescents,respectively.Before modeling,variables were screened by LASSO regression to avoid multicollinearity.Root mean square error(RMSE),mean error(Bias),mean absolute percentage error(MAPE),and intraclass correlation coefficient(ICC)were used to assessed the accuracy of the models.The IR-risk predicted model of obese adolescents was constructed by body morphology,body composition,and blood lipids index.Physical activity of 90 obese IR or NIR adolescents were assessed by physical activity assessment models.The isotemporal substitution model(ISM)was used to investigate the effectiveness of different intensity physical activity in IR and IR-risk improvment in obese adolescents,the dose-response relationship between physical activity and IR or IR-risk improvement was investigated by restricted cubic spline(RCS).Results1.The rest energy expenditure(1.73±0.33 kcal/min vs 1.50±0.27 kcal/min,P<0.001)and 1MET(4.28 m L/kg/min vs 3.93 m L/kg/min,P<0.001)were significantly higher in obese IR adolescents than NIR adolescents.During exercise test,PAEE was significantly higher in obese IR group(P<0.05),while the proportion of fat energy supply was significantly lower(P<0.05).After adjusted for body weight,there was no statistical difference in PAEE for obese IR and NIR adolescents between different gender.2.By LASSO regression,age,height,waist circumference(WC),body fat(BF),body weight(BW),VM3 were selected to construct GLM,LMM and BP-ANNs model for METs in obese IR adolescents.GLM:METs=(6.23E-4)×VM3(count/min)-0.031×BW(kg)+0.023×Height(cm)-0.225;LMM:METs=1.710-(1.28E-4)×BW(kg)~2+(1.456E-5)×VM3(count/min);And the BP-ANNs model was modelled for a total of7875 iterations.The Bias,RMSE,MAPE and ICC of BP-ANNs were 0,0.49 and8.71%and 0.97,respectively,which were better than GLM and LMM.By LASSO regression,age,height,WC,BW and VM3 were selected to construct GLM,LMM and BP-ANNs model of METs in obese NIR adolescents.GLM:METs=(6.39E-4)×VM3(counts/min)-0.023×BW(kg)+3.207;LMM:METs=(-6.864E-6)×BW(kg)×VM3(counts/min)+0.00124×VM3(counts/min)+0.0382×Height(cm)-4.896;And the BP-ANNs model was modelled for a total of 4794 iterations.The Bias,RMSE,MAPE and ICC of BP-ANNs were 0,0.66 and 9.77%and 0.97,respectively,which were better than GLM and LMM.3.By LASSO regression,age,height,WC,FFM,BW,VM3 were selected to construct GLM,LMM and BP-ANNs model for PAEE in obese IR adolescents.GLM:PAEE=0.001×VM3(counts/min)+0.037×BW(kg)-3.011;LMM:PAEE=2.247-0.035×BW(kg)+(1.456E-5)×VM3(counts/min)×BW(kg);And the BP-ANNs model was modelled for a total of 9043 iterations.The Bias,RMSE,MAPE and ICC of BP-ANNs were 0,0.60,13.35%and 0.97,respectively,which were better than GLM and LMM.By LASSO regression,age,height,HC,BF,BW and VM3 were selected to construct GLM,LMM and BP-ANNs model of PAEE in obese NIR adolescents.GLM:PAEE=(1.028E-3)×VM3(counts/min)+0.0305×BW(kg)+0.0358×Height(cm)-8.113;LMM:PAEE=-3.2245+(1.078E-3)×VM3(counts/min)+(2.39E-4)×Height(cm)×BW(kg);And the BP-ANNs model was modelled for a total of 5251 iterations.The Bias,RMSE,MAPE and ICC of BP-ANNs were 0.02,0.94,9.41%and 0.97,respectively,which were better than GLM and LMM.4.After univariate and multivariate logistic regression analysis,IR risk factors were body mass index(BMI),WC,body fat percentage(PBF),total cholesterol(TC),triglyceride(TG)and low-density lipoprotein cholesterol(LDL-c).The AUC of model was 0.828(95%CI:0.784-0.872,P<0.001).The internal verification showed that the C-Index of training set was 0.810,the absolute error was 0.021,and the Brier was 0.146.The decision clinical analysis showed that the threshold probability was0.1.5.10 min of moderate-intensity physical activity(MPA)replaced sedentary time(SB)(β=-0.456,P=0.012)or low-intensity physical activity(LPA)(β=-0.444,P=0.018),respectively,and 10 min of vigorous physical activity(VPA)replaced SB(β=-0.560,P=0.005)or LPA(β=-0.548,P=0.005),respectively,were effective in improving HOMA-IR in obese IR adolescents.6.10 min MPA replaced SB(β=-0.035,P=0.032)or LPA(β=-0.039,P=0.025),respectively,and 10 min VPA replaced SB(β=-0.036,P=0.025)or LPA(β=-0.040,P=0.033),respectively,were effective in reduced IR-risk in obese NIR adolescents.7.The optimal MVPA dose for reduced IR-risk in obese NIR adolescents was53.0 min/day.When MVPA≤53.0 min/day,the IR-risk improved by 0.012(95%CI:0.004,0.020)for 1 min increased;the optimal physical activity dose was 16.6 MET-h/day,and when physical activity≤16.6MET-h/day,the IR-risk improved by 0.023(95%CI:0.007,0.040)for 1 MET-h/day increased;the optimal PAEE dose was1643.9 kcal/day,and when PAEE≤1643.9 kcal/day,the IR-risk by 0.020(95%CI:0.004,0.020)for each 100 kcal/day improved.8.The optimal MVPA dose for HOMA-IR improved in obese IR adolescents was48.9min/day.When MVPA≤48.9 min/day,the HOMA-IR improved by 0.044(95%CI:0.004,0.085)for 1 min increased;the optimal physical activity was 14.5 MET-h/day,and when physical activity≤14.5MET-h/day,the HOMA-IR improved by0.220(95%CI:0.081,0.360)for 1 MET-h/day increased;the optimal PAEE dose was1296.3 kcal/day,and when PAEE≤1296.3 kcal/day,the HOMA-IR by 0.260(95%CI:0.069,0.445)for each 100 kcal/day improved.Conclusion1.The difference of REE and PAEE in obese IR and NIR adolescents were significant,and the PAEE of obese IR adolescents was significantly higher than obese NIR adolescents,and the proportion of fat energy supply was also significantly lower.2.The BP-ANNs for obese IR and NIR adolescents in physical activity assessment can effectively improved the prediction accuracy and reduced the errors in LMM and GLM predictions;3.Compared with LPA,MPA or VPA were the best physical activity intensity to improved IR and IR-risk in obese IR and NIR adolescents;4.The optimal MVPA for reducing IR-risk in obese NIR adolescents was 53min/day or the optimal physical activity level was 17 MET-h/day or the optimal PAEE was 1644 kcal/day.5.The optimal MVPA for improving IR in obese IR adolescents was 49 min/day or the optimal physical activity was 15 MET-h/day or the optimal PAEE was 1297kcal/day.