Use of compositional data analysis to show estimated changes in cardiometabolic health by reallocating time to light-intensity physical activity in older adults.

Journal:Sports Medicine

Authors: Cormac Powell*, Leonard D. Browne, Brian P. Carson, Kieran P. Dowd, Ivan J. Perry, Patricia M. Kearney, Janas M. Harrington and Alan E. Donnelly

Instrcuctions

Please do the following:

Anthropometric Measures

BMI

The relationship between BMI and daily activity behaviours as modelled by Compositional multiple linear regression with physical activity expressed as isometric log-ratio (ilr) coordinates are shown in Supplementary Table 4. This model included the ilr coordinates of activity behaviours and terms for age, sex, smoking status, alcohol consumption, work status, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditions along with diabetic,lipid and BP medications. Five sets of ilr-coordinate systems were constructed, each time rotating the sequence of activity behaviours, so that each behaviour was iteratively represented as the first compositional part.

Supplementary Table 4.0 Multiple linear regression analyses of the relationship between first ilr coordinates and BMI
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) -2.820 1.825 -1.546 0.124
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) 3.068 1.245 2.464 0.014
ilr1 / ln (Standing: geometric mean of remaining behaviours) 3.410 1.099 3.102 0.002
ilr1 / ln (LIPA: geometric mean of remaining behaviours) -3.331 0.968 -3.442 0.001
ilr1 / ln (MVPA: geometric mean of remaining behaviours) -0.327 0.365 -0.897 0.370
Note:
ilr: Isometric log-ratio; BMI: Body mass index transformed, Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption, work status,DASH score, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 260
§ Adjusted R-squared:
¶ 0.14 
## Anova Table (Type II tests)
## 
## Response: y
##                         Sum Sq  Df F value    Pr(>F)    
## ilr_comps                483.2   4  5.5632 0.0002667 ***
## dataf$Age                 36.7   1  1.6895 0.1948982    
## dataf$Sex_Code            58.2   1  2.6825 0.1027544    
## dataf$Smoking_Code       310.1   1 14.2796 0.0001985 ***
## dataf$Consume_Alco_Code   48.6   1  2.2399 0.1357932    
## dataf$Work_Status_Code    86.9   1  4.0029 0.0465380 *  
## dataf$Hypertension_Code   27.4   1  1.2614 0.2624906    
## dataf$BPMed_Code           5.3   1  0.2437 0.6219982    
## dataf$High_Chol_Code       0.4   1  0.0181 0.8931522    
## dataf$CholMed_Code         7.8   1  0.3601 0.5489853    
## dataf$Diabetes_Code        3.4   1  0.1549 0.6942054    
## dataf$DiabetesMed_Code    12.5   1  0.5762 0.4485344    
## dataf$Heart_Cond_Code      9.5   1  0.4383 0.5085727    
## dataf$dash                35.0   1  1.6108 0.2055929    
## Residuals               5254.9 242                      
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Supplementary Figure 1. The relationship between daily behaviours and BMI as estimated by compositional linear regression models.Difference in minutes modelled around the population mean composition.

Supplementary Figure 2. The relationship between daily behaviours and BMI as estimated by compositional linear regression models.Predictions based on realloacting each actvity type for another.

% Body Fat

Supplementary Table 5.0 Multiple linear regression analyses of the relationship between first ilr coordinates and Body Fat %
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) 1.904 2.221 0.857 0.392
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) 2.156 1.513 1.425 0.155
ilr1 / ln (Standing: geometric mean of remaining behaviours) 2.857 1.349 2.117 0.035
ilr1 / ln (LIPA: geometric mean of remaining behaviours) -6.350 1.201 -5.285 0.000
ilr1 / ln (MVPA: geometric mean of remaining behaviours) -0.567 0.467 -1.214 0.226
Note:
ilr: Isometric log-ratio; Body_Fat: Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption, work status,DASH score, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 253
§ Adjusted R-squared:
¶ 0.52 
## Anova Table (Type II tests)
## 
## Response: y
##                         Sum Sq  Df  F value    Pr(>F)    
## ilr_comps               1246.7   4   9.8927 2.052e-07 ***
## dataf$Age                262.1   1   8.3182  0.004289 ** 
## dataf$Sex_Code          4542.1   1 144.1633 < 2.2e-16 ***
## dataf$Smoking_Code        82.4   1   2.6161  0.107126    
## dataf$Consume_Alco_Code   44.0   1   1.3961  0.238568    
## dataf$Work_Status_Code   303.5   1   9.6329  0.002146 ** 
## dataf$Hypertension_Code    0.2   1   0.0056  0.940438    
## dataf$BPMed_Code           6.9   1   0.2185  0.640654    
## dataf$High_Chol_Code       0.6   1   0.0186  0.891678    
## dataf$CholMed_Code         3.7   1   0.1180  0.731562    
## dataf$Diabetes_Code        1.4   1   0.0453  0.831711    
## dataf$DiabetesMed_Code     2.4   1   0.0769  0.781837    
## dataf$Heart_Cond_Code      3.9   1   0.1250  0.724042    
## dataf$dash                 2.8   1   0.0894  0.765161    
## Residuals               7404.1 235                       
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Supplementary Figure 3. The relationship between daily behaviours and Body Fat % as estimated by compositional linear regression models. Difference in minutes modelled around the population mean composition.

Supplementary Figure 4. The relationship between daily behaviours and Body_Fat as estimated by compositional linear regression models.Predictions based on realloacting each actvity type for another.

Fat Free Mass (kg)

Supplementary Table 6.0 Multiple linear regression analyses of the relationship between first ilr coordinates and Fat_Free_Mass
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) -1.497 2.327 -0.643 0.521
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) 2.624 1.585 1.655 0.099
ilr1 / ln (Standing: geometric mean of remaining behaviours) 0.413 1.414 0.292 0.770
ilr1 / ln (LIPA: geometric mean of remaining behaviours) -1.309 1.259 -1.040 0.299
ilr1 / ln (MVPA: geometric mean of remaining behaviours) -0.231 0.490 -0.472 0.638
Note:
ilr: Isometric log-ratio; Fat Free Mass: , Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption, work status,DASH score, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 253
§ Adjusted R-squared:
¶ 0.73 
## Anova Table (Type II tests)
## 
## Response: y
##                          Sum Sq  Df  F value    Pr(>F)    
## ilr_comps                 183.7   4   1.3276 0.2603705    
## dataf$Age                 225.5   1   6.5165 0.0113212 *  
## dataf$Sex_Code          15511.3   1 448.3334 < 2.2e-16 ***
## dataf$Smoking_Code        458.5   1  13.2532 0.0003347 ***
## dataf$Consume_Alco_Code    51.4   1   1.4861 0.2240521    
## dataf$Work_Status_Code     63.7   1   1.8416 0.1760633    
## dataf$Hypertension_Code     0.8   1   0.0230 0.8795518    
## dataf$BPMed_Code           45.8   1   1.3232 0.2511931    
## dataf$High_Chol_Code       83.8   1   2.4222 0.1209752    
## dataf$CholMed_Code          1.7   1   0.0496 0.8239837    
## dataf$Diabetes_Code         3.0   1   0.0858 0.7698945    
## dataf$DiabetesMed_Code      0.1   1   0.0025 0.9602786    
## dataf$Heart_Cond_Code      18.6   1   0.5374 0.4642447    
## dataf$dash                 18.2   1   0.5252 0.4693351    
## Residuals                8130.5 235                       
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Supplementary Figure 5. The relationship between daily behaviours and Fat_Free_Mass as estimated by compositional linear regression models. Difference in minutes modelled around the population mean composition.

Supplementary Figure 6. The relationship between daily behaviours and Fat_Free_Mass as estimated by compositional linear regression models.Predictions based on realloacting each actvity type for another.

Fat Mass (kg)

Supplementary Table 7.0 Multiple linear regression analyses of the relationship between first ilr coordinates and Fat_Mass
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) 0.626 3.128 0.200 0.842
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) 4.024 2.131 1.888 0.060
ilr1 / ln (Standing: geometric mean of remaining behaviours) 4.237 1.901 2.229 0.027
ilr1 / ln (LIPA: geometric mean of remaining behaviours) -8.089 1.692 -4.779 0.000
ilr1 / ln (MVPA: geometric mean of remaining behaviours) -0.797 0.658 -1.211 0.227
Note:
ilr: Isometric log-ratio; Fat Free Mass: , Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption, work status,DASH score, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 253
§ Adjusted R-squared:
¶ 0.18 
## Anova Table (Type II tests)
## 
## Response: y
##                          Sum Sq  Df F value    Pr(>F)    
## ilr_comps                2171.8   4  8.6843 1.488e-06 ***
## dataf$Age                 604.2   1  9.6632  0.002112 ** 
## dataf$Sex_Code            432.1   1  6.9106  0.009134 ** 
## dataf$Smoking_Code        228.2   1  3.6506  0.057267 .  
## dataf$Consume_Alco_Code   110.3   1  1.7649  0.185304    
## dataf$Work_Status_Code    226.2   1  3.6177  0.058392 .  
## dataf$Hypertension_Code     0.2   1  0.0032  0.954919    
## dataf$BPMed_Code           24.2   1  0.3874  0.534284    
## dataf$High_Chol_Code       17.9   1  0.2862  0.593151    
## dataf$CholMed_Code         18.0   1  0.2875  0.592311    
## dataf$Diabetes_Code         5.2   1  0.0832  0.773286    
## dataf$DiabetesMed_Code      1.1   1  0.0178  0.894059    
## dataf$Heart_Cond_Code       4.5   1  0.0722  0.788453    
## dataf$dash                 27.6   1  0.4407  0.507414    
## Residuals               14692.6 235                      
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Supplementary Figure 5. The relationship between daily behaviours and Fat_Mass as estimated by compositional linear regression models. Difference in minutes modelled around the population mean composition.

Supplementary Figure 6. The relationship between daily behaviours and Fat_Mass as estimated by compositional linear regression models.Predictions based on realloacting each actvity type for another.

Lipids

Total cholesterol

Supplementary Table 8.0 Multiple linear regression analyses of the relationship between first ilr coordinates and Total_Cholesterol
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) 0.582 0.439 1.328 0.186
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) -0.197 0.301 -0.654 0.514
ilr1 / ln (Standing: geometric mean of remaining behaviours) -0.168 0.260 -0.647 0.518
ilr1 / ln (LIPA: geometric mean of remaining behaviours) -0.225 0.242 -0.929 0.354
ilr1 / ln (MVPA: geometric mean of remaining behaviours) 0.008 0.089 0.093 0.926
Note:
ilr: Isometric log-ratio; Total_Cholesterol: Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption,fat mass,DASH score, work status, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 246
§ Adjusted R-squared:
¶ 0.01 
## Anova Table (Type II tests)
## 
## Response: y
##                          Sum Sq  Df F value  Pr(>F)  
## ilr_comps                 2.994   4  0.6677 0.61504  
## dataf$Age                 0.385   1  0.3434 0.55843  
## dataf$Sex_Code            3.114   1  2.7777 0.09696 .
## dataf$Smoking_Code        0.059   1  0.0530 0.81812  
## dataf$Consume_Alco_Code   0.304   1  0.2714 0.60288  
## dataf$Work_Status_Code    1.379   1  1.2301 0.26856  
## dataf$Hypertension_Code   0.068   1  0.0608 0.80550  
## dataf$BPMed_Code          0.223   1  0.1989 0.65604  
## dataf$High_Chol_Code      0.177   1  0.1578 0.69158  
## dataf$CholMed_Code        2.369   1  2.1126 0.14748  
## dataf$Diabetes_Code       0.934   1  0.8330 0.36238  
## dataf$DiabetesMed_Code    1.749   1  1.5601 0.21294  
## dataf$Heart_Cond_Code     0.553   1  0.4929 0.48336  
## dataf$Fat_Mass            4.673   1  4.1682 0.04235 *
## dataf$dash                0.054   1  0.0484 0.82611  
## Residuals               254.507 227                  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Supplementary Figure 7. The relationship between daily behaviours and Total_Cholesterol as estimated by compositional linear regression models. Difference in minutes modelled around the population mean composition.

Supplementary Figure 8. The relationship between daily behaviours and Total_Cholesterol as estimated by compositional linear regression models.Predictions based on realloacting each actvity type for another.

HDL-C

Supplementary Table 9.0 Multiple linear regression analyses of the relationship between first ilr coordinates and HDL
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) 0.164 0.155 1.062 0.289
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) -0.129 0.106 -1.211 0.227
ilr1 / ln (Standing: geometric mean of remaining behaviours) 0.036 0.092 0.388 0.698
ilr1 / ln (LIPA: geometric mean of remaining behaviours) -0.043 0.086 -0.504 0.615
ilr1 / ln (MVPA: geometric mean of remaining behaviours) -0.028 0.031 -0.898 0.370
Note:
ilr: Isometric log-ratio; HDL: , Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption,fat mass,DASH score, work status, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 246
§ Adjusted R-squared:
¶ 0.03 
## Anova Table (Type II tests)
## 
## Response: y
##                         Sum Sq  Df F value   Pr(>F)   
## ilr_comps                0.384   4  0.6875 0.601301   
## dataf$Age                0.062   1  0.4430 0.506337   
## dataf$Sex_Code           0.014   1  0.0996 0.752608   
## dataf$Smoking_Code       0.297   1  2.1296 0.145860   
## dataf$Consume_Alco_Code  0.365   1  2.6132 0.107365   
## dataf$Work_Status_Code   0.001   1  0.0040 0.949625   
## dataf$Hypertension_Code  0.264   1  1.8873 0.170867   
## dataf$BPMed_Code         0.086   1  0.6159 0.433385   
## dataf$High_Chol_Code     0.485   1  3.4740 0.063631 . 
## dataf$CholMed_Code       1.476   1 10.5750 0.001321 **
## dataf$Diabetes_Code      0.130   1  0.9318 0.335418   
## dataf$DiabetesMed_Code   0.171   1  1.2257 0.269418   
## dataf$Heart_Cond_Code    0.284   1  2.0371 0.154877   
## dataf$Fat_Mass           0.024   1  0.1693 0.681113   
## dataf$dash               0.012   1  0.0847 0.771277   
## Residuals               31.694 227                    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Supplementary Figure 9. The relationship between daily behaviours and HDL as estimated by compositional linear regression models.Difference in minutes modelled around the population mean composition.

Supplementary Figure 10. The relationship between daily behaviours and HDL as estimated by compositional linear regression models.Predictions based on realloacting each actvity type for another.

LDL-C

Supplementary Table 10.0 Multiple linear regression analyses of the relationship between first ilr coordinates and LDL
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) 0.265 0.414 0.641 0.522
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) -0.119 0.284 -0.418 0.677
ilr1 / ln (Standing: geometric mean of remaining behaviours) -0.017 0.247 -0.067 0.947
ilr1 / ln (LIPA: geometric mean of remaining behaviours) -0.201 0.230 -0.873 0.383
ilr1 / ln (MVPA: geometric mean of remaining behaviours) 0.071 0.086 0.824 0.411
Note:
ilr: Isometric log-ratio; LDL: , Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption,fat mass,DASh score, work status, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 235
§ Adjusted R-squared:
¶ -0.03 
## Anova Table (Type II tests)
## 
## Response: y
##                          Sum Sq  Df F value Pr(>F)
## ilr_comps                 1.606   4  0.4156 0.7973
## dataf$Age                 0.024   1  0.0247 0.8752
## dataf$Sex_Code            2.112   1  2.1868 0.1407
## dataf$Smoking_Code        0.205   1  0.2118 0.6458
## dataf$Consume_Alco_Code   0.392   1  0.4063 0.5245
## dataf$Work_Status_Code    0.234   1  0.2422 0.6231
## dataf$Hypertension_Code   0.042   1  0.0438 0.8345
## dataf$BPMed_Code          0.018   1  0.0185 0.8918
## dataf$High_Chol_Code      0.244   1  0.2525 0.6158
## dataf$CholMed_Code        0.314   1  0.3254 0.5690
## dataf$Diabetes_Code       0.017   1  0.0181 0.8932
## dataf$DiabetesMed_Code    0.115   1  0.1193 0.7302
## dataf$Heart_Cond_Code     1.003   1  1.0388 0.3092
## dataf$Fat_Mass            2.454   1  2.5410 0.1124
## dataf$dash                0.113   1  0.1174 0.7322
## Residuals               208.646 216

Supplementary Figure 11. The relationship between daily behaviours and LDL as estimated by compositional linear regression models.Difference in minutes modelled around the population mean composition.

Supplementary Figure 12. The relationship between daily behaviours and LDL as estimated by compositional linear regression models.Predictions based on realloacting each actvity type for another.

VLDL-C

Supplementary Table 11.0 Multiple linear regression analyses of the relationship between first ilr coordinates and VLDL
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) -0.072 0.116 -0.616 0.539
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) 0.106 0.080 1.326 0.186
ilr1 / ln (Standing: geometric mean of remaining behaviours) -0.056 0.072 -0.776 0.439
ilr1 / ln (LIPA: geometric mean of remaining behaviours) -0.002 0.066 -0.027 0.979
ilr1 / ln (MVPA: geometric mean of remaining behaviours) 0.023 0.026 0.901 0.369
Note:
ilr: Isometric log-ratio; VLDL: , Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption,fat mass,DASH score, work status, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 209
§ Adjusted R-squared:
¶ 0 
## Anova Table (Type II tests)
## 
## Response: y
##                          Sum Sq  Df F value  Pr(>F)  
## ilr_comps                0.2608   4  0.9103 0.45901  
## dataf$Age                0.0521   1  0.7269 0.39495  
## dataf$Sex_Code           0.0012   1  0.0171 0.89624  
## dataf$Smoking_Code       0.3292   1  4.5954 0.03333 *
## dataf$Consume_Alco_Code  0.0217   1  0.3035 0.58235  
## dataf$Work_Status_Code   0.2266   1  3.1634 0.07691 .
## dataf$Hypertension_Code  0.0065   1  0.0904 0.76397  
## dataf$BPMed_Code         0.0005   1  0.0072 0.93242  
## dataf$High_Chol_Code     0.0818   1  1.1422 0.28654  
## dataf$CholMed_Code       0.0676   1  0.9434 0.33265  
## dataf$Diabetes_Code      0.3187   1  4.4494 0.03622 *
## dataf$DiabetesMed_Code   0.4004   1  5.5901 0.01907 *
## dataf$Heart_Cond_Code    0.0258   1  0.3598 0.54931  
## dataf$Fat_Mass           0.0028   1  0.0387 0.84427  
## dataf$dash               0.0001   1  0.0010 0.97512  
## Residuals               13.6101 190                  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Supplementary Figure 13. The relationship between daily behaviours and VLDL as estimated by compositional linear regression models. Difference in minutes modelled around the population mean composition.

Supplementary Figure 14. The relationship between daily behaviours and VLDL as estimated by compositional linear regression models.Predictions based on realloacting each actvity type for another.

Triglycerides

Supplementary Table 12.0 Multiple linear regression analyses of the relationship between first ilr coordinates and Triglycerides
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) 0.044 0.301 0.146 0.884
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) 0.140 0.207 0.680 0.497
ilr1 / ln (Standing: geometric mean of remaining behaviours) -0.354 0.178 -1.986 0.048
ilr1 / ln (LIPA: geometric mean of remaining behaviours) 0.199 0.166 1.196 0.233
ilr1 / ln (MVPA: geometric mean of remaining behaviours) -0.029 0.061 -0.475 0.635
Note:
ilr: Isometric log-ratio; Triglycerides: , Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption,fat mass,DASH score, work status, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 246
§ Adjusted R-squared:
¶ 0.02 
## Anova Table (Type II tests)
## 
## Response: y
##                          Sum Sq  Df F value  Pr(>F)  
## ilr_comps                 2.893   4  1.3712 0.24475  
## dataf$Age                 0.013   1  0.0253 0.87374  
## dataf$Sex_Code            0.000   1  0.0009 0.97608  
## dataf$Smoking_Code        2.297   1  4.3559 0.03800 *
## dataf$Consume_Alco_Code   1.118   1  2.1197 0.14680  
## dataf$Work_Status_Code    1.614   1  3.0603 0.08158 .
## dataf$Hypertension_Code   2.207   1  4.1843 0.04195 *
## dataf$BPMed_Code          0.539   1  1.0223 0.31306  
## dataf$High_Chol_Code      1.810   1  3.4311 0.06528 .
## dataf$CholMed_Code        0.706   1  1.3388 0.24847  
## dataf$Diabetes_Code       0.630   1  1.1946 0.27557  
## dataf$DiabetesMed_Code    0.850   1  1.6126 0.20542  
## dataf$Heart_Cond_Code     0.754   1  1.4292 0.23315  
## dataf$Fat_Mass            0.044   1  0.0825 0.77417  
## dataf$dash                0.006   1  0.0122 0.91228  
## Residuals               119.719 227                  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Supplementary Figure 15. The relationship between daily behaviours and Triglycerides as estimated by compositional linear regression models. Difference in minutes modelled around the population mean composition.

Supplementary Figure 16. The relationship between daily behaviours and Triglycerides as estimated by compositional linear regression models.Predictions reallocating compositional values pairwise amungsnst the components model for each activity component

Diabetic markers

Fasting Glucose

Supplementary Table 13.0 Multiple linear regression analyses of the relationship between first ilr coordinates and Glucose
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) 0.240 0.535 0.449 0.654
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) -0.238 0.367 -0.649 0.517
ilr1 / ln (Standing: geometric mean of remaining behaviours) -0.203 0.329 -0.617 0.538
ilr1 / ln (LIPA: geometric mean of remaining behaviours) 0.250 0.303 0.825 0.410
ilr1 / ln (MVPA: geometric mean of remaining behaviours) -0.049 0.113 -0.438 0.662
Note:
ilr: Isometric log-ratio; Glucose: , Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption,fat mass,DASh score, work status, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 253
§ Adjusted R-squared:
¶ -0.01 
## Anova Table (Type II tests)
## 
## Response: y
##                         Sum Sq  Df F value Pr(>F)
## ilr_comps                 2.06   4  0.2808 0.8902
## dataf$Age                 2.22   1  1.2128 0.2719
## dataf$Sex_Code            1.65   1  0.9028 0.3430
## dataf$Smoking_Code        1.48   1  0.8064 0.3701
## dataf$Consume_Alco_Code   2.67   1  1.4592 0.2283
## dataf$Work_Status_Code    0.17   1  0.0937 0.7597
## dataf$Hypertension_Code   0.00   1  0.0023 0.9614
## dataf$BPMed_Code          2.85   1  1.5582 0.2132
## dataf$High_Chol_Code      1.56   1  0.8503 0.3574
## dataf$CholMed_Code        4.04   1  2.2088 0.1386
## dataf$Diabetes_Code       1.35   1  0.7366 0.3916
## dataf$DiabetesMed_Code    1.42   1  0.7769 0.3790
## dataf$Heart_Cond_Code     4.56   1  2.4947 0.1156
## dataf$Fat_Mass            1.14   1  0.6252 0.4299
## dataf$dash                0.36   1  0.1994 0.6556
## Residuals               428.09 234

Supplementary Figure 17. The relationship between daily behaviours and Glucose as estimated by compositional linear regression models.Difference in minutes modelled around the population mean composition.

Supplementary Figure 18. The relationship between daily behaviours and Glucose as estimated by compositional linear regression models.Predictions based on realloacting each actvity type for another.

HbA1c

Supplementary Table 14.0 Multiple linear regression analyses of the relationship between first ilr coordinates and HbA1C
term estimate std.error statistic p.value
ilr1 / ln (Sleep: geometric mean of remaining behaviours) -2.413 2.975 -0.811 0.418
ilr1 / ln (Sedentary: geometric mean of remaining behaviours) 0.060 2.041 0.029 0.977
ilr1 / ln (Standing: geometric mean of remaining behaviours) 0.949 1.828 0.519 0.604
ilr1 / ln (LIPA: geometric mean of remaining behaviours) 1.811 1.687 1.073 0.284
ilr1 / ln (MVPA: geometric mean of remaining behaviours) -0.407 0.629 -0.647 0.518
Note:
ilr: Isometric log-ratio; HbA1C: , Beta: Unstandardised regression coefficient estimate, SE: Standard error; LIPA: Light-intensity physical activity; MVPA: Moderate- to vigorous physical activity.
* Note; All models adjusted for age, sex, smoking status, alcohol consumption,fat mass,DASH score, work status, comorbidities including hypertension,hyperlipidemia, diabetes and history of heart conditons along with diabetic,lipd and BP medications.
† Number of observations:
‡ 250
§ Adjusted R-squared:
¶ 0.02 
## Anova Table (Type II tests)
## 
## Response: y
##                          Sum Sq  Df F value  Pr(>F)  
## ilr_comps                 162.1   4  0.7181 0.58030  
## dataf$Age                  65.4   1  1.1587 0.28286  
## dataf$Sex_Code             40.8   1  0.7236 0.39586  
## dataf$Smoking_Code         77.0   1  1.3641 0.24403  
## dataf$Consume_Alco_Code    35.7   1  0.6328 0.42716  
## dataf$Work_Status_Code      8.9   1  0.1573 0.69205  
## dataf$Hypertension_Code    12.4   1  0.2194 0.63996  
## dataf$BPMed_Code          192.8   1  3.4174 0.06579 .
## dataf$High_Chol_Code       87.8   1  1.5568 0.21340  
## dataf$CholMed_Code        311.9   1  5.5270 0.01957 *
## dataf$Diabetes_Code        54.3   1  0.9627 0.32752  
## dataf$DiabetesMed_Code     53.6   1  0.9497 0.33081  
## dataf$Heart_Cond_Code      66.9   1  1.1855 0.27737  
## dataf$Fat_Mass            141.2   1  2.5017 0.11509  
## dataf$dash                  0.1   1  0.0016 0.96812  
## Residuals               13034.5 231                  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Supplementary Figure 19. The relationship between daily behaviours and HbA1C as estimated by compositional linear regression models.Difference in minutes modelled around the population mean composition.

Supplementary Figure 20. The relationship between daily behaviours and HbA1C as estimated by compositional linear regression models.Predictions based on realloacting each actvity type for another.

Predicted changes in Cardiometabolic Indicators