1. General Characteristics of the Study Subjects
A total of 3525 study subjects were selected out of 8110 subjects. A total of 3525 subjects were included in the study; 1480 were males, 2045 were females, and all individuals were over the age of 40. In the 2019 KNHANES, NC was measured among people aged over 40 years. A total of 3302 subjects who were under the age of 40 years were excluded from the study. A total of 249 subjects who did not have NC measurement data were excluded from the study. Seventy subjects who did not have data on alcohol consumption and smoking were excluded from the study. A total of 196 subjects who did not submit medical diagnosis records were excluded from the study. A total of 615 subjects without PFT results were excluded from the study. A total of 153 subjects without laboratory results were excluded from the study. The patient flow chart is shown in Figure 1.
The prevalence of obesity was more common in females than males 2.9% and 1.4% respectively The prevalence of normal weight was more common in females than males 42.8% and 27.8% respectively.. Preobese was more common in males than females 28.8% and 23.4% respectively. The prevalence of obese class 1 was more common in males than females 37.0% and 25.8% respectively. The prevalence of obese class 2 was more common in females than males 4.5% and 4.2% recpectively.. The prevalence of obese class 3 was more common in males than females 0.8% and 0.6%, respectively. Obesity class was define by the Korean Society for the Study of Obesity 2018 guideline12.
The PFT results differed by sex. The rates of normal PFT results were more common in females than males 72.1% and 57.8% respectively.. restrictive lung disease were more common in males than females 15.9% and 8.0%, respectively. obstructive lung disease were more common in males than females 14.9% and 5.4%, respectively. Forced expiratory flow at 25-75% (FEF25-75) was higher in males than females 2.7 L/sec and 2.3 L/sec, respectively.
Total cholesterol was higher in females than males 199.4mg/dL and 194.6mg/dL respectively.. HDL cholesterol level was higher in females than males 55.8mg/dL and 48.2mg/dL respectively.. LDL cholesterol level was higher in females than males 120.5mg/dL and 115.7mg/dL respectively.. TG level was higher in males than females 166.4 mg/dL and 119.1 mg/dL, respectively. Fasting glucose level was higherin males than females 165.8 mg/dL and 117.5 mg/dL, respectively. HbA1c level was higher in males than females 5.9 mmol/mol and 5.8 mmol/mol, respectively. Fasting serum insulin level was higher in males than females 8.7 mIU/L and 8.2 mIU/L, respectively.
Serum BUN level was higher in males than females 16.5 mg/dL and 15.5 mg/dL, respectively. Serum creatinine level was higher in males than 0.9 mg/dL and 0.7 mg/dL, respectively. Serum aspartate transaminase (AST) level was higher in males than females 27.0 U/ml and 23.7 U/ml, respectively. Serum alanine transaminase (ALT) level was higher in males than females 27.8 U/ml and 20.0 U/ml, respectively. The hemoglobin level was higher in males than females 15.2 g/dL and 13.2 g/dL, respectively. The hematocrit percentages was higher in males than females 45.2% and 40.0%, respectively. The uric acid level was higher in males than females 5.9 mg/dL and 4.5 mg/dL, respectively.
2. Correlations of Anthropometric Measurements and CMAMs with Laboratory Data in All Study Subjects (Males and Females)
2.1.1 Correlations between NC and Laboratory Data in All Study Subjects (Males and Females)
To find if NC had a positive correlation with cardiometabolic risk factors and PFT, NC and its correlation with various laboratory results and measurements were analyzed. NC had positive correlations with BMI, WC, total cholesterol, TGs, LDL, fasting glucose, HbA1c, serum insulin, BUN, creatinine, AST, ALT, hemoglobin, hematocrit, uric acid, FVC, FEV1, FEV1/FVC and FEF25_75. The highest beta coefficient was observed for TGs at 10.62, followed by WC at 3.23. NC had a negative correlation with HDL, with a beta coefficient of -1.28. The above results all had p-values <0.05 (Table 2). he observed findings suggest that NC had a positive correlation with cardiometabolic risk factors and also had a positive correlation with normal PFT function.
2.1.2 Correlations of NC/BMI with Laboratory Data in All Study Subjects (Males and Females)
Next, to find if combination of NC and BMI had a positive correlation with cardiometabolic risk factors and PFT, NC devided by BMI and its correlation with various laboratory results and measurements were analyzed.. NC/BMI was positively correlated with HDL, FVC and FEV1, with HDL showing the highest beta coefficient (17.71). NC/BMI was negatively correlated with BMI, WC, total cholesterol, TGs, LDL, fasting glucose, HbA1c, serum insulin, BUN, creatinine, AST, ALT, hemoglobin, hematocrit, uric acid, FEV1/FVC and FEF25-75. The lowest beta coefficient was observed for TGs at -118.66, followed by WC at -50.45 (Table 2).
2.1.3 Correlations of NC/WC with Laboratory Data in All Study Subjects (Males and Females)
Next to find if combination of NC and WC had a positive correlation with cardiometabolic risk factors and PFT, NC devided by WC and its correlation with various laboratory results and measurements were analyzed. NC/WC was positively correlated with HDL and FVC, with beta coefficients of 82.65 and 0.88, respectively. NC/WC was negatively correlated with BMI, WC, TGs, fasting glucose, HbA1c, serum insulin, AST, ALT, hemoglobin, hematocrit, uric acid, FEV1/FVC and FEF25_75. The lowest beta coefficient was observed for TGs (-577.75), followed by WC (-252.16), fasting glucose (-117.88), ALT (-116.65), BMI (-73.77), AST (-44.14), hematocrit (-13.09) and uric acid (-6.10) (Table 2).
2.2 Correlations of Anthropometric Measurements(NC, BMI, WC) and CMAMs(NC/BMI, NC/WC) with Laboratory Data in Male Subjects
2.2.1 Correlations of NC and Laboratory Data in Male Subjects
NC and its correlations with various laboratory results and measurements were further analyzed among male subjects. NC was positively correlated with BMI, WC, total cholesterol, TGs, fasting glucose, HbA1c, serum insulin, BUN, creatinine, AST, ALT, hemoglobin, hematocrit, uric acid, FVC, FEV1, FEV1/FVC and FEF25_75. The highest beta coefficient was observed for TGs (9.77), followed by WC (2.95) and ALT (2.55). There was negative correlation between NC and HDL, with a beta coefficient of -0.94 (Table 3). NC had a positive correlation with cardiometabolic risk factors in males.
2.2.2 Correlations of NC/BMI with Laboratory Data in Male Subjects
Regarding NC/BMI and its correlations with various laboratory results and measurements among male subjects, NC/BMI was positively correlated with HDL, FVC and FEV1, with the highest beta coefficient being observed for HDL at 15.15. NC/BMI was negatively correlated with BMI, WC, TGs, fasting glucose, HbA1c, serum insulin, creatinine, AST, ALT, hemoglobin, hematocrit, uric acid, FEV1/FVC and FEF25_75. The lowest beta coefficient was observed for TGs (-122.50), followed by WC (-49.70), ALT (-33.78), BMI (-20.95), fasting glucose (-19.66) and AST (-10.78) (Table 3).NC/BMI had a negative correlation with cardiometabolic risk factors in males.
2.2.3 Correlations of NC/ WC with Laboratory Data in Male Subjects
Then, NC/WC and its correlations with various laboratory results and measurements among male subjects were analyzed. NC/WC was positively correlated with HDL, FVC, and FEV1, and the highest beta coefficient was for HDL at 68.15. NC/WC was negatively correlated with WC, TGs, fasting glucose, HbA1c, serum insulin, AST, ALT, hemoglobin, hematocrit, uric acid and FEF25_75. The lowest coefficient was observed for TG at -658.32, followed by WC (-249.09), ALT (-153.46), fasting glucose (-103.41), serum insulin (-88.75), AST (-60.97) and hematocrit (-11.79) (Table 3). NC/WC had a negative correlation with cardiometabolic risk factors in males.
The observed findings suggested that NC had a positive correlation with cardiometabolic risk factors whereas NC/BMI and NC/WC had a negative correlation with cardiometabolic risk factors.
2.3 Correlations of Anthropometric Measurements(NC, BMI, WC) and CMAMs(NC/BMI, NC/WC) with Laboratory Data in female Subjects
2.3.1 Correlations of NC with Laboratory Data in Female Subjects
NC and its correlations with various laboratory results and measurements among female subjects were analyzed. NC alone had positive correlations with BMI, WC, TGs, fasting glucose, HbA1c, serum insulin, creatinine, AST, ALT, hemoglobin, hematocrit, uric acid and FEF25_75 among female subjects. The highest beta coefficient was observed for TG (10.68), followed by WC (3.54) and fasting glucose (3.07). There was negative correlation between NC and HDL, with a beta coefficient of -1.67 (Table 4. NC had positive correlation with cardiometabolic risk factors in females.
2.3.2 Correlations of NC/BMI with Laboratory Data in Female Subjects
NC/BMI and its correlations with various laboratory results and measurements among female subjects were also analyzed. NC/BMI had positive correlations with HDL and FVC (highest beta coefficient was observed for HDL at 18.94) and negative correlations with BMI, WC, TGs, fasting glucose, HbA1c, serum insulin, AST, ALT, hemoglobin, hematocrit, uric acid, FEV1/FVC and FEF 25_75. The lowest beta coefficients were observed for TGs (-100.92), WC (-50.43), BMI (-22.18), ALT (-21.09) and serum insulin (-15.65). NC/BMI had negative correlation with cardiometabolic risk factors in females.
2.3.3 Correlations of NC/WC with Laboratory Data in Female Subjects
NC/WC and its correlations with various laboratory results and measurements among female subjects were analyzed. NC/WC was positively correlated with HDL, with a beta coefficient of 90.83, and negatively correlated with BMI, WC, TGs, fasting glucose, HbA1c, serum insulin, AST, ALT, hemoglobin, hematocrit and uric acid. The lowest beta coefficient was observed for TG (-456.27), followed by WC (-251.61), fasting glucose (-128.92), BMI (-77.66), ALT (-77.17), serum insulin (-63.97) and AST (-26.72). NC/WC had negative correlation with cardiometabolic risk factors in females.
The observed correlations of NC, NC/BMI and NC/WC with various laboratory results and measurements showed that the beta coefficients were not the same across the three variables but were unique for certain laboratory results and measurements.
3. Determining the Precision of Anthropometric Measurements and CMAMs with Precision-recall Plots
Since correlation studies do not indicate precision, precision-recall plot analysis was performed.
3.1.1. Precision Among All Study Subjects (Both Males and Females)
Precision-recall plots and areas under the curve (AUCs) were used to evaluate the precision of the anthropometric measurements and CMAMs at each present threshold. These data can be seen in Figure 2. When the BMI thresholds were applied, WC had the highest AUCs (0.73 and 0.729), followed by NC/BMI (0.603 and 0.591). In contrast, when WC thresholds were used, BMI had the highest AUCs (0.779 and 0.907), followed by NC/WC (0.534 and 0.905). These results show that BMI 30 kg/m2 and over was best predicted with WC whereas WC for male 90cm and over for female 80cm and over was best predicted with BMI.
WC/BMI had the highest AUCs (0.121 and 0.514), followed by NC/BMI (0.119 and 0.478) when total cholesterol thresholds were applied, but NC/WC had the highest AUCs (0.159 and 0.418) when HDL thresholds were used, while NC/BMI (0.113) and WC/BMI (0.396) had the highest AUCs depending on if LDL_y1 (>90th ) or LDL_y2 (>130) was applied. Thus, Total cholesterol 200mg/dL and over was best predicted with WC/BMI where as HDL for male under 40mg/dL and for female under 50mg/dL was best predicted with NC/WC. LDL was best predicted with NC/BMI and WC/BMI. LDL of 90% and over was best predicted with NC/BMI whereas LDL 130mg/dL and over was best predicted with WC/BMI.
Regardless of the TG threshold, NC had the highest AUCs (0.194 and 0.46), while WC had the highest AUCs (0.209 and 0.615) for both glucose thresholds. Similarly, WC also had the highest AUCs (0.236 and 0.259) for both HbA1c thresholds, while BMI had the highest AUCs (0.359 and 0.359) for both insulin thresholds.
WC/BMI had the highest AUCs (0.141 and 0.108) for both BUN thresholds. However, NC/BMI had the highest AUC (0.154) for crea_y1 (creatinine >90th ), but NC had the highest AUC (0.061) for crea_y2 (>1.2), showing that kidney dysfunction which is characterized by elevation of serum creatinine of more than 1.2mg/dL is best predicted with NC. Moreover, WC had the highest AUCs (0.187 and 0.143) for both AST thresholds and for both ALT thresholds (AUCs of 0.233 and 0.223) showing that liver dysfunction such as fatty liver can be best predicted with WC.
For hemoglobin levels, both the upper and lower ends of the distribution were analyzed. The highest AUCs were observed for BMI (0.133), NC (0.065), WC/BMI (0.161) and NC (0.166) for the hemoglobin thresholds (Hb >90th percentile, >17, <10th percentile, and < 13.5), respectively showing that excess red blood cell (polycythemia) was best predicted with BMI whereas anemia was best predicted with NC. Similarly, the highest AUCs were observed for BMI (0.139), NC (0. 094), WC/BMI (0. 0.145) and WC/BMI (0.151) for the Hct (Hematocrit) thresholds (>90th percentile, >50, <10th percentile and <41, respectively). For the uric acid threshold of >90th percentile, NC had the highest AUC (0.234), but the number of cases was too small for analysis for uacid_y2. Therefore, for elevation of uric acids such as in gout, NC had the highest predictive value.
Regarding PFT results, WC had the highest AUCs (0.319 and 0.487) for both PFT_y1 (aged between 40 and 60 years, FEV1/FVC>70 and FVC pred %<80 equivalent to restrictive pattern) and PFT_y2 (aged 60 years and over, FEV1/FVC>60 and FVC pred %<80 equivalent to restrictive pattern). However, BMI had the highest AUC of 0.35 for PFT_y3 (aged between 40 and 60 years, FEV1/FVC<70 and FVC pred %<80 equivalent to combined restrictive pattern), while NC/WC had the highest AUC of 0.445 for PFT_y4 (aged 60 years and over, FEV1/FVC<60 and FVC pred %<80 equivalent to combined restrictive pattern). For PFT_y5 (aged between 40 and 60 years, FEV1/FVC<70, FVC pred %>80 and FEV1<100 equivalent to obstructive pattern) and PFT_y6 (aged 60 years and over, FEV1/FVC<60, FVC pred %>80 and FEV1<100 equivalent to obstructive pattern), the number of cases was too small to be analyzed. These results indicate thatthat people age over 40 with restrictive lung disease was best predicted with WC, whereas people ages between 40 to 60 with combined restrictive disease was best predicted with BMI and for people age over 60 with combined restrictive disease was best predicted with NC/WC. Since NC/WC is negatively correlated with FEV1/FVC and FEF25_75 as been found in this study, elderlies(people age over 60) with high NC/WC ratio should get PFT screening in order to early detect lung disease.
In all subjects, total cholesterol was best predicted by WC/BMI, HDL was best predicted by NC/WC, TGs were best predicted by NC, LDL was best predicted by WC/BMI, fasting glucose was best predicted by WC, HbA1c was best predicted by WC, serum insulin was best predicted by BMI, BUN was best predicted by WC/BMI, creatinine was best predicted by NC, AST was best predicted by WC, ALT was best predicted by WC, Hb (upper range) and Hb (lower range) were both best predicted by NC, Hct (upper range) was best predicted by NC, and Hct (lower range) was best predicted by WC/BMI. For PFT, a restrictive pattern was best predicted by WC in age groups of 40 to 60 years and 60 years and older. The combined restrictive pattern was best predicted by BMI in patients aged 40 to 60 years old, whereas in patients aged 60 years and older, the combined restrictive pattern was best predicted by NC/WC. The numbers for obstructive patterns were too small to be analyzed.
3.1.2 Precision Among Male Subjects
A similar analysis was carried out in male subjects. These data can be seen in Figure 3. Total cholesterol was best predicted with NC (AUC of 0.113) for chol_y1(>90th) and with WC/BMI (AUC of 0.471)for chol_y2(total cholesterol>200). HDL_y1 (<10th ) was best predicted with NC and WC (AUC of 0.132) whereas HDL_y2 (HDL<40) was best predicted with WC (AUC of 0.305). TG_y1 (>90th ) and TG_y2(TG>150)was both best predicted with NC (AUC of 0.228 and 0.512 respectively). LDL_y1 (>90th ) was best predicted with NC (AUC of 0.087) and LDL_y2 (>130) was best predicted with WC/BMI ( AUC of 0.36. glucose_y1 (>90th ) was best predicted with WC and NC (AUCs, both 0.233whereas glucose_y2 (>100) was best predicted with WC ( AUC of 0.652), HbA1c_y1 (>90th ) and HbA1c_y2(6.5 and over) was best predicted with WC (AUC of 0.257 and 0.278 respectively). insulin_y1 (>90th ) and insulin_y2(15uIU/mL and over) was best predicted with BMI(AUC of 0.392 and 0.39 respectively). BUN_y1 (>90th ) and BUN_y2(23mg/dL and over) was best predicted with WC/BMI (AUC of 0.159 and 0.126 respectively) crea_y1 (creatinine >90th ) and crea_y2(Creatinine>1.2mg/dL) was best predicted with NC/WC(AUC of 0.231 and 0.086 respectively)., AST_y1 (>90th ) and AST_y2(40mg/dL) was best predicted with WC (AUC of 0.224 and 0.161 respectively), ALT_y1 (>90th ) and ALT_y2(>41mg/dL) BMI had the highest predictive value, AUC of 0.294 and 0.287 respectively.
For hemoglobin levels, both the upper and lower ends of the distribution were analyzed. Hb_y1 (Hemoglobin>90th ) was best predicted with NC( AUC of 0.138), whereas Hb_y2 (Hemoglobin>17g/dL) was best predicted with BMI (AUC of 0.073)Hb_y3 (Hemoglobin<10th ) was best predicted with NC and WC/BMI(both AUC of 0.202) whereas Hb_y4 (Hemoglobin<13.5g/dL) was best predicted with NC (AUC of 0.175). Hct_y1 (Hematocrit >90th ) was best predicted with BMI and WC (AUC of 0.133) Hct_y2 (Hematocrit >50%) was best predicted with WC ( AUC of 0.111 Hct_y3 (Hematocrit<10th ) and Hct_y4 (Hematocrit<41) was best predicted withNC ( AUC of 0.2 and 0.209 respectively). uacid_y1 (Uric acid >90th ) was best predicted with WC and BMI (both AUC 0.267), for uacid_y2, the number of cases was too small for analysis.
For PFT_y1 (aged between 40 and 60 years, FEV1/FVC>70 and FVC pred %<80 equivalent to restrictive pattern) and PFT_y2 (aged 60 years and over, FEV1/FVC>60 and FVC pred %<80 equivalent to restrictive pattern) was best predicted with WC(AUC of 0.351 and 0.6 respectively) PFT_y3 (aged between 40 and 60 years, FEV1/FVC<70 and FVC pred %<80 equivalent to combined restrictive pattern) was best predicted with ( AUC of 0.348), PFT_y4 (aged 60 years and over, FEV1/FVC<60 and FVC pred %<80 equivalent to combined restrictive pattern) was best predicted with NC/BMI PFT_y5 (aged between 40 and 60 years, FEV1/FVC<70, FVC pred %>80 and FEV1<100 equivalent to obstructive pattern) and PFT_y6 (aged 60 years and over, FEV1/FVC<60, FVC pred %>80 and FEV1<100 equivalent to obstructive pattern), the number of cases was too small to be analyzed.
In male subjects, the predictions were the same as in the total number of subjects except for HDL, creatinine, ALT, Hb (upper range), Hct (lower range) and combined restrictive pattern in individuals aged 60 years and over. HDL was best predicted by WC. Creatinine was best predicted by NC/WC. ALT was best predicted by WC, and Hb (hemoglobin>17) was best predicted by BMI. Hct (<41) was best predicted by NC. The combined restrictive PFT pattern in individuals aged 60 years and over was best predicted by NC/BMI.
3.1.3 Precision Among Female Subjects
Finally, the same PR plots and AUCs were analyzed in female subjects. These data can be seen in Figure 4. chol_y1 (>90th ) was best predicted with NC (AUC of 0.137) whereas chol_y2 (total cholesterol>200) was best predicted with WC/BMI (AUC of 0.535) HDL_y1 (<10th ) was best predicted with NC (AUC of 0.197) whereas HDL_y2(HDL<50mg/dL) was best predicted with WC (AUC of 0.487) TG_y1 (>90th ) and TG_y2(TG>150mg/dL) was best predicted with NC (AUC of 0.228 and 0.369 respectively) LDL_y1 (>90th ) was best predicted with NC ( AUC of 0.126) whereas LDL_y2 (>130mg/dL) was best predicted with WC/( AUC of 0.413) glucose_y1 (>90th ) and glucose_y2(>100mg/dL) was best predicted with WC (AUC of 0.176 and 0.566 respectively) HbA1c_y1 (>90th ) and HbA1c_y2(>6.5) was best predicted with WC (AUC of 0.216 and 0.202 respectively) insulin_y1 (>90th ) and insulin_y2(15uIU/mL) was best predicted with WC (AUC of 0.345 and 0.347 respectively) BUN_y1 (>90th ) and BUN_y2(23mg/dL) was best predicted with NC/WC (AUC of 0.136 and 0.106 respectively). crea_y1 (creatinine >90th ) was best predicted with NC/WC and WC (both AUC of 0.028) whereas crea_y2 (>1.2) was best predicted with WC/BMI (AUC of 0.018 AST_y1 (>90th ) was best predicted with NC and BMI (both AUCof 0.138) whereas AST_y2 (>40mg/dL) was best predicted with BMI ( AUC of 0.128) ALT_y1 (>90th ) and ALT_y2(41mg/dL) was best predicted with BMI ( AUC of 0.173 and 0.162 respectively).
For hemoglobin levels, both the upper and lower ends of the distribution were analyzed. Hb_y1 (Hemoglobin>90th ) was best predicted with NC ( AUC of 0.149whereas Hb_y2 (Hemoglobin>16g/dL) was best predicted with BMI ( AUC of 0.01) Hb_y3 (Hemoglobin<10th ) was best predicted with ( AUC of 0.129) whereas Hb_y4 (Hemoglobin<12g/dL) was best predicted with WC ( AUC of 0.169) Hct_y1 (Hematocrit >90th ) and Hct_y2(Hematocrit >47%) was best predicted with NC ( AUC of 0.165 and 0.02 respectively). Hct_y3 (Hematocrit<10th ) and Hct_y4(Hematocrit< 36g/dL) was best predicted with WC ( both AUC of 0.124) uacid_y1 (Uric acid >90th ) was best predicted with BMI ( AUC of 0.04For uacid_y2, the number of cases was too small for analysis.
PFT_y1 (aged between 40 and 60 years, FEV1/FVC>70 and FVC pred %<80 equivalent to restrictive pattern) and PFT_y2 (aged 60 years and over, FEV1/FVC>60 and FVC pred %<80 equivalent to restrictive pattern) was best predicted with WC ( AUC of 0.292 and 0.371 respectively) PFT_y3 (aged between 40~60, FEV1/FVC<70 and FVC pred %<80 equivalent to combined restrictive pattern) was best predicted with WC (AUC of 0.374 PFT_y4 (aged 60 years and over, FEV1/FVC<60 and FVC pred %<80 equivalent to combined restrictive pattern) was best predicted with WC and NC/WC (both AUC of 0.818For PFT_y5 (aged between 40 and 60 years, FEV1/FVC<70, FVC pred %>80 and FEV1<100 equivalent to obstructive pattern) and PFT_y6 (aged 60 years and over, FEV1/FVC<60, FVC pred %>80 and FEV1<100 equivalent to obstructive pattern), the number of cases was too small to be analyzed.
In female subjects, the predictions were the same as in the total number of subjects except for HDL, serum insulin, BUN, creatinine, AST, ALT, Hb (upper range), Hb (lower range), Hct (lower range) and combined restrictive pattern in individuals aged 40 to 60 years old. HDL was best predicted by WC. Serum insulin was best predicted by WC. BUN was best predicted by NC/WC, creatinine was best predicted by WC/BMI, AST was best predicted by BMI, ALT was best predicted by BMI, and Hb (Hemoglobin>16) was predicted by BMI. Hct (Hematocrit<36) was best predicted by WC. The combined restrictive pattern in individuals aged 40 to 60 years old was best predicted by WC.
Correlation Analysis of Anthropometric Measurements with the Number of Metabolic Syndrome Criteria Met
NC, BMI, WC, NC/WC, WC/BMI, NC/BMI were analyzed based on the number of metabolic syndrome criteria present. These data are shown in Table 6. Clear trends were shown with NC, BMI and WC.
Among all subjects, the mean NC in subjects with zero metabolic syndrome criteria was 33.4cm, and increased to 34.7cm for one criterion, 35.7cm for two criteria, 36.7cm for three criteria, 37.5cm for four criteria and 37.3cm for five criteria. Similarly, the mean BMI increased from 21.9 kg/m2 to 23.1 kg/m2, 24.5 kg/m2, 25.7 kg/m2, 27.1 kg/m2, and 27.0 kg/m2 for subjects with zero to five metabolic syndrome criteria, respectively. Similarly, the mean WC increased from 77.1cm to 81.7cm, 86.8cm, 90.5cm, 93.8cm and 95.3cm for subjects with zero to five metabolic syndrome criteria, respectively.
In male subjects, the mean NC in subjects with zero metabolic syndrome criteria was 36.6cm, and increased to 37.4cm for one criterion, 38.2cm for two criteria, 38.8cm for three criteria, 39.9cm for four criteria and 39.8cm for five criteria. Similiarly the mean BMI increased from 22.4 kg/m2 to 23.6 kg/m2, 24.6 kg/m2, 25.6 kg/m2, 27.2 kg/m2 and 27.0 kg/m2 for subjects with zero to five metabolic syndrome criteria, respectively. The mean WC increased from 81.4cm to 85.3cm, 89.3cm, 91.8cm, 96.1cm and 96.6cm for subjects with zero to five metabolic syndrome criteria, respectively.
Similarly, In female subjects, the mean NC increased from31.7cm, 32.4cm, 33.0cm, 33.9cm, 34.5cm and 34.8cm for subjects with zero to five metabolic syndrome criteria, respectively. The mean BMI increased from 21.6 kg/m2 to 22.7 kg/m2, 24.3 kg/m2, 25.9 kg/m2, 26.8 kg/m2 and 27.0 kg/m2 for subjects with zero to five metabolic syndrome criteria, respectively. The mean WC increased from 74.9cm to 79.0cm, 84.0cm, 88.5cm, 91.4cm and 92.9cm for subjects with zero to five metablic syndrome criteria, respectively.
A positive linear pattern was found between NC/BMI, NC/WC, BMI, WC and the number of metabolic syndrome criteria present. Linear trend p value between NC/BMI, NC/WC, BMI, WC and the number of metabolic syndrome criteria was statistically significant before adjusted, adjusted for alcohol and smoking and adjusted for alcohol, smoking, age and gender(in total subjects). For further analysis of the precision of anthropometric measurements in predicting the number of metabolic syndrome criteria present. precision-recall plot analysis was performed.
Precision-recall plot of anthropometric measurements and CMAMs with metabolic syndrome criteria
Precision-recall plots showing the precision of anthropometric measurements and CMAMs for metabolic syndrome criteria were generated. These data are shown in Figure 5.
Among all study subjects, including both males and females, in the presence of more than 3, more than 4 and all 5 metabolic syndrome criteria, WC had the highest AUCs (0.62, 0.342 and 0.09), followed by BMI (0.587, 0.317 and 0.083), followed by NC (0.519, 0.263, 0.063), followed by NC/WC (0.49, 0.229, 0.064)
In male study subjects, in males, in the presence of more than 3, more than 4 and all metabolic syndrome criteria, WC had the highest AUCs (0.626, 0.354, 0.083), followed by BMI (0.601, 0.331, 0.076), followed by NC (0.599, 0.329, 0.083) (Figure 6).
In female study subjects, in females, in the presence of more than 3, more than 4 and all metabolic syndrome criteria, WC had the highest AUCs (0.628, 0.349, 0. 108) followed by BMI (0.575), NC(0.321, 0.093). (Figure 7).
There was a difference in the AUC of each anthropometric measurement evaluated according to the number of metabolic syndrome criteria present and gender The highest AUC observed in WC through all metabolic syndrome criteria is probably because NC is one of the factors in metabolic syndrome. In males, the second highest prediction value was BMI but AUCs of BMI did not show much difference with NC. But, when all metabolic syndrome criteria were met, NC had the highest prediction value.
Similarly, in female, the highest AUC was observed in WC. The second highest prediction value was observed with BMI and NC. The obove finding suggests that NC has more prediction value in females than in males with metabolic syndrome.