A data set of 63 severely obese Europid women were analysed and a group of 17 were identified as outliers by trimmed means on the 5% level. Of these, 10 proved to be significant. A set of two subjects with glucose values >230 mg/dl were completely taken out. Two had CRP values >24 mg/dl and were excluded, three data were missing. TG, LDL and HDL each had two values >310 mg/dl, >240 mg/dl, >100 mg/dl, respectively and were excluded. The remaining cases are summarized in Table 1.
Table 1 Anthropometric and metabolic risk factors of obese German women
|
n
|
Mean
|
±SD
|
Median
|
Min
|
Max
|
Age (years)
|
61
|
41.6
|
12.1
|
42.0
|
18.0
|
65.0
|
Height (cm)
|
61
|
166.5
|
6.3
|
166.1
|
153.7
|
179.2
|
Weight (kg)
|
61
|
108.1
|
16.5
|
105.4
|
80.6
|
147.0
|
HdC (cm)
|
61
|
19.4
|
1.0
|
19.5
|
16.5
|
21.4
|
AC (cm)
|
61
|
118.3
|
12.2
|
118.7
|
94.0
|
147.0
|
AC/Ht (cm/cm)
|
61
|
0.71
|
0.07
|
0.71
|
0.55
|
0.89
|
BMI (kg/m²)
|
61
|
38.9
|
4.9
|
38.2
|
29.7
|
51.7
|
%FM-BIA (%)
|
61
|
49.8
|
4.5
|
49.6
|
36.9
|
59.2
|
%FM-DBA (%)
|
61
|
49.1
|
4.8
|
49.4
|
35.9
|
57.9
|
|
|
|
|
|
|
|
Glucose (mg/dl)
|
61
|
96.4
|
14.5
|
93.0
|
76.0
|
149.0
|
Insulin (mU/l)
|
61
|
19.1
|
9.8
|
18.2
|
3.8
|
56.9
|
HOMA-IR (mU/l, mmol/l)
|
61
|
4.6
|
2.8
|
3.9
|
0.7
|
18.6
|
TG (mg/dl)
|
59
|
141.7
|
55.9
|
134.0
|
43.0
|
286.0
|
HDL-C (mg/dl)
|
59
|
52.6
|
13.1
|
51.0
|
32.0
|
85.0
|
LDL-C (mg/dl)
|
59
|
126.7
|
25.9
|
127.0
|
76.0
|
190.0
|
CRP (mg/l)
|
56
|
5.9
|
4.1
|
4.6
|
0.9
|
19.0
|
SBP (mmHg)
|
61
|
131.4
|
17.2
|
130.0
|
100.0
|
179,0
|
Abbreviations: BMI, body mass index; AC, abdomen circumference; HdC, hand circumference; Ht, height; %FM, percentage fat mass; BIA, bioimpedance analysis; DBA, Dahlmann-Body-Analysis; HOMA-IR, homeostasis model assessment of insulin resistance; TG, triglycerides; HDL, high-density lipoproteins; LDL, low-density lipoproteins; CRP, C-reactive protein; SBP, systolic blood pressure.
Age ranged between 18 and 65 years. All subjects had a BMI >30kg/m², an abdomen circumference > 94 cm and a relationship AC/Ht >0.55 cm/cm. The mean of %FM was about 50%, measured by the two methods BIA and DBA, respectively. With the exception of CRP and SBP the mean values of all the other metabolic risk factors are lying within the recommended normal range. The HOMA-IR value 4.6 is in the same order of magnitude (4.7) compared to a Brazilian study of obese women (BMI = 40.5) [23].
We furthermore performed correlation analyses to elucidate associations between obesity indices and age. The results are presented in Table 2.
Table 2 Results of the correlation analyses between obesity indices and age.
|
Age
|
Height
|
Weight
|
HdC
|
AC
|
AC/Ht
|
BMI
|
%FM-BIA
|
%FM-DBA
|
Age
|
1.00
|
|
|
|
|
|
|
|
|
Height
|
-0.17
|
1.00
|
|
|
|
|
|
|
|
Weight
|
-0.13
|
0.56
|
1.00
|
|
|
|
|
|
|
HdC
|
0.12
|
0.23
|
0.36
|
1.00
|
|
|
|
|
|
AC
|
-0.05
|
0.15
|
0.70
|
0.39
|
1.00
|
|
|
|
|
AC/Ht
|
0.02
|
-0.21
|
0.49
|
0.30
|
0.93
|
1.00
|
|
|
|
BMI
|
-0.06
|
0.07
|
0.86
|
0.28
|
0.75
|
0.71
|
1.00
|
|
|
%FM-BIA
|
-0.28
|
0.05
|
0.72
|
0.14
|
0.70
|
0.67
|
0.83
|
1.00
|
|
%FM-DBA
|
-0.13
|
0.11
|
0.82
|
0.12
|
0.74
|
0.69
|
0.91
|
0.85
|
1.00
|
Abbreviations: BMI, body mass index; AC, abdomen circumference; HdC, hand circumference; Ht, height; %FM, percentage fat mass; BIA, bioimpedance analysis; DBA, Dahlmann-Body-Analysis;
The correlation coefficients between age and %FM are negative for the BIA as well as the DBA measurement, indicating that younger women are more obese than older ones. All obesity indices were highly correlated with each other. The correlations between BMI and the indices of central adiposity AC (r=0.75) and AC/Ht (r=0.71) were lower than the one between BMI and the general adiposity marker (%FM). The highest correlation coefficient was found between BMI and %FM-DBA (r=0.91). Fat mass measurement, either performed by a BIA device or the Dahlmann model, showed an excellent agreement expressed by a correlation coefficient of 0.85 as a sign that both methods are equal to calculate the %FM.
Associations between body fat mass measured by the DBA system (%FM-DBA) and the systolic blood pressure and seven metabolic risk factors are plotted in Figs. 1a – 1h.
Figures 1a to 1h
The graphical representation shows a linear relationship. The equations of the corresponding regression lines are given in Table 3. The indicated regression coefficient represents the slope ß, which is a measure of the contribution of the body fat volume toward the depending metabolic variables. A positive relationship is represented by a rising (ß > 0) or falling (ß < 0) regression line. The t-test (null hypothesis ß = 0) revealed significant rising slopes for the parameters Insulin, HOMA-IR, and CRP and HDL-C showing an inverse relation. The corresponding correlation coefficients are all I > 0,30 I. The traits Glucose, TG, LDL-C and SBP have correlation coefficients r < 0.12.
Table 3 Regression analysis of metabolic parameters vs. %FM-DBA
Metabolic Parameters
|
Linear Regr. Equation
|
R²
|
r
|
Slope
|
p-value
|
Glucose (mg/dl)
|
y = 0,20x + 86,4
|
0,005
|
0,07
|
ß
|
0,604
|
Insulin (mU/l)
|
y = 0,95x - 27,5
|
0,217
|
0,47
|
ß*
|
0,001
|
HOMA-IR (mU/l, mmol/l
|
y = 0,24x - 6,9
|
0,162
|
0,40
|
ß*
|
0,001
|
TG (mg/dl)
|
y = 1,20x + 79,6
|
0,012
|
0,11
|
ß
|
0,411
|
HDL-C (mg/dl)
|
y = -0,82x + 92,8
|
0,094
|
-0,31
|
ß*
|
0,018
|
LDL (mg/dl)
|
y = 0,08x + 122,7
|
0,000
|
0,02
|
ß
|
0,909
|
CRP (mg/l)
|
y = 0,33x - 10,1
|
0,153
|
0,39
|
ß*
|
0,003
|
SBP (mmHg)
|
y = 0,42x + 111,1
|
0,013
|
0,12
|
ß
|
0,376
|
* a p-value < 0,05 was considered significant
|
All of the study participants were obese, independent of the obesity index measured (Table 4). Consequently, there was an alarming high prevalence of morbidity given that more than 80% of subjects were insulin resistant, about 40% had dyslipidemia, three quarter showed signs of inflammation and nearly half of them suffered from MetS (Table 4). Comparing subjects with MetS to subjects without MetS, there were no significant differences in age and obesity indices except AC and AC/Ht, who were different on a weak significant level. Mean values of Glucose, TG, HDL and SBP were significantly higher compared to subjects without MetS.
Table 4 Prevalence and descriptive characteristics of metabolic risk factors in the study population with and without MetS
|
All
|
Prevalence
|
MetS
|
Without-MetS
|
|
|
n
|
n
|
%
|
n
|
mean
|
±SD
|
n
|
mean
|
±SD
|
p
|
Age, years
|
|
|
|
27
|
42.2
|
11.6
|
34
|
41.1
|
12.6
|
|
AC, >88 cm
|
61
|
61
|
100
|
27
|
122.4
|
13.1
|
34
|
115.0
|
10.7
|
*
|
AC/Ht, >0,5 cm/cm
|
61
|
61
|
100
|
27
|
0.74
|
0.08
|
34
|
0.69
|
0.06
|
*
|
BMI, >30 kg/m²
|
61
|
61
|
100
|
27
|
40.1
|
5.8
|
34
|
38.0
|
4.1
|
|
%FM-BIA, >30%
|
61
|
61
|
100
|
27
|
50.7
|
4.5
|
34
|
48.9
|
4.4
|
|
%FM-DBA, >30%
|
61
|
61
|
100
|
27
|
49.8
|
4.7
|
34
|
48.5
|
4.9
|
|
|
|
|
|
|
|
|
|
|
|
|
Glucose >100 (mg/dl)
|
61
|
16
|
26.2
|
27
|
104.6
|
15.6
|
34
|
89.9
|
9.5
|
*
|
Insulin >25 (mU/l)
|
61
|
14
|
23.0
|
27
|
21.6
|
8.0
|
34
|
17.2
|
10.8
|
|
HOMA-IR >2,61
|
61
|
49
|
80.3
|
27
|
5.5
|
2.1
|
34
|
3.9
|
3.1
|
|
TG, >150 (mg/dl)
|
59
|
21
|
35.6
|
26
|
177.5
|
51.3
|
33
|
113.5
|
38.6
|
*
|
HDL-C < 50 (mg/dl)
|
59
|
28
|
47.5
|
27
|
47.1
|
11.4
|
33
|
57.3
|
12.7
|
*
|
LDL-C >130 mg/dl
|
59
|
26
|
44.1
|
25
|
126.7
|
25.9
|
34
|
126.7
|
26.2
|
|
CRP >3,0 (mg/l)
|
56
|
39
|
69.6
|
27
|
6.4
|
3.6
|
29
|
5.3
|
4.5
|
|
SBP >130 (mmHg)
|
61
|
33
|
54.1
|
27
|
140.7
|
14.9
|
34
|
124.1
|
15.2
|
*
|
MetS
|
61
|
27
|
44.3
|
|
|
|
|
|
|
|
Abbreviations: BMI. body mass index; AC. abdomen circumference; HC. hand circumference; Ht. height; %FM. percentage fat mass; BIA. bioimpedance analysis; DBA. Dahlmann-Body-Analysis; HOMA-IR. homeostasis model assessment of insulin resistance; TG. triglycerides; HDL. high-density lipoproteins; LDL. low-density lipoproteins; CRP. C-reactive protein; SBP. systolic blood pressure. * p > 0.05.
The associations between different obesity indices and the logarithmically transformed metabolic risk factors are shown in Table 5 as a matrix of absolute values of correlation coefficients. The HOMA variable is excluded as a non-independent variable calculated out of Glucose and Insulin. The correlation coefficients were tested to be different to zero as a proof of a significant relationship. This holds true for all obesity indices and the parameters Insulin, HDL and CRP. There are two exceptions, namely the relationship TG vs. AC/Ht and SBP vs. AC. The overall pattern of correlation coefficients reveals that none of the adiposity indices is of crucial advantage to detect metabolic risk factors. Notably, BMI and the %FM measurements (BIA and DBA) spread out a homogeneous picture. This impression is confirmed by the Friedman test. The test statistic is Q = 1.60 and the corresponding p-value is p = 0.81 and with that exceeds far the critical level of 0.05. The result gives sufficient evidence to conclude that there is no significant difference between the means of all obesity indices calculated out of all metabolic parameters, meaning that none of the obesity indices is of superior quality to detect MetS.
Table 5 Pearson correlation coefficients between anthropometric and metabolic risk factors, tested different to zero
|
BMI
|
AC
|
AC/Ht
|
%FM-BIA
|
%FM-DBA
|
CRP, log
|
*0.41
|
*0.45
|
*0.41
|
*0.51
|
*0.41
|
Glucose
|
0.10
|
0.18
|
0.26
|
0.07
|
0.07
|
Insulin, log
|
*0.44
|
*0.35
|
*0.33
|
*0.48
|
*0.48
|
TG, log
|
0.13
|
0.24
|
*0.34
|
0.19
|
0.14
|
HDL
|
*-0.43
|
*-0.37
|
*-0.44
|
*-0.36
|
*-0.31
|
LDL, log
|
0.00
|
-0.04
|
0.01
|
-0.07
|
0.04
|
SBP
|
0.17
|
*0.27
|
0.25
|
0.18
|
0.12
|
Mean
|
0.12
|
0.15
|
0.17
|
0.14
|
0.14
|
Abbreviations: BMI, body mass index; AC, abdomen circumference; Ht, height; %FM, percentage fat mass; BIA, bioimpedance analysis; DBA, Dahlmann-Body-Analysis; log: values are logarithmic transformed; * p < 0.05
The sensitivity, specificity, Youden index and the diagnostic odds ratio (DOR) of the systolic blood pressure and the metabolic risk factors to identify those classified as MetS positive are presented in Table 6.
Table 6 The sensitivity, specificity, Youden-Index and the diagnostic odds ratio of metabolic risk factors classifying individuals as MetS positive
Classsification based on:
|
Sensitivity (%)
|
Specificity (%)
|
Youden-Index (%)
|
DOR
|
Glucose >100 (mg/dl)
|
51.9
|
94.1
|
46.0
|
17.2
|
Insulin >25 (mU/l)
|
33.3
|
85.3
|
18.6
|
2.9
|
HOMA-IR >2,61
|
92.6
|
29.4
|
22.0
|
5.2
|
TG >150 (mg/dl)
|
69.2
|
90.9
|
60.1
|
22.5
|
HDL-C < 50 (mg/dl)
|
77.8
|
78.1
|
55.9
|
12.5
|
LDL-C >130 mg/dl
|
48.0
|
58.8
|
6.8
|
1.3
|
CRP >3,0 (mg/l)
|
77.8
|
37.9
|
15.7
|
2.1
|
SBP >130 (mmHg)
|
81.5
|
67.6
|
49.1
|
9.2
|
DOR: diagnostic odds ratio
HOMA-IR showed good sensitivity but poor specificity resulting in a Youden Index of 22.0%. The Youden Index of Glucose, TG, HDL and SBP were all above 45%. This corresponds to DOR values about 10 and greater, meaning that the chance of a positive result in individuals with MetS is 10 times greater than in individuals without MetS. The values of all other risk factors like Insulin, LDL-C and CRP were below 22% corresponding to DOR values <5.2.
Based on the same assumption of thresholds, the false-positive tests (FP) represent the false-positive test results in a cohort of individuals defined as MetS-negative. As depicted in Figure 2, almost 95% of MetS-negative subjects had a positive test result of at least one parameter, in the average 1.9. The candidates with the highest values >40% were CRP, HOMA and LDL, which are the parameters not being involved in the definition of MetS. Compared to results based on the MetS definition that is, based on the traits Glucose, TG, HDL, and SBP, the FP-rate was still 67.7% with an average of 1.0 parameter. That meets the criteria of subjects to be defined as metabolically healthy [24].
Figure 2
In Figure 3, the accuracy of adiposity indices with respect to the prediction of >2 component traits of MetS (elevated blood pressure, TG or glucose level) is compared by using plots of receiver operating curves (ROC). AUC values for the obesity indices BMI, AC/Ht and %FM-DBA are shown (0.63; 0.75; 0.65). The AUC values for AC and %FM-BIA were calculated as 0.70 for both parameters each.
Figure 3
AUC values of the different obesity indices were all in a similar range and reached values between 0.63 and 0.75. Furthermore, with the exception of AC/Ht (AUC = 0.75), all given AUC values fall below the discrimination level of 0.7, what means that they have a poor class separation capacity supporting the finding that they cannot provide an adequate power in the diagnose of MetS.