J Clin Epidemiol Vol. 43, No. I, pp. 21-34, 1990 Printed in Great Britain. All rights reserved
0895-4356/90 $3.00 + 0.00 Copyright C 1990 Pergamon Press plc
ANDROGENICITY IN RELATION TO BODY FAT DISTRIBUTION AND METABOLISM IN 3%YEAR-OLD WOMEN-THE EUROPEAN FAT DISTRIBUTION STUDY JACOB
C.
SEIDELL,’ MASSIMO CIGOLINI,’JADVIGACHARZEWSKA,~
BRITT-MARIEELLSINGER,~ GIUSEPPEDI BIASE,~PER BJ~RNTORP, JOSEPHG. A. J. HAUTVAST,’FRANCOCONTALDO,’ VIKTOR SZOSTAK~and LUDOVICOA. SCURO~ ‘Department of Human Nutrition, Agricultural University, PO Box 8129, 6700 EV Wageningen, The Netherlands, ‘Institute of Clinical Medicine, University of Verona, Italy, 3National Institute of Food Science, Warsaw, Poland, 4Department of Medicine I, University of Gothenburg, Sweden and 5Department of Internal Medicine and Metabolic Diseases, University of Naples, Italy (Receiced in revisedform
24 July 1989)
Abstract-We studied fat distributionand metabolic risk factors in 434 38-year old women selected from population registars in 5 cities in different parts of Europe. In the present study we focussed on the geographical variation in serum concentrations of free testosterone and its relation to measures of obesity, fat distribution and indicators of cardiovascular risk (serum lipids, insulin, and blood pressure). There were significant differences in free testosterone levels (F = 5.4, p < 0.001) with lowest levels in Polish women (mean + SEM: 1.56 + 0.08 pg/ml) and highest in women from Italy (2.07 + 0.12 pg/ml). In the pooled data, free testosterone levels were correlated with several anthropometric variables (strongest with subscapular/triceps ratio r = 0.27, with subscapular skinfold and waist/thigh circumference ratio Y = 0.25 p-values < 0.001). In addition, free testosterone was positively correlated with serum total cholesterol (r = 0.1 l), HDL/total cholesterol fraction (r = 0.12), serum insulin (r = 0.20) and diastolic blood pressure (r = 0.15). These associations remained significant after adjustment for body mass index and waist/thigh ratio (not for diastolic blood pressure) but were no longer significant after further adjustment for insulin levels. There were considerable differences in strength of the associations mentioned between the 5 centers. We conclude that degree of obesity, fat distribution and serum levels of free testosterone all, to a limited degree, contribute to the metabolic profile of randomly selected 38-year old women but that adjustment for such variables increses the differences in metabolic profiles between women from different centers of Europe. Obesity Androgens Fat distribution
Testosterone
Insulin
INTRODUCTION
Cholesterol
Blood pressure
141. Evans et al. [5] demonstrated that, in premenopausal women selected to cover a wide range of obesity, increased abdominal fat mass was associated with increased androgenicity (decreased serum levels of sex hormone binding globulin and increased percent free testosterone). These associations were independent of the degree of obesity. The investigators
Adipose tissue distribution in women is known to be related to a predisposition towards various important diseases such as coronary heart disease [l], stroke [l], non-insulin dependent diabetes mellitus [2, 31, gallbladder disease, menstrual disorders [2], and endometrial cancer 21
22
JACOBCSEIDELL~~
speculated that abdominal fat distribution could be one of many symptoms of a hyperandrogenic state which in itself is likely to cause metabolic aberrations. Indeed, subsequent work from the same group showed that the associations between fat distribution and both hepatic insulin clearance and peripheral insulin resistance were largely dependent on the androgenic activity of the serum [6]. Haffner et al. [7] confirmed that in Mexican-Americans and nonHispanic whites SHBG levels were negatively correlated with fasting glucose and insulin concentrations in serum, independently from indicators of body fatness and body fat distribution. We recently observed that women from southern parts of Europe tend to have higher BMIs and, apart from that, more abdominal fat distribution compared to women from northern parts of Europe [8]. In the present study we investigate the possible role of serum free testosterone in the explanation of the considerable variation of body fat distribution and metabolic risk factors in European women, all born in 1948, randomly selected from healthy populations in 5 different centers in Europe. POPULATION
AND METHODS
Populations
In 5 different centers in 4 European countries (Ede in The Netherlands, Warsaw in Poland, Gothenburg in Sweden, and Verona and Afracola (near Naples in Italy) a list of all women born in 1948 was obtained (voting lists or birth registers). From these lists a random selection was made, excluding all immigrants and those who had at least one parent who was an immigrant. An informative latter was send to the women and those who could be reached by telephone were asked to participate in the study. Those who refused to participate were asked to answer to items on a “non-responders questionnaire” which included items such as height, weight, socioeconomic status and chronic illness), Table 1 shows the number of participants and participation in the telephone questionnaire. In the present study we excluded all women who were pregnant, had diabetes mellitus, had a resection of both ovaries or used medicines known to affect serum lipids. For blood pressure analysis we excluded all those who use antihypertensive medication. Methods Anthropometry. All measurements were performed by one observer from each participating
al.
center. All observers (excluding the one from Naples) received a training in these measurements in The Netherlands and all worked according to a detailed working plan describing all the measurements. This working plan is available on request to one of the authors (JCS). All subjects were measured in the morning in the fasting state wearing only underwear. Height (to the nearest mm) and weight (to the nearest 0.5 kg) were recorded. Body mass index (BMI) was calculated as weight/height*. Skinfolds were measured in triplicate with a Harpender skinfold caliper and values were read when they started to stabilize (usually 2-3 seconds after applying the full pressure of the caliper jaws) and recorded to the nearest 0.2 mm. Of the 7 different skinfolds that were measured we present here only the results obtained for the triceps and subscapular skinfolds. Circumferences were measured in duplicate on subjects in standing position at the end of gentle aspiration. Of the 7 circumferences that were measured we concentrate in this paper on the waist circumference (midway between the lower rib margin and the iliac crest) the hip circumference (widest circumference over the great trochanters) and the thigh circumference (horizontal circumference at the level of the gluteal gold on the right thigh). More details about the anthropometric variables and socioeconomic and demographic characteristics of the population have been presented elsewhere [8,9]. Blood samples. 35 ml of blood was obtained in the morning after a fast of at least 10 hours by puncture of the anticubital vein. After 1 hour standing at room temperature sedimentation was measured and serum was obtained after centrifugation. The obtained serum was stored at -70°C and sent in the frozen state to Wageningen, The Netherlands, for laboratory analysis, total cholesterol [lo], HDL-cholesterol [ 1l] and triglycerides [ 121 were determined with semi-automized methods. Reproducibility of the laboratory with blind control sera is 2.3% (coefficient of variation) for HDL-cholesterol and 1.2% for total cholesterol. Accuracy in comparison with international control-sera is within 1%. Serum samples kept at - 70°C were sent to Verona, Italy, for analysis of total testosterone, free testosterone, and insulin. Total serum testosterone was measured by 3HRIA assay after diethyl-ether extraction (Test0 KR, Sorin Biomedical, Saluggia, Italy). Free testosterone was determined by a solid phase ‘Z51-RIA (Coat A-count, DPC, Los Angeles,
23
Androgenicity and Body Fat Distribution Table I. Details of selection and response in the participating European centers Population size n
Country (city) Sweden (Gothenburg) Poland (Warsaw) The Netherlands @de) Italy (Verona) Italy (Naples)t
Number of selected women Unreachable II n
400,000
140
I ,300,000
100
90,000
140
400,000
100
60,000
169
Non-participants telephone Participants n
yes
no
Percent
88
15*
4
82.2
92
8*
-
92.0
10
85
40*
5
64.4
3
87
4*
6
87.0
100
4l.t
28
59.2
33
-
*No significant differences in BMI and socioeconomic status (education, occupation, occupation of husband) between the participants and the non-participating telephone-responders. TCity of Afragola (near Naples): difference (p < 0.0001) in BMI between participants (average BMI: 27.8 kg/m*) and non-participating telephone responders (average BMI: 25.6 kg/m*) but no differences in socioeconomic
status.
Calif.). The intra-assay and inter-assay coefficients of variation were 4%. RIA insulin was measured by using Insik-5R (Sorin Biomedical, Saluggia, Italy). Both intra-assay and interassay coefficients of variation were 7.5%. Blood pressure. Blood pressure was measured with a sphygmomanometer in the sitting position after at least 30 minutes rest. Diastolic blood pressure was recorded at the disappearance of sounds (Korotkov phase V). Cuff-size was usually 12 x 22 cm but we switched to larger cuff-size appropriate for the arm circumference (this was extremely rare). The measurements were repeated with 5 minute intervals. The average of the measurements was used in the analysis. Statistical analysis. Pearson product-moment correlations and partial correlations were calculated using univariate and multiple linear regression. Deviations from normality of the distributions of the variables and the linearity of the relationships (analysis of residuals) were checked. Although some of the variables used in analysis were slightly skewed, transformation of these variables did not improve the fit of the regression models. We therefore present analysis with untransformed variables. Analysis of variance was performed for testing differences in means and regression coefficients between the centers. RESULTS
Table 1 shows the participation rate of the randomly selected women. The participation rate differed from 59.2% (Naples) to 92.0% (Poland). If we include the telephone responders the overall response rate was 92.9%. For these
women we have information on height, weight, indicators of social class. The common reason for not participating was “no time” (because of work, small children at home or other practical reasons). In none of the centers was there a selection on the basis of indicators for social class and in only one center (Naples) was there a selection bias in BMI (the non-participants being less overweight compared to the participants). Table 2 shows means and standard errors of the anthropometric measurements and serum lipids in the 5 centers. Analysis of variance showed that there were differences in all anthropometric variables except the triceps skinfold. The largest variance was observed in height (the women in Naples being, on the average, 13 cm shorter than the women in Sweden and The Netherlands). The BMI was, even when the selection bias in Napes was taken into account (analysis not shown), highest in the women from Naples and, in addition, they had the largest circumferences, circumference ratios, skinfolds, and skinfold ratio. Table 3 shows that here were no significant differences between the centers in serum triglycerides the HDL-cholesterol. Also in separate t-tests for all pairs of centers there were no significant differences. The analysis of variance showed that there were significant differences in total cholesterol values between the centers. In separate t-tests it was shown that this was due to the lower values in Naples compared to all other centers (all t-tests: p < 0.0001). There were no differences in total cholesterol between the other centers. The HDL/total cholesterol fraction was highest in Naples and lowest in Poland. There were also striking differences in fasting insulin levels with the higher levels in
20.2 17.0 1.25
80.2 99.6 62.4 0.81 1.29
168 66.1 23.1
0.7 0.7 0.03
1.0 0.8 0.6 0.006 0.01
0.7 1.0 0.4
SEM
Sweden Mean
22.3 18.9 1.25
77.6 97.4 57.7 0.80 1.35
160 61.5 23.9
Mean
Poland
21.8 16.8 1.42
77.1 98.1 58.3 0.79 1.33
168 66.8 23.3
22.3 20.3 1.18
77.5 98.7 56.7 0.78 1.37
160 65.1 24.1
0.8 1.0 0.04
1.0 0.8 0.5 0.007 0.01
0.7 1.2 0.4
SEM
Verona Mean
23.1 28.0 0.88
83.8 102.2 59.1 0.82 1.42
155 67.3 27.8
Mean
0.7 1.1 0.02
0.8 0.9 0.5 0.004 0.01
0.5 1.2 0.5
SEM
Naples
2.3 (NS) 26.9* 23.8*
9.9* 5.1* 16.9* 8.0’ 21.9*
80.5* 5.3* 23.9*
F-value
1.6 0.9
0.02
0.56
123.3 78.4
0.54 0.08
118.8 75.9
0.48
15.96 1.56
1.00 1.40 5.66 0.25
Mean
Poland
1.3 1.0
0.02
0.58 0.08
0.05 0.03 0.09 0.007
SEM
116.7 72.6
0.50
15.18 1.91
0.88 1.42 5.58 0.26
Mean
1.3 1.1
0.02
0.71 0.10
004 0.03 0.11 0.006
SEM
Netherlands
*F-value for the analysis of variance yielded a p-value 0.05.
Systolic Diastolic
Blood pressure (mm Hg)
Insulin (UI/ml) Free testosterone (pg/ml) Total testosterone (nglml)
0.05 0.04 0.08 0.009
SEM
10.42 1.64
0.87 1.48 5.36 0.28
Serum lipidr (mmol/l) Triglycerides HDL-cholesterol Total cholesterol HDL/tot. cholesterol
Hormones
Mean
Variable
Sweden
120.2 80.6
0.49
10.52 2.07
0.95 1.47 5.29 0.28
Mean
1.8 1.1
0.01
0.32 0.12
0.05 0.04 0.09 0.007
SEM
Verona
127.6 78.0
0.47
12.72 2.06
0.99 1.38 4.86 0.29
Mean
1.6 1.3
0.02
0.69 0.12
0.05 0.03 0.09 0.006
SEM
Naples
8.1* 7.5*
3.4*
18.5’ 5.4*
1.5 (NS) 1.9 (NS) 11.8* 4.6*
F-value
Table 3. Description of some metabolic variables (serum lipids, insulin, androgens and blood pressure) in European women born in 1948. Mean and SEM
0.6 0.7 0.05
0.8 0.7 0.5 0.005 0.01
0.6 1.1 0.3
SEM
Netherlands Mean
(NS) = p > 0.05.
0.8 0.9 0.03
0.3 0.9 0.6 0.005 0.01
0.5 1.2 0.4
SEM
variables in European women born in 1948; Mean and SEM
*F-value for the analysis of variance yielded a p-value t0.001.
Triceps (mm) Subscapular (mm) Triceps/subscapular
Skinfolh
Circumferences Waist (cm) Hip (cm) Thigh (cm) Waist/hip Waist/thigh
Height (cm) Weight (kg) BMI (kg/m’)
Variable
Table 2. Description of anthropometric
Androgenicity and Body Fat Distribution BMI
WTR
1kg/m*)
1451
S PI NL I
SPINL
I
(Vrl (Npl
I I (VrlINpl
Fig. 1. Average waist/thigh circumference ratios (WTR) and body mass index (BMI) in randomly selected 38-year shown by bars). women (SEM old European S = Gothenburg in Sweden; PI = Warsaw in Poland; Nl = Wageningen in The Netherlands; I(Vr) = Verona in Italy; I(Np) = Naples in Italy.
Poland and The Netherlands and relatively low values in Swedish and Italian women. Free testosterone levels were highest in Italian women whereas total testosterone levels were among the lowest. Systolic and diastolic blood pressure were highest in Swedish and Italian women and lowest in women from Poland and The Netherlands. The differences in BMI, waist/thigh ratio and free testosterone levels are illustrated in Figs 1 and 2. Table 4 shows the correlations of serum free testosterone and anthropometric variables in the 5 centers and the women of all centers combined. In the pooled material the highest correlations were observed with indices of body
fat distribution (especially with subscapular/ triceps ratio and waist/thigh circumference ratio) and subscapular skinfold and BMI. When looking at the countries separately we found considerable inconsistencies in the strength of associations but also the magnitude of slopes of the regression lines between free testosterone and anthropometric variables. In the cohorts from Sweden and Verona we did not seen any significant association whereas associations between free testosterone and subscapular skinfold, skinfold ratio and circumference ratios were similar in strength in the women from Poland, The Netherlands, and Naples (with the exception for the associations with circumferences in The Netherlands and circumference ratios in Naples). Figure 3 illustrates the associations of free testosterone and waist/thigh circumference ratio. In all centers, except Sweden, the highest free-testosterone levels were observed in the women with the highest waist/thigh ratios. Table 5 shows the associations of free testosterone and metabolic risk factors. In the pooled material significant positive associations were observed with total cholesterol, serum insulin and diastolic blood pressure. A negative association was found between free testosterone and the HDL-cholesterol fraction. The associations are quite weak and it is therefore not surprising that we find inconsistencies in the strength of associations and the regression lines between the different centers. Most striking is the negative association between free testosterone and diastolic blood pressure in the Swedish sample. It seems thus that in the Swedish cohort free
free/total
free-T
25
5.00
T
1
3.00
S
PL
NL
I(Vr)
I(NpI
0
S
PI
NI
IF.+)
I (Np)
Fig. 2. Average serum free testosterone levels (free-T) and free: total testosterone ratio (free/total T) in randomly selected 38-year old European women (SEM shown by bars). S = Gothenburg in Sweden; Pl = Warsaw in Poland; Nl = Wageningen in The Netherlands; I(Vr) = Verona in Italy; I(Np) = Naples in Italy.
26
JACOBC. SEIDELLet al. Table 4. Correlations of free testosterone with anthropometric variables in European women born in 1948 in different parts of Europe Independent variable
Dependent variable
Anthropometric variables
Free testosterone
Center Sweden
Poland
Netherlands
Verona
Naples
0.02 -0.13 -0.16
0.13 0.40** 0.36**
-0.15 0.01 0.09
0.06 0.13 0.11
0.06 0.35** 0.34**
Skinfolds Triceps skinfold Subscapular skinfold Subscapular/triceps
-0.03 0.03 0.09
0.36** 0.43** 0.27*
-0.02 0.27; 0.37**
0.05 0.08 0.09
0.05 0.30** 0.41**
0.10* 0.25’; 0.27**
Circumferences Waist circumference Hip circumference Thigh circumference
-0.06 -0.19 -0.16
0.44**
0.15 -0.05 -0.10
0.16 0.11 0.11
0.32** 0.29** 0.19
0.21** 0.15** 0.03
0.14 0.15
0.09 0.17
0.18** 0.25**
Height Weight BMI
Waist/hip ratio Waist/thigh ratio
0.32** 0.24* 0.39** 0.40**
0.13 0.09
0.28** 0.32**
All -0.05
0.19** 0.22**
*p < 0.05; **p < 0.01.
testosterone is neither a determinant of body fat distribution, nor of metabolic risk factors. Table 6 shows the independent contribution of free testosterone and degree of overweight (BMI) to measures of fat distribution in the pooled data. Both BMI and, to a lesser degree, free testosterone were independently positively associated with all indicators of fat distribution. Waist circumference was much more strongly associated to BMI compared to the other indicators of fat distribution. In Tables 7(a), (b) and (c) it is shown that free testosterone has a positive association with total cholesterol and triglycerides independently from BMI and 3 indicators of fat distribution. All 3 indicators are negatively associated to HDLcholesterol concentrations and the HDL/total cholesterol ratio and positively associated with triglyceride concentrations independently from BMI and free testosterone concentrations. The adjusted associations of waist/hip ratio and
waist circumference to serum cholesterol are significant. This is not so for waist/thigh ratio. The 3 models with different indicators of fat distribution do not show appreciable differences in the percentage of explained variance. Tables 8(a), (b) and (c) show the same regression models as shown in Tables 7(a)-(c) but now insulin has been added as an additional independent variable. The partial regression coefficients of free testosterone levels were now reduced to non-significant levels except for the association between free testosterone and triglycerides when adjusted for waist/hip ratio or waist circumference. Insulin was independently associated with all serum lipids and lipoproteins. No indicator of fat distribution was associated to serum cholesterol in these models and only waist/hip and waist/ thigh ratios, but not waist circumference alone, to serum triglycerides. All indicators of fat distribution remained negatively associated to HDLcholesterol and the HDL/total cholesterol ratio.
free-T
1.6
0
LM H Sweden
L M H Poland
L M H Netherlands
LMH (Verona)
=WTR(1.30
y
:1.30 WlR(lLo
;
L
M H Italy (Naples)
:WTR>l.40
Fig. 3. Average serum free testosterone levels (free-T) in categories of waist/thigh circumference ratio (WTR) in randomly selected 38-year old European women (SEM shown by bars).
Androgenicity and Body Fat Distribution
21
Table 5. Correlations of free testosterone (free T) with metabolic variables (triglycerides, HDL-cholesterol, total cholesterol, and HDL/total cholesterol ratio, insulin and blood pressure) in European women born in 1948, in different parts of Europe Independent variable
Dependent variable
Metabolic variables
Free testosterone
Center Sweden
Poland
Netherlands
Verona
Naples
All
0.21* 0.08 -0.05
0.23* -0.15 -0.26’
0.22* -0.01 -0.20
0.11* -0.06 -0.12**
Serum lipids
Total cholesterol HDL-cholesterol HDL/total cholesterol
0.21* -0.04 -0.03 -0.20 -0.11 -0.13
Serum insulin
Insulin
-0.07
0.57**
0.19
0.13
0.34**
0.20**
-0.23’ -0.26*
0.16 0.23*
0.15 0.23*
0.19 0.19
0.05 0.16
0.09 0.15**
Blood pressure
Systolic Diastolic *p < 0.05: **p < 0.01.
Table 6. Independent contributions of degree of androgenicity (free testosterone) and BMI to measures of fat distribution in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient P
Dependent variables
Independent variables
Waist/thigh
Constant BMI Free Tt Multiple r2
1.050 0.011 8.9*** 0.021 4.2*‘* 0.22 (F = 4.20, p < O.OOOl)t
Waist/hip
Constant BMI Free T Multiple r2
0.0654 0.006 0.006 0.21 (F = 4.73, p < o.OOol)t
Subscapular/ triceps
Constant BMI Free T Multiple r2
0.080 9.3*** 0.030 4 8*** 0.064 0.24 (F = 8.30, p < O.OOOl)t ’
Waist
Constant BMI Free T Multiple r2
1.050 32.0*** 1.825 2.4’ 0.575 0.72 (F = 4.50, p < O.OOOl)t
t-Value
9 5*** 2:3*
*p < 0.05; ***p < 0.0001. tFree T = serum free testosterone.
Tables 9(a), (b) and (c) show that free testosterone was, in multiple regression analysis, independently associated with serum insulin but not with blood pressure. BMI was independently associated with serum insulin (not when adjusted for waist circumference [Table 9(c)] and blood pressure whereas body fat distribution was not related to blood pressure in multiple regression analysis. Only waist circumference was associated with serum insulin levels. Insulin levels were not associated with blood pressure independently of BMI, free testosterone, and indicators of fat distribution (not shown). DISCUSSION
This study shows that in randomly selected premenopausal women all of the same age, free
testosterone levels in serum are lowest in the north of Europe (Sweden, Poland, and The Netherlands) and highest in Italian women. This trend has some overlap with differences in fat distribution in the same centers. As we have discussed elsewhere, it is difficult to assess the optimal indicator of fat distribution [9]. The results from the multiple linear regression analysis from which 3 different indicators can be compared with respect to their associations to metabolic risk factors suggest the same (Tables 7-9). Waist circumference alone is more strongly associated to serum cholesterol, the HDL/total cholesterol ratio, and serum insulin compared to waist/thigh and waist/hip ratios. The waist/thigh ratio, on the other hand, shows the strongest partial associations with HDLcholesterol and triglycerides. The differences in
28
JACOBC. SEIDELLet al.
strength of associations of waist/hip ratio, waist/thigh ratio, and waist circumference with metabolic risk factors were not very large and the proportion of explained variance in metabolic risk factors does not change very
much when different indicators of fat distribution are exchanged for each other. Correlational analysis in the pooled material showed significant associations between free testosterone levels with indicators of fatness and body
Table 7(a). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist/thigh), and BMI to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient P
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist/thigh BMI Multiple r*
4.62 0.10 0.41 -0.0003 0.02 (F = 5.36, p < O.OOOl)t
2.16* 0.88 -0.03
HDL cholesterol (mmol/l)
Constant Free testosterone Waist/thigh BMI Multiple r2
2.46 0.008 -0.52 -0.014 0.09 (F = 1.39, p = 0.14)?
0.49 -3.42*** -3..52”*
HDL/total cholesterol
Constant Free testosterone Waist/thigh BMI Multiple r*
0.49 -0.003 -0.11 -0.003 0.09 (F = 4.88, p < O.OOOl)t
-0.85 -3.15** -3.38***
Triglycerides (mmol/l)
Constant Free testosterone Waist/thigh BMI Multiple r*
-0.61 0.054 0.76 0.017 0.12 (F = 0.75, p = 0.74)t
*p < 0.05; **p < 0.01; ***p
t-Value
2.43* 3.56*** 3.06**
< 0.001.
tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers. Table 7(b). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist/hip), and BMI to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient P
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist/hip BMI Multiple r*
3.77 0.10 1.96 -0.007 0.03 (F = 5.16, p < O.OOOl)t
2.15* 2.05; -0.54
HDL cholesterol (mmol/l)
Constant Free testosterone Waist/hip BMI Multiple r*
2.60 0.003 - 1.05 -0.014 0.09 (F = 0.83, p = 0.65)t
0.18 -3.39*** -3.41***
HDL/total cholesterol
Constant Free testosterone Waist/hip BMI Multiple r*
0.56 - 0.0036 -0.27 -0.003 0.15 (F = 3.99, p < o.oool)t
-1.07 -4.01*** - 2.94**
Triglycerides (mmol/l)
Constant Free testosterone Waist/hip BMI Multiple r2
-0.64 0.063 1.27 0.019 0.11 (F = 0.77, p = 0.72)t
*p < 0.05; **p < 0.01; ***p
t-Value
2.85** 2.87** 3.21**
< 0.001.
tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers.
Androgenicity and Body Fat Distribution
29
Table 7(c). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist circumference), and BMI to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient B
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist BMI Multiple r2
4.18 0.10 0.026 -0.044 0.03 (F = 5.03, p < o.oOOl)t
2.08* 2.69*’ -2.08’
HDL cholesterol (mmol/l)
Constant Free testosterone Waist BMI Multiple r2
2.21 0.002 -0.009 -0.003 0.08 (F = 1.17, p = 0.29)t
0.13 -2.85** -0.45
HDL/total cholesterol
Constant Free testosterone Waist BMI Multiple r2
0.48 -0.004 - 0.003 -0.001 0.11 (F = 3.35, p < o.oool)t
- 1.04 -4.11*** -0.72
Triglycerides (mmol/l)
Constant Free testosterone Waist BMI Multiple r’
-0.20 0.063 0.012 0.004 0.10 (F = 0.79, p = 0.70)t
t-Value
2.87** 2.60** 0.42
*p < 0.05; **p < 0.01; ***p < 0.001. tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers.
Table 8(a). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist/thigh), BMI, and fasting insulin concentrations to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient B
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist/thigh BMI Insulin Multiple r2
4.73 0.08 0.27 -0.01 0.03 0.06 (F = 4.73, p < O.OOOl)t
1.74 0.59 -1.01 4.14***
HDL cholesterol (mmol/l)
Constant Free testosterone Waist/thigh BMI Insulin Multiple r2
2.44 0.01 -0.49 -0.01 -0.007 0.11 (F = 0.94, p = 0.54)?
0.77 - 3.25** -2.83** -2.55*
HDL/total cholesterol
Constant Free testosterone Waist/thigh BMI Insulin Multiple r2
0.49 -0.001 -0.09 - 0.002 - 0.002 0.14 (F = 14.69, p < O.OOOl)t
-0.32 -2.86** -2.18** - 5.03***
Triglycerides (mmol/l)
Constant Free testosterone Waist/thigh BMI Insulin Multiple r’
-0.54 0.04 0.67 0.009 0.02 0.19 (F = 0.77, p = 0.75)t
f-Value
1.86 3.25** 1.69 5.95***
*p < 0.05; **p < 0.01; ***p < 0.001. tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers.
30
JACOBC. SEIDELLet al. Table 8(b). Independent contributions of degree. of androgenicity (free testosterone), body fat distribution (waist/hip), BMI, and fasting insulin concentrations to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient B
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist/hip BMI Insulin Multiple r*
3.89 0.08 1.71 -0.02 0.03 0.06 (F = 3.91, p < O.OOOl)t
1.69 1.82 - 1.50 4.08***
HDL cholesterol (mmol/l)
Constant Free testosterone Waist/hip BMI Insulin Multiple r2
2.57 0.007 -1.00 -0.01 -0.007 0.11 (F = 0.53, p = 0.95)?
0.48 - 3.24** -2.72*+ -2.57”
HDL/total cholesterol
Constant Free testosterone Waist/hip BMI Insulin Multiple r*
0.55 -0.002 -0.25 - 0.002 -0.003 0.16 (F = 2.63, p < O.OOOl)t
-0.49 - 3.79*** -1.75 -5.03***
Triglycerides (mmol/l)
Constants Free testosterone Waist/hip BMI Insulin Multiple r*
-0.56 0.05 1.10 0.010 0.02 0.18 (F = 0.81, p < 0.70)t
f-Value
2.24+ 2.59* 1.83 5.99***
***p < 0.001. *p < 0.05; **p < 0.01; tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers.
Table 8(c). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist circumference), BMI, and fasting insulin concentrations to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient B
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist BMI Insulin Multiple r*
4.36 0.08 0.02 -0.04 0.03 0.06 (F = 3.78, p < O.OOOl)t
1.69 2.01* -2.15* 3.78***
HDL cholesterol (mmol/l)
Constant Free testosterone Waist BMI Insulin Multiple r*
2.17 0.006 -0.008 -0.003 -0.006 0.10 (F = 0.85, p = 0.65)t
0.37 -2.40; -0.43 - 2.30’
HDL/total cholesterol
Constant Free testosterone Waist BMI Insulin Multiple r*
0.46 -0.019 -0.002 0.001 -0.003 0.15 (F = 2.30, p < 0.0013)t
-0.57 - 3.32** 0.79 -4.59***
Triglycerides (mmol/l)
Constant Free testosterone Waist BMI Insulin Multiple r2
-0.08 0.05 0.007 0.004 0.02 0.17 (F = 0.85, p = 0.65)-f
t-Value
2.35* 1.63 0.38 5.76+++
*p c 0.05; **p < 0.01; ***p < 0.001. tf-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slope and/or intercepts differ beyond chance between the centers.
Androgenicity and Body Fat Distribution
fat distribution and a number of metabolic variables (total cholesterol, fasting insulin, and diastolic blood pressure). We observed interactions in the relationships between the independent variables and dependent variables in the separate centers when we performed analysis of variance of the regression coefficients. Differences in linear relationships between the centers cannot be readily explained but it must be noted that the variation in all measurements in the pooled material is very much larger than in the separate centers and the associations, even in the pooled material, were weak. The reason why
31
in the Swedish cohort, for example, no associations could be found between free testosterone and fat distribution and metabolic risk factors may be because the Swedish women were all at the lower end of the range of testosterone levels and fat distribution. Performing analysis in the pooled data which comprises the total range of fat distribution and metabolic risk factors in European women is more likely to give a reliable estimate of the “true” relationships than analysis in all centers separately. With regard to the testosterone values we must emphasize that blood samples were taken at various phases of
Table 9(a). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist/thigh), and BMI to insulin and blood pressure in European women born in 1948 (oooled data). Results of multinle linear regression Regression coefficient P
Dependent variables
Independent variables
Serum insulin (ID/l)
Constant Free testosterone Waist/thigh BMI Multiple r*
-3.10 0.66 4.24 0.37 0.12 (F = 8.90, p < O.OOOl)t
2.24’ 1.49 4.99***
Diastolic blood pressure (mmHg)
Constant Free testosterone Waist/thigh BMI Multiple r*
53.18 0.79 1.77 0.82 0.12 (F = 2.66, p < 0.0005)t
1.52 0.35 6.21***
Systolic blood pressure (mmHg)
Constant Free testosterone Waist/thigh BMI Multiple r*
79.96 -0.08 9.71 1.17 0.13 (F = 1.661, p < 0.064)t
-0.11 1.40 6.38***
f-Value
*p < 0.05; **p < 0.01; ***p < 0.001. tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers. Table 9(b). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist/hip), and BMI to insulin and blood pressure in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient B
Dependent variables
Independent variables
Serum insulin (ID/l)
Constant Free testosterone Waist/hip BMI Multiple r*
-3.79 0.70 7.89 0.37 0.11 (F = 8.94, p < O.OOOl)t
2.42* 1.35 4.95*+*
Diastolic blood pressure (mmHg)
Constant Free testosterone Waist/hip BMI Multiple r*
52.98 0.81 1.77 0.83 0.12 (F = 2.84, p c 0.0002)t
1.58 0.31 6.16***
Systolic blood pressure (mmHg)
Constant Free testosterone Waist/hip BMI Multiple r2
75.25 -0.007 22.86 1.14 0.13 (F = 1.51, p c 0.093)-f
f-Value
-0.01 1.61 6.19***
*p < 0.05; **p < 0.01; ***p < 0.001. TF-values calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers.
32
JACOBC. SEIDELLef al. Table 9(c). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist circumference), and BMI to insulin and blood pressure in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient P
Dependent variables
Independent variables
Serum insulin (ID/l)
Constant Free testosterone Waist BMI Multiple r2
-5.92 0.62 0.22 0.024 0.14 (F = 9.79, p < o.oOOl)t
2.18* 3.71*** 0.19
Diastolic blood pressure (mmHg)
Constant Free testosterone Waist BMI Multiple rz
51.31 0.76 0.11 0.64 0.12 (F=3.19, p
0895-4356/90 $3.00 + 0.00 Copyright C 1990 Pergamon Press plc
ANDROGENICITY IN RELATION TO BODY FAT DISTRIBUTION AND METABOLISM IN 3%YEAR-OLD WOMEN-THE EUROPEAN FAT DISTRIBUTION STUDY JACOB
C.
SEIDELL,’ MASSIMO CIGOLINI,’JADVIGACHARZEWSKA,~
BRITT-MARIEELLSINGER,~ GIUSEPPEDI BIASE,~PER BJ~RNTORP, JOSEPHG. A. J. HAUTVAST,’FRANCOCONTALDO,’ VIKTOR SZOSTAK~and LUDOVICOA. SCURO~ ‘Department of Human Nutrition, Agricultural University, PO Box 8129, 6700 EV Wageningen, The Netherlands, ‘Institute of Clinical Medicine, University of Verona, Italy, 3National Institute of Food Science, Warsaw, Poland, 4Department of Medicine I, University of Gothenburg, Sweden and 5Department of Internal Medicine and Metabolic Diseases, University of Naples, Italy (Receiced in revisedform
24 July 1989)
Abstract-We studied fat distributionand metabolic risk factors in 434 38-year old women selected from population registars in 5 cities in different parts of Europe. In the present study we focussed on the geographical variation in serum concentrations of free testosterone and its relation to measures of obesity, fat distribution and indicators of cardiovascular risk (serum lipids, insulin, and blood pressure). There were significant differences in free testosterone levels (F = 5.4, p < 0.001) with lowest levels in Polish women (mean + SEM: 1.56 + 0.08 pg/ml) and highest in women from Italy (2.07 + 0.12 pg/ml). In the pooled data, free testosterone levels were correlated with several anthropometric variables (strongest with subscapular/triceps ratio r = 0.27, with subscapular skinfold and waist/thigh circumference ratio Y = 0.25 p-values < 0.001). In addition, free testosterone was positively correlated with serum total cholesterol (r = 0.1 l), HDL/total cholesterol fraction (r = 0.12), serum insulin (r = 0.20) and diastolic blood pressure (r = 0.15). These associations remained significant after adjustment for body mass index and waist/thigh ratio (not for diastolic blood pressure) but were no longer significant after further adjustment for insulin levels. There were considerable differences in strength of the associations mentioned between the 5 centers. We conclude that degree of obesity, fat distribution and serum levels of free testosterone all, to a limited degree, contribute to the metabolic profile of randomly selected 38-year old women but that adjustment for such variables increses the differences in metabolic profiles between women from different centers of Europe. Obesity Androgens Fat distribution
Testosterone
Insulin
INTRODUCTION
Cholesterol
Blood pressure
141. Evans et al. [5] demonstrated that, in premenopausal women selected to cover a wide range of obesity, increased abdominal fat mass was associated with increased androgenicity (decreased serum levels of sex hormone binding globulin and increased percent free testosterone). These associations were independent of the degree of obesity. The investigators
Adipose tissue distribution in women is known to be related to a predisposition towards various important diseases such as coronary heart disease [l], stroke [l], non-insulin dependent diabetes mellitus [2, 31, gallbladder disease, menstrual disorders [2], and endometrial cancer 21
22
JACOBCSEIDELL~~
speculated that abdominal fat distribution could be one of many symptoms of a hyperandrogenic state which in itself is likely to cause metabolic aberrations. Indeed, subsequent work from the same group showed that the associations between fat distribution and both hepatic insulin clearance and peripheral insulin resistance were largely dependent on the androgenic activity of the serum [6]. Haffner et al. [7] confirmed that in Mexican-Americans and nonHispanic whites SHBG levels were negatively correlated with fasting glucose and insulin concentrations in serum, independently from indicators of body fatness and body fat distribution. We recently observed that women from southern parts of Europe tend to have higher BMIs and, apart from that, more abdominal fat distribution compared to women from northern parts of Europe [8]. In the present study we investigate the possible role of serum free testosterone in the explanation of the considerable variation of body fat distribution and metabolic risk factors in European women, all born in 1948, randomly selected from healthy populations in 5 different centers in Europe. POPULATION
AND METHODS
Populations
In 5 different centers in 4 European countries (Ede in The Netherlands, Warsaw in Poland, Gothenburg in Sweden, and Verona and Afracola (near Naples in Italy) a list of all women born in 1948 was obtained (voting lists or birth registers). From these lists a random selection was made, excluding all immigrants and those who had at least one parent who was an immigrant. An informative latter was send to the women and those who could be reached by telephone were asked to participate in the study. Those who refused to participate were asked to answer to items on a “non-responders questionnaire” which included items such as height, weight, socioeconomic status and chronic illness), Table 1 shows the number of participants and participation in the telephone questionnaire. In the present study we excluded all women who were pregnant, had diabetes mellitus, had a resection of both ovaries or used medicines known to affect serum lipids. For blood pressure analysis we excluded all those who use antihypertensive medication. Methods Anthropometry. All measurements were performed by one observer from each participating
al.
center. All observers (excluding the one from Naples) received a training in these measurements in The Netherlands and all worked according to a detailed working plan describing all the measurements. This working plan is available on request to one of the authors (JCS). All subjects were measured in the morning in the fasting state wearing only underwear. Height (to the nearest mm) and weight (to the nearest 0.5 kg) were recorded. Body mass index (BMI) was calculated as weight/height*. Skinfolds were measured in triplicate with a Harpender skinfold caliper and values were read when they started to stabilize (usually 2-3 seconds after applying the full pressure of the caliper jaws) and recorded to the nearest 0.2 mm. Of the 7 different skinfolds that were measured we present here only the results obtained for the triceps and subscapular skinfolds. Circumferences were measured in duplicate on subjects in standing position at the end of gentle aspiration. Of the 7 circumferences that were measured we concentrate in this paper on the waist circumference (midway between the lower rib margin and the iliac crest) the hip circumference (widest circumference over the great trochanters) and the thigh circumference (horizontal circumference at the level of the gluteal gold on the right thigh). More details about the anthropometric variables and socioeconomic and demographic characteristics of the population have been presented elsewhere [8,9]. Blood samples. 35 ml of blood was obtained in the morning after a fast of at least 10 hours by puncture of the anticubital vein. After 1 hour standing at room temperature sedimentation was measured and serum was obtained after centrifugation. The obtained serum was stored at -70°C and sent in the frozen state to Wageningen, The Netherlands, for laboratory analysis, total cholesterol [lo], HDL-cholesterol [ 1l] and triglycerides [ 121 were determined with semi-automized methods. Reproducibility of the laboratory with blind control sera is 2.3% (coefficient of variation) for HDL-cholesterol and 1.2% for total cholesterol. Accuracy in comparison with international control-sera is within 1%. Serum samples kept at - 70°C were sent to Verona, Italy, for analysis of total testosterone, free testosterone, and insulin. Total serum testosterone was measured by 3HRIA assay after diethyl-ether extraction (Test0 KR, Sorin Biomedical, Saluggia, Italy). Free testosterone was determined by a solid phase ‘Z51-RIA (Coat A-count, DPC, Los Angeles,
23
Androgenicity and Body Fat Distribution Table I. Details of selection and response in the participating European centers Population size n
Country (city) Sweden (Gothenburg) Poland (Warsaw) The Netherlands @de) Italy (Verona) Italy (Naples)t
Number of selected women Unreachable II n
400,000
140
I ,300,000
100
90,000
140
400,000
100
60,000
169
Non-participants telephone Participants n
yes
no
Percent
88
15*
4
82.2
92
8*
-
92.0
10
85
40*
5
64.4
3
87
4*
6
87.0
100
4l.t
28
59.2
33
-
*No significant differences in BMI and socioeconomic status (education, occupation, occupation of husband) between the participants and the non-participating telephone-responders. TCity of Afragola (near Naples): difference (p < 0.0001) in BMI between participants (average BMI: 27.8 kg/m*) and non-participating telephone responders (average BMI: 25.6 kg/m*) but no differences in socioeconomic
status.
Calif.). The intra-assay and inter-assay coefficients of variation were 4%. RIA insulin was measured by using Insik-5R (Sorin Biomedical, Saluggia, Italy). Both intra-assay and interassay coefficients of variation were 7.5%. Blood pressure. Blood pressure was measured with a sphygmomanometer in the sitting position after at least 30 minutes rest. Diastolic blood pressure was recorded at the disappearance of sounds (Korotkov phase V). Cuff-size was usually 12 x 22 cm but we switched to larger cuff-size appropriate for the arm circumference (this was extremely rare). The measurements were repeated with 5 minute intervals. The average of the measurements was used in the analysis. Statistical analysis. Pearson product-moment correlations and partial correlations were calculated using univariate and multiple linear regression. Deviations from normality of the distributions of the variables and the linearity of the relationships (analysis of residuals) were checked. Although some of the variables used in analysis were slightly skewed, transformation of these variables did not improve the fit of the regression models. We therefore present analysis with untransformed variables. Analysis of variance was performed for testing differences in means and regression coefficients between the centers. RESULTS
Table 1 shows the participation rate of the randomly selected women. The participation rate differed from 59.2% (Naples) to 92.0% (Poland). If we include the telephone responders the overall response rate was 92.9%. For these
women we have information on height, weight, indicators of social class. The common reason for not participating was “no time” (because of work, small children at home or other practical reasons). In none of the centers was there a selection on the basis of indicators for social class and in only one center (Naples) was there a selection bias in BMI (the non-participants being less overweight compared to the participants). Table 2 shows means and standard errors of the anthropometric measurements and serum lipids in the 5 centers. Analysis of variance showed that there were differences in all anthropometric variables except the triceps skinfold. The largest variance was observed in height (the women in Naples being, on the average, 13 cm shorter than the women in Sweden and The Netherlands). The BMI was, even when the selection bias in Napes was taken into account (analysis not shown), highest in the women from Naples and, in addition, they had the largest circumferences, circumference ratios, skinfolds, and skinfold ratio. Table 3 shows that here were no significant differences between the centers in serum triglycerides the HDL-cholesterol. Also in separate t-tests for all pairs of centers there were no significant differences. The analysis of variance showed that there were significant differences in total cholesterol values between the centers. In separate t-tests it was shown that this was due to the lower values in Naples compared to all other centers (all t-tests: p < 0.0001). There were no differences in total cholesterol between the other centers. The HDL/total cholesterol fraction was highest in Naples and lowest in Poland. There were also striking differences in fasting insulin levels with the higher levels in
20.2 17.0 1.25
80.2 99.6 62.4 0.81 1.29
168 66.1 23.1
0.7 0.7 0.03
1.0 0.8 0.6 0.006 0.01
0.7 1.0 0.4
SEM
Sweden Mean
22.3 18.9 1.25
77.6 97.4 57.7 0.80 1.35
160 61.5 23.9
Mean
Poland
21.8 16.8 1.42
77.1 98.1 58.3 0.79 1.33
168 66.8 23.3
22.3 20.3 1.18
77.5 98.7 56.7 0.78 1.37
160 65.1 24.1
0.8 1.0 0.04
1.0 0.8 0.5 0.007 0.01
0.7 1.2 0.4
SEM
Verona Mean
23.1 28.0 0.88
83.8 102.2 59.1 0.82 1.42
155 67.3 27.8
Mean
0.7 1.1 0.02
0.8 0.9 0.5 0.004 0.01
0.5 1.2 0.5
SEM
Naples
2.3 (NS) 26.9* 23.8*
9.9* 5.1* 16.9* 8.0’ 21.9*
80.5* 5.3* 23.9*
F-value
1.6 0.9
0.02
0.56
123.3 78.4
0.54 0.08
118.8 75.9
0.48
15.96 1.56
1.00 1.40 5.66 0.25
Mean
Poland
1.3 1.0
0.02
0.58 0.08
0.05 0.03 0.09 0.007
SEM
116.7 72.6
0.50
15.18 1.91
0.88 1.42 5.58 0.26
Mean
1.3 1.1
0.02
0.71 0.10
004 0.03 0.11 0.006
SEM
Netherlands
*F-value for the analysis of variance yielded a p-value 0.05.
Systolic Diastolic
Blood pressure (mm Hg)
Insulin (UI/ml) Free testosterone (pg/ml) Total testosterone (nglml)
0.05 0.04 0.08 0.009
SEM
10.42 1.64
0.87 1.48 5.36 0.28
Serum lipidr (mmol/l) Triglycerides HDL-cholesterol Total cholesterol HDL/tot. cholesterol
Hormones
Mean
Variable
Sweden
120.2 80.6
0.49
10.52 2.07
0.95 1.47 5.29 0.28
Mean
1.8 1.1
0.01
0.32 0.12
0.05 0.04 0.09 0.007
SEM
Verona
127.6 78.0
0.47
12.72 2.06
0.99 1.38 4.86 0.29
Mean
1.6 1.3
0.02
0.69 0.12
0.05 0.03 0.09 0.006
SEM
Naples
8.1* 7.5*
3.4*
18.5’ 5.4*
1.5 (NS) 1.9 (NS) 11.8* 4.6*
F-value
Table 3. Description of some metabolic variables (serum lipids, insulin, androgens and blood pressure) in European women born in 1948. Mean and SEM
0.6 0.7 0.05
0.8 0.7 0.5 0.005 0.01
0.6 1.1 0.3
SEM
Netherlands Mean
(NS) = p > 0.05.
0.8 0.9 0.03
0.3 0.9 0.6 0.005 0.01
0.5 1.2 0.4
SEM
variables in European women born in 1948; Mean and SEM
*F-value for the analysis of variance yielded a p-value t0.001.
Triceps (mm) Subscapular (mm) Triceps/subscapular
Skinfolh
Circumferences Waist (cm) Hip (cm) Thigh (cm) Waist/hip Waist/thigh
Height (cm) Weight (kg) BMI (kg/m’)
Variable
Table 2. Description of anthropometric
Androgenicity and Body Fat Distribution BMI
WTR
1kg/m*)
1451
S PI NL I
SPINL
I
(Vrl (Npl
I I (VrlINpl
Fig. 1. Average waist/thigh circumference ratios (WTR) and body mass index (BMI) in randomly selected 38-year shown by bars). women (SEM old European S = Gothenburg in Sweden; PI = Warsaw in Poland; Nl = Wageningen in The Netherlands; I(Vr) = Verona in Italy; I(Np) = Naples in Italy.
Poland and The Netherlands and relatively low values in Swedish and Italian women. Free testosterone levels were highest in Italian women whereas total testosterone levels were among the lowest. Systolic and diastolic blood pressure were highest in Swedish and Italian women and lowest in women from Poland and The Netherlands. The differences in BMI, waist/thigh ratio and free testosterone levels are illustrated in Figs 1 and 2. Table 4 shows the correlations of serum free testosterone and anthropometric variables in the 5 centers and the women of all centers combined. In the pooled material the highest correlations were observed with indices of body
fat distribution (especially with subscapular/ triceps ratio and waist/thigh circumference ratio) and subscapular skinfold and BMI. When looking at the countries separately we found considerable inconsistencies in the strength of associations but also the magnitude of slopes of the regression lines between free testosterone and anthropometric variables. In the cohorts from Sweden and Verona we did not seen any significant association whereas associations between free testosterone and subscapular skinfold, skinfold ratio and circumference ratios were similar in strength in the women from Poland, The Netherlands, and Naples (with the exception for the associations with circumferences in The Netherlands and circumference ratios in Naples). Figure 3 illustrates the associations of free testosterone and waist/thigh circumference ratio. In all centers, except Sweden, the highest free-testosterone levels were observed in the women with the highest waist/thigh ratios. Table 5 shows the associations of free testosterone and metabolic risk factors. In the pooled material significant positive associations were observed with total cholesterol, serum insulin and diastolic blood pressure. A negative association was found between free testosterone and the HDL-cholesterol fraction. The associations are quite weak and it is therefore not surprising that we find inconsistencies in the strength of associations and the regression lines between the different centers. Most striking is the negative association between free testosterone and diastolic blood pressure in the Swedish sample. It seems thus that in the Swedish cohort free
free/total
free-T
25
5.00
T
1
3.00
S
PL
NL
I(Vr)
I(NpI
0
S
PI
NI
IF.+)
I (Np)
Fig. 2. Average serum free testosterone levels (free-T) and free: total testosterone ratio (free/total T) in randomly selected 38-year old European women (SEM shown by bars). S = Gothenburg in Sweden; Pl = Warsaw in Poland; Nl = Wageningen in The Netherlands; I(Vr) = Verona in Italy; I(Np) = Naples in Italy.
26
JACOBC. SEIDELLet al. Table 4. Correlations of free testosterone with anthropometric variables in European women born in 1948 in different parts of Europe Independent variable
Dependent variable
Anthropometric variables
Free testosterone
Center Sweden
Poland
Netherlands
Verona
Naples
0.02 -0.13 -0.16
0.13 0.40** 0.36**
-0.15 0.01 0.09
0.06 0.13 0.11
0.06 0.35** 0.34**
Skinfolds Triceps skinfold Subscapular skinfold Subscapular/triceps
-0.03 0.03 0.09
0.36** 0.43** 0.27*
-0.02 0.27; 0.37**
0.05 0.08 0.09
0.05 0.30** 0.41**
0.10* 0.25’; 0.27**
Circumferences Waist circumference Hip circumference Thigh circumference
-0.06 -0.19 -0.16
0.44**
0.15 -0.05 -0.10
0.16 0.11 0.11
0.32** 0.29** 0.19
0.21** 0.15** 0.03
0.14 0.15
0.09 0.17
0.18** 0.25**
Height Weight BMI
Waist/hip ratio Waist/thigh ratio
0.32** 0.24* 0.39** 0.40**
0.13 0.09
0.28** 0.32**
All -0.05
0.19** 0.22**
*p < 0.05; **p < 0.01.
testosterone is neither a determinant of body fat distribution, nor of metabolic risk factors. Table 6 shows the independent contribution of free testosterone and degree of overweight (BMI) to measures of fat distribution in the pooled data. Both BMI and, to a lesser degree, free testosterone were independently positively associated with all indicators of fat distribution. Waist circumference was much more strongly associated to BMI compared to the other indicators of fat distribution. In Tables 7(a), (b) and (c) it is shown that free testosterone has a positive association with total cholesterol and triglycerides independently from BMI and 3 indicators of fat distribution. All 3 indicators are negatively associated to HDLcholesterol concentrations and the HDL/total cholesterol ratio and positively associated with triglyceride concentrations independently from BMI and free testosterone concentrations. The adjusted associations of waist/hip ratio and
waist circumference to serum cholesterol are significant. This is not so for waist/thigh ratio. The 3 models with different indicators of fat distribution do not show appreciable differences in the percentage of explained variance. Tables 8(a), (b) and (c) show the same regression models as shown in Tables 7(a)-(c) but now insulin has been added as an additional independent variable. The partial regression coefficients of free testosterone levels were now reduced to non-significant levels except for the association between free testosterone and triglycerides when adjusted for waist/hip ratio or waist circumference. Insulin was independently associated with all serum lipids and lipoproteins. No indicator of fat distribution was associated to serum cholesterol in these models and only waist/hip and waist/ thigh ratios, but not waist circumference alone, to serum triglycerides. All indicators of fat distribution remained negatively associated to HDLcholesterol and the HDL/total cholesterol ratio.
free-T
1.6
0
LM H Sweden
L M H Poland
L M H Netherlands
LMH (Verona)
=WTR(1.30
y
:1.30 WlR(lLo
;
L
M H Italy (Naples)
:WTR>l.40
Fig. 3. Average serum free testosterone levels (free-T) in categories of waist/thigh circumference ratio (WTR) in randomly selected 38-year old European women (SEM shown by bars).
Androgenicity and Body Fat Distribution
21
Table 5. Correlations of free testosterone (free T) with metabolic variables (triglycerides, HDL-cholesterol, total cholesterol, and HDL/total cholesterol ratio, insulin and blood pressure) in European women born in 1948, in different parts of Europe Independent variable
Dependent variable
Metabolic variables
Free testosterone
Center Sweden
Poland
Netherlands
Verona
Naples
All
0.21* 0.08 -0.05
0.23* -0.15 -0.26’
0.22* -0.01 -0.20
0.11* -0.06 -0.12**
Serum lipids
Total cholesterol HDL-cholesterol HDL/total cholesterol
0.21* -0.04 -0.03 -0.20 -0.11 -0.13
Serum insulin
Insulin
-0.07
0.57**
0.19
0.13
0.34**
0.20**
-0.23’ -0.26*
0.16 0.23*
0.15 0.23*
0.19 0.19
0.05 0.16
0.09 0.15**
Blood pressure
Systolic Diastolic *p < 0.05: **p < 0.01.
Table 6. Independent contributions of degree of androgenicity (free testosterone) and BMI to measures of fat distribution in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient P
Dependent variables
Independent variables
Waist/thigh
Constant BMI Free Tt Multiple r2
1.050 0.011 8.9*** 0.021 4.2*‘* 0.22 (F = 4.20, p < O.OOOl)t
Waist/hip
Constant BMI Free T Multiple r2
0.0654 0.006 0.006 0.21 (F = 4.73, p < o.OOol)t
Subscapular/ triceps
Constant BMI Free T Multiple r2
0.080 9.3*** 0.030 4 8*** 0.064 0.24 (F = 8.30, p < O.OOOl)t ’
Waist
Constant BMI Free T Multiple r2
1.050 32.0*** 1.825 2.4’ 0.575 0.72 (F = 4.50, p < O.OOOl)t
t-Value
9 5*** 2:3*
*p < 0.05; ***p < 0.0001. tFree T = serum free testosterone.
Tables 9(a), (b) and (c) show that free testosterone was, in multiple regression analysis, independently associated with serum insulin but not with blood pressure. BMI was independently associated with serum insulin (not when adjusted for waist circumference [Table 9(c)] and blood pressure whereas body fat distribution was not related to blood pressure in multiple regression analysis. Only waist circumference was associated with serum insulin levels. Insulin levels were not associated with blood pressure independently of BMI, free testosterone, and indicators of fat distribution (not shown). DISCUSSION
This study shows that in randomly selected premenopausal women all of the same age, free
testosterone levels in serum are lowest in the north of Europe (Sweden, Poland, and The Netherlands) and highest in Italian women. This trend has some overlap with differences in fat distribution in the same centers. As we have discussed elsewhere, it is difficult to assess the optimal indicator of fat distribution [9]. The results from the multiple linear regression analysis from which 3 different indicators can be compared with respect to their associations to metabolic risk factors suggest the same (Tables 7-9). Waist circumference alone is more strongly associated to serum cholesterol, the HDL/total cholesterol ratio, and serum insulin compared to waist/thigh and waist/hip ratios. The waist/thigh ratio, on the other hand, shows the strongest partial associations with HDLcholesterol and triglycerides. The differences in
28
JACOBC. SEIDELLet al.
strength of associations of waist/hip ratio, waist/thigh ratio, and waist circumference with metabolic risk factors were not very large and the proportion of explained variance in metabolic risk factors does not change very
much when different indicators of fat distribution are exchanged for each other. Correlational analysis in the pooled material showed significant associations between free testosterone levels with indicators of fatness and body
Table 7(a). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist/thigh), and BMI to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient P
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist/thigh BMI Multiple r*
4.62 0.10 0.41 -0.0003 0.02 (F = 5.36, p < O.OOOl)t
2.16* 0.88 -0.03
HDL cholesterol (mmol/l)
Constant Free testosterone Waist/thigh BMI Multiple r2
2.46 0.008 -0.52 -0.014 0.09 (F = 1.39, p = 0.14)?
0.49 -3.42*** -3..52”*
HDL/total cholesterol
Constant Free testosterone Waist/thigh BMI Multiple r*
0.49 -0.003 -0.11 -0.003 0.09 (F = 4.88, p < O.OOOl)t
-0.85 -3.15** -3.38***
Triglycerides (mmol/l)
Constant Free testosterone Waist/thigh BMI Multiple r*
-0.61 0.054 0.76 0.017 0.12 (F = 0.75, p = 0.74)t
*p < 0.05; **p < 0.01; ***p
t-Value
2.43* 3.56*** 3.06**
< 0.001.
tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers. Table 7(b). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist/hip), and BMI to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient P
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist/hip BMI Multiple r*
3.77 0.10 1.96 -0.007 0.03 (F = 5.16, p < O.OOOl)t
2.15* 2.05; -0.54
HDL cholesterol (mmol/l)
Constant Free testosterone Waist/hip BMI Multiple r*
2.60 0.003 - 1.05 -0.014 0.09 (F = 0.83, p = 0.65)t
0.18 -3.39*** -3.41***
HDL/total cholesterol
Constant Free testosterone Waist/hip BMI Multiple r*
0.56 - 0.0036 -0.27 -0.003 0.15 (F = 3.99, p < o.oool)t
-1.07 -4.01*** - 2.94**
Triglycerides (mmol/l)
Constant Free testosterone Waist/hip BMI Multiple r2
-0.64 0.063 1.27 0.019 0.11 (F = 0.77, p = 0.72)t
*p < 0.05; **p < 0.01; ***p
t-Value
2.85** 2.87** 3.21**
< 0.001.
tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers.
Androgenicity and Body Fat Distribution
29
Table 7(c). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist circumference), and BMI to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient B
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist BMI Multiple r2
4.18 0.10 0.026 -0.044 0.03 (F = 5.03, p < o.oOOl)t
2.08* 2.69*’ -2.08’
HDL cholesterol (mmol/l)
Constant Free testosterone Waist BMI Multiple r2
2.21 0.002 -0.009 -0.003 0.08 (F = 1.17, p = 0.29)t
0.13 -2.85** -0.45
HDL/total cholesterol
Constant Free testosterone Waist BMI Multiple r2
0.48 -0.004 - 0.003 -0.001 0.11 (F = 3.35, p < o.oool)t
- 1.04 -4.11*** -0.72
Triglycerides (mmol/l)
Constant Free testosterone Waist BMI Multiple r’
-0.20 0.063 0.012 0.004 0.10 (F = 0.79, p = 0.70)t
t-Value
2.87** 2.60** 0.42
*p < 0.05; **p < 0.01; ***p < 0.001. tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers.
Table 8(a). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist/thigh), BMI, and fasting insulin concentrations to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient B
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist/thigh BMI Insulin Multiple r2
4.73 0.08 0.27 -0.01 0.03 0.06 (F = 4.73, p < O.OOOl)t
1.74 0.59 -1.01 4.14***
HDL cholesterol (mmol/l)
Constant Free testosterone Waist/thigh BMI Insulin Multiple r2
2.44 0.01 -0.49 -0.01 -0.007 0.11 (F = 0.94, p = 0.54)?
0.77 - 3.25** -2.83** -2.55*
HDL/total cholesterol
Constant Free testosterone Waist/thigh BMI Insulin Multiple r2
0.49 -0.001 -0.09 - 0.002 - 0.002 0.14 (F = 14.69, p < O.OOOl)t
-0.32 -2.86** -2.18** - 5.03***
Triglycerides (mmol/l)
Constant Free testosterone Waist/thigh BMI Insulin Multiple r’
-0.54 0.04 0.67 0.009 0.02 0.19 (F = 0.77, p = 0.75)t
f-Value
1.86 3.25** 1.69 5.95***
*p < 0.05; **p < 0.01; ***p < 0.001. tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers.
30
JACOBC. SEIDELLet al. Table 8(b). Independent contributions of degree. of androgenicity (free testosterone), body fat distribution (waist/hip), BMI, and fasting insulin concentrations to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient B
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist/hip BMI Insulin Multiple r*
3.89 0.08 1.71 -0.02 0.03 0.06 (F = 3.91, p < O.OOOl)t
1.69 1.82 - 1.50 4.08***
HDL cholesterol (mmol/l)
Constant Free testosterone Waist/hip BMI Insulin Multiple r2
2.57 0.007 -1.00 -0.01 -0.007 0.11 (F = 0.53, p = 0.95)?
0.48 - 3.24** -2.72*+ -2.57”
HDL/total cholesterol
Constant Free testosterone Waist/hip BMI Insulin Multiple r*
0.55 -0.002 -0.25 - 0.002 -0.003 0.16 (F = 2.63, p < O.OOOl)t
-0.49 - 3.79*** -1.75 -5.03***
Triglycerides (mmol/l)
Constants Free testosterone Waist/hip BMI Insulin Multiple r*
-0.56 0.05 1.10 0.010 0.02 0.18 (F = 0.81, p < 0.70)t
f-Value
2.24+ 2.59* 1.83 5.99***
***p < 0.001. *p < 0.05; **p < 0.01; tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers.
Table 8(c). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist circumference), BMI, and fasting insulin concentrations to serum lipids in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient B
Dependent variables
Independent variables
Total cholesterol (mmol/l)
Constant Free testosterone Waist BMI Insulin Multiple r*
4.36 0.08 0.02 -0.04 0.03 0.06 (F = 3.78, p < O.OOOl)t
1.69 2.01* -2.15* 3.78***
HDL cholesterol (mmol/l)
Constant Free testosterone Waist BMI Insulin Multiple r*
2.17 0.006 -0.008 -0.003 -0.006 0.10 (F = 0.85, p = 0.65)t
0.37 -2.40; -0.43 - 2.30’
HDL/total cholesterol
Constant Free testosterone Waist BMI Insulin Multiple r*
0.46 -0.019 -0.002 0.001 -0.003 0.15 (F = 2.30, p < 0.0013)t
-0.57 - 3.32** 0.79 -4.59***
Triglycerides (mmol/l)
Constant Free testosterone Waist BMI Insulin Multiple r2
-0.08 0.05 0.007 0.004 0.02 0.17 (F = 0.85, p = 0.65)-f
t-Value
2.35* 1.63 0.38 5.76+++
*p c 0.05; **p < 0.01; ***p < 0.001. tf-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slope and/or intercepts differ beyond chance between the centers.
Androgenicity and Body Fat Distribution
fat distribution and a number of metabolic variables (total cholesterol, fasting insulin, and diastolic blood pressure). We observed interactions in the relationships between the independent variables and dependent variables in the separate centers when we performed analysis of variance of the regression coefficients. Differences in linear relationships between the centers cannot be readily explained but it must be noted that the variation in all measurements in the pooled material is very much larger than in the separate centers and the associations, even in the pooled material, were weak. The reason why
31
in the Swedish cohort, for example, no associations could be found between free testosterone and fat distribution and metabolic risk factors may be because the Swedish women were all at the lower end of the range of testosterone levels and fat distribution. Performing analysis in the pooled data which comprises the total range of fat distribution and metabolic risk factors in European women is more likely to give a reliable estimate of the “true” relationships than analysis in all centers separately. With regard to the testosterone values we must emphasize that blood samples were taken at various phases of
Table 9(a). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist/thigh), and BMI to insulin and blood pressure in European women born in 1948 (oooled data). Results of multinle linear regression Regression coefficient P
Dependent variables
Independent variables
Serum insulin (ID/l)
Constant Free testosterone Waist/thigh BMI Multiple r*
-3.10 0.66 4.24 0.37 0.12 (F = 8.90, p < O.OOOl)t
2.24’ 1.49 4.99***
Diastolic blood pressure (mmHg)
Constant Free testosterone Waist/thigh BMI Multiple r*
53.18 0.79 1.77 0.82 0.12 (F = 2.66, p < 0.0005)t
1.52 0.35 6.21***
Systolic blood pressure (mmHg)
Constant Free testosterone Waist/thigh BMI Multiple r*
79.96 -0.08 9.71 1.17 0.13 (F = 1.661, p < 0.064)t
-0.11 1.40 6.38***
f-Value
*p < 0.05; **p < 0.01; ***p < 0.001. tF-value calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers. Table 9(b). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist/hip), and BMI to insulin and blood pressure in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient B
Dependent variables
Independent variables
Serum insulin (ID/l)
Constant Free testosterone Waist/hip BMI Multiple r*
-3.79 0.70 7.89 0.37 0.11 (F = 8.94, p < O.OOOl)t
2.42* 1.35 4.95*+*
Diastolic blood pressure (mmHg)
Constant Free testosterone Waist/hip BMI Multiple r*
52.98 0.81 1.77 0.83 0.12 (F = 2.84, p c 0.0002)t
1.58 0.31 6.16***
Systolic blood pressure (mmHg)
Constant Free testosterone Waist/hip BMI Multiple r2
75.25 -0.007 22.86 1.14 0.13 (F = 1.51, p c 0.093)-f
f-Value
-0.01 1.61 6.19***
*p < 0.05; **p < 0.01; ***p < 0.001. TF-values calculated from analysis of variance of regression coefficients over groups. A significant F-value indicates that the slopes and/or intercepts differ beyond chance between the centers.
32
JACOBC. SEIDELLef al. Table 9(c). Independent contributions of degree of androgenicity (free testosterone), body fat distribution (waist circumference), and BMI to insulin and blood pressure in European women born in 1948 (pooled data). Results of multiple linear regression Regression coefficient P
Dependent variables
Independent variables
Serum insulin (ID/l)
Constant Free testosterone Waist BMI Multiple r2
-5.92 0.62 0.22 0.024 0.14 (F = 9.79, p < o.oOOl)t
2.18* 3.71*** 0.19
Diastolic blood pressure (mmHg)
Constant Free testosterone Waist BMI Multiple rz
51.31 0.76 0.11 0.64 0.12 (F=3.19, p
