Gender differences in factors influencing insulin resistance in elderly hyperlipemic non-diabetic subjects

Background The increase in the prevalence of insulin resistance-related metabolic syndrome, a disorder that greatly increases the risk of diabetes, heart attack and stroke, is alarming. One of the most frequent and early symptoms of metabolic syndrome is hypertriglyceridemia. We examined the gender differences between various metabolic factors related to insulin resistance in elderly non-diabetic men and postmenopausal women of comparable age suffering from hypertriglyceridemia, and compared them with healthy subjects of equal age. Results The indexes of insulin resistance HOMA IR and QUICKI were significantly higher in both hyperlipemic men and women than in controls; 95% confidence limits of hyperlipemic subjects did not overlap with controls. In both normolipemic and hyperlipemic men and women serum leptin correlated significantly with insulin resistance, while HDL-cholesterol correlated inversely with HOMA-IR only in women (both normo- and hyperlipemic), and serum tumor necrosis factor α (TNFα) only in hyperlipemic women. According to results of multiple regression analysis with HOMA-IR as a dependent variable, leptin played a significant role in determining insulin resistance in both genders, but – aside from leptin – triglycerides, TNFα and decreased HDL-cholesterol were significant determinants in women, while body mass index and decreased HDL-cholesterol were significant determinants in men. The coefficient of determination (R2) of HOMA IR by above mentioned metabolic variables was in women above 60%, in men only about 40%. Conclusion The significant role of serum leptin in determination of insulin resistance in both elderly men and postmenopausal women of equal age was confirmed. However, the study also revealed significant gender differences : in women a strong influence of triglycerides, TNFα and decreased HDL-cholesterol, in men only a mild role of BMI and decreased HDL-cholesterol.


Background
In association with pandemic obesity the prevalence of the insulin resistance-related metabolic syndrome is constantly growing [1]. As a consequence of this fact, type 2 diabetes mellitus and cardiovascular mortality occurs in much younger age groups [2]. A typical hyperlipemia, consisting of an increase of serum triglycerides and a decrease of serum HDL-cholesterol, is a characteristic and an early symptom of this syndrome [3].
With increasing age, body mass index (BMI) and adiposity, insulin sensitivity declines and the number of cardio-vascular risk factors increases in both genders [4][5][6]. It was repeatedly demonstrated that plasma concentration of leptin -a hormone produced mainly by adipose tissueis substantially higher in all age groups of women than in men [7][8][9][10]. This may be caused by different size and/or distribution of fat tissue compartments influenced by hormones: estrogens stimulate, whereas testosterone inhibits leptin secretion. In women subcutaneous fat mass prevails -and during augmentation of overweight it increaseswhile in men intra-abdominal fat mass prevails [11][12][13]. Subcutaneous fat in particular serves as a substantial source of tumor necrosis factor a (TNFa), which represents one of the factors that interfere with insulin signal transduction into the cells [14][15][16]. Leptin, TNFa and some other factors are abundantly expressed in adipose tissue and contribute to the insulin resistance that accompanies overweight and obesity. Leptin correlates positively with hyperinsulinemia, BMI, fat mass and hypertriglyceridemia, respectively, and correlates inversely with HDL-cholesterol and lean body mass [17][18][19][20][21][22][23][24][25].
The incidence and mortality of ischemic heart disease and of other consequences of atherosclerosis increases with age in both genders, especially after the age of sixty. In premenopausal women, however, the incidence of these disorders is considerably less frequent than in men of appropriate age. After the menopause the prevalence of metabolic syndrome and cardiovascular mortality in women gradually increases, attaining values comparable to men at about the age of 70 [2,26]. Paradoxically, it takes place at the time when serum leptin concentration in women has relatively decreased [27,28].
The aim of this study was to analyze the interrelations between several metabolic variables and factors related to insulin resistance in groups of both normal and hyperlipemic postmenopausal women and men of appropriate age, and to attempt to elucidate the gender differences and some pathophysiologic mechanisms of these differences. We compared homeostatic indexes of insulin resistance HOMA IR and QUICKI, serum lipid and insulin parameters, uric acid, leptin and TNFa between groups of subjects without apparent symptoms of metabolic syndrome, and groups showing mild hypertriglyceridemia with decreased HDL-cholesterol. In addition, serum concentration of the heart fraction of fatty acid binding protein (hFATP) was explored as a factor that might reflect the regulative role of PPAR gamma in lipid homeostasis [29,30], and serum IgG anticardiolipin (ACL-IgG) was investigated as an indirect indicator of oxidized lipid fractions related to atherosclerotic complications [31,32].

Subjects
The study was carried out on 70 out-patients of the Metabolic Center at the hospital in Sternberk, Czech Republic. From these, 40 patients (20 men and 20 women) were selected with mild hyperlipidemia, i.e. with plasma triglyceride concentration exceeding 2.0 mmol/l, total cholesterol exceeding 6.0 mmol/l, LDL cholesterol exceeding 4.0 mmol/l, and with HDL cholesterol concentration in men under 1.0 mmol/l, and in women under 1.2 mmol/l. These groups were denominated as "hyperlipemic". Two other groups (10 men and 20 women) with approximately normal serum values of these variables were taken as "controls". The average age in men was 59.1 ± 10.6 y, and in women 59.4 ± 10.1 y, respectively. The differences between lipid parameters of hyperlipemic and control groups were highly statistically significant, while the age differences were insignificant (see Table 1). None of the patients had clinically apparent diabetes mellitus, but some of the hyperlipemic patients exerted impaired glucose tolerance or impaired fasting glucose (values between 6.1 and 7.0 mmol/l, or between 6.1 and 7.8 mmol/ l, respectively). None of the patients was treated with insulin, peroral antidiabetics or antihyperlipemic drugs; some of them were treated with antihypertensive therapy. No signs of major clinical or laboratory symptoms of other diseases were present in any group of the explored patients. Blood samples were obtained in the morning via a venipuncture after overnight fasting. After clotting the serum was separated and stored at -20° until used. An informed consent was obtained from all probands.
Body mass indexes (BMI), defined as weight in kilograms divided by the square of height in meters, were calculated.

Biochemical methods
Serum leptin concentrations were measured by a sandwich ELISA test kit (Human Leptin ELISA, BioVendor Laboratory Medicine, Inc, Czech Republic). Its sensitivity limit was 0.2 ng/ml, intraassay CV 6.1% at the level of 7.5 ng/m, inter-assay CV 8.5% at the level of 4.8 ng/ml. Tetramethylbenzidine was used as a substrate; quality controls were human based. Several other hormones and peptides were estimated by routine immunochemical The homeostatic indexes of insulin resistance (HOMA IR and QUICKI) were calculated according to the homeostasis model of assessment [33][34][35] as follows: HOMA IR = fasting insulin (mU/ml) * fasting glucose (mmol/l) / 22.5; QUICKI = 1 / [log fasting insulin (mU/ml) + log fasting glucose (mg/100 ml)].

Statistics
Statistical analysis was performed using the Version 6 SAS/STAT software (SAS Institute, Inc., Cary, NC, U.S.A.). The Shapiro-Wilks tests were used in testing the normality of distribution. Some of the data obtained were not normally distributed. The statistical significance of differences between the means in the hyperlipemic and control groups were evaluated using the unpaired Student's T-test in the case of normal distribution of data sets, and using the Kolmogorov-Smirnov test when at least in one of the data sets the normal distribution was excluded. Spearman's rank-order correlation was used for correlation analysis. Multiple regression analysis was performed using HOMA IR indexes of insulin resistance as dependent variables, and other metabolic and hormonal factors (lipid parameters, BMI, leptin, TNFa, hFABP, ACL-IgG) as independent variables. The so-called step-down regression model was used to select dominant independent variables. Various four-member groups of independent (explanatory) variables were used for the analysis and the non-zero intercept was taken into account. The independ-ent variables were then dropped, one at a time; at each stage one variable making the least contribution to the dependent variable (i.e. that showed the least p-value in the test of the regression coefficient being zero) was excluded. The coefficient of determination R 2 , which can be viewed as a percentage explaining the total variance, was simultaneously monitored. A great drop in R 2 after excluding some independent variable enabled selection of those independent variables that could be thought to be the most important determinants of the dependent variable. Table 1 demonstrates mean parameters in individual groups of subjects matched according to sex, lipid parameters and age. While the age of all four groups did not differ substantially, the concentrations of total serum cholesterol, triglycerides, HDL-cholesterol and LDL-cholesterol differ very significantly in both male and female hyperlipemic groups as compared with controls. In addition, the concentration of triglycerides in control women was significantly higher than in control men, the concentration of triglycerides in hyperlipemic women was lower than in hyperlipemic men, and the concentration of HDLcholesterol in hyperlipemic women was very significantly higher when compared with hyperlipemic men. Table 2 shows the values of other metabolic and insulin parameters, factors related to insulin resistance and indexes of insulin resistance, respectively. Body mass indexes and uric acid concentration were significantly higher in hyperlipemic men as compared to controls, but not in hyperlipemic women. Uric acid concentration was substantially lower in hyperlipemic women than in hyperlipemic men. Plasma concentrations of glycemia, insulin and intact proinsulin were significantly higher in both hyperli- Statistical significance between control and hyperlipemic groups was tested using the unpaired Student's T-test in the case of normal distribution of compared data sets, and using Kolmogorov-Smirnov's test when at least in one of the data sets compared the normal distribution was excluded (marked with KS ). The significant p values are denoted by thick underlined numbers. With asterisks statistically significant differences between controls and hypelipemic patients of different gender are denoted (*, **, *** = p < 0,05, 0,01 and 0,001, respectively).

Results
pemic men and women as compared with controls of identical gender, while the concentration of leptin increased only in hyperlipemic men. However, serum leptin concentrations of both control and hyperlipemic women were significantly higher than in corresponding groups of men. Serum concentrations of TNFa, hFABP and ACL-IgG in hyperlipemic groups of both men and women were not significantly different from control groups. On the other hand, the indexes of insulin resistance HOMA IR and QUICKI differed very significantly in hyperlipemic groups of both men and women as compared with corresponding control groups, more distinctly in women.
From Fig. 1, presenting 95% confidence limits of insulin resistance indexes HOMA IR and QUICKI, we concluded that in groups of hyperlipemic patients of both genders the insulin resistance was substantially higher than in control groups; the groups did not overlap each other.
In Table 3 the results of Spearman's correlations between insulin resistance index HOMA IR and various metabolic parameters are presented. In the control group of men, positive significant correlation between HOMA IR and serum leptin concentration, and inverse significant correlation between HOMA IR and ACL IgG, respectively, were found. In the control group of women, the significance of Spearman's correlation between HOMA IR and leptin was more expressive; inverse correlation between HOMA IR and HDL-cholesterol was also present.
In the hyperlipemic group of men, the significance of the correlation between HOMA IR was more expressive in relation to the control group, and no significant correlation between HOMA IR and ACL IgG was found. In the hyperlipemic group of women, however, the significance of Spearman's correlation between HOMA IR and serum leptin concentration weakened, the inverse correlation between HOMA IR and HDL-cholesterol remained

Figure 1
Schematic presentation of 95% confidence limits of insulin resistance indexes HOMA IR and QUICKI in groups of control and hyperlipemic women and men. = Controls (women) = Controls(men) = Hyperlipemic women = Hyperlipemic men approximately unchanged, and a positive correlation between HOMA IR and serum concentration of TNFa appeared. Table 4 shows results of multiple regression analysis, when data from both control and hyperlipemic groups of each gender were judged together. HOMA IR was considered as a dependent variable and differently changed constellations of metabolic and other factors were taken as independent variables.
In men, BMI and leptin seemed to play a main role in influencing the insulin resistance index HOMA IR, while TGL, ACL IgG and LDL-cholesterol didn't play any significant role (see left columns of Table 4). The decreasing of HDL-cholesterol concentration may also have some influence (see a significant drop of R 2 after exclusion of this factor in Table 4A, 4B). But in the presence of leptin in the group of independent factors, the drop of R 2 after exclusion of HDL-cholesterol from these factors was minimal (see Table 4C). On the other hand, after the exclusion of TNFa from the group of independent variables (see Table  4B, 4D) the value of R 2 has unexpectedly risen, which could reflect the interference of TNFa with factors increasing the insulin resistance.
In women (see right columns of Table 4), the maximal values of R 2 were achieved with combination of independent variables containing TGL, leptin and HDL-cholesterol (about 60% influence on HOMA IR! -see Table  4A, 4B, 4C). TNFa seemed to play quite a different role than in men: after exclusion of this factor from the group of independent factors R 2 significantly decreased (see Table 4B, 4D). In contrast to men, the role of BMI seemed to be minimal. As in men, the role of ACL IgG and LDL-cholesterol in influencing HOMA IR was negligible, but in contrast to men, hFABP might play a certain role in this process (see Table 4D).   Generally, the insulin resistance (represented by HOMA IR) was in men much less influenced by metabolic variables than in women; while in women in some combinations of dependent variables R 2 reached 64 %, in men the maximal value of R 2 was only 39 %.

Discussion
In our previous paper [36], the mean value of HOMA IR in healthy subjects of both genders and of age comparable with our controls was 1.57 ± 0.87, and the mean value of index QUICKI 0.366 ± 0.029, respectively. These values, as well as the 95 % confidence limits, correspond to values found in controls in this study.
In accordance with many previous papers, serum concentrations of leptin in women (both control and hyperlipemic) were substantially higher than in men. In the control group of women the correlation between leptin and HOMA IR was highly significant. However, in hyperlipemic women the significance of this correlation lessened, because HOMA IR increased considerably (and significantly) but serum concentration of leptin only slightly (insignificantly). In men the significance of correlations between serum leptin and HOMA IR was high and approximately the same in both the control and hyperlipemic groups, because the values of HOMA IR as well as serum leptin have nearly doubled in hyperlipemic in relation to control groups. In non-hyperlipemic postmenopausal women the high concentration of serum leptin was not associated with higher insulin resistance: HOMA IR did not differ substantially from men. A significant increase of insulin resistance in hyperlipemic women was associated by only slight and insignificant increase of leptin concentration. According to Spearman's correlations, an increase of serum TNFa and/or a decrease of HDL-cholesterol might also play a distinct role in this respect. (see Table 3). In contrast to women, in hyperlipemic men the increase of insulin resistance index was approximately proportional with the increase of leptin concentration.
Multiple regression analysis affirmed the importance of leptin serum in increasing of insulin resistance in both genders. In men, only BMI and HDL-cholesterol from other factors studied seemed to play a certain role, but the maximal values of influencing HOMA IR reached only 39%, with leptin and BMI being the more important factors. On the other hand, in women the maximal determination of HOMA IR as high as 60% was registered in combination of serum leptin, TGL and decreased HDLcholesterol as independent factors; the role of BMI was insignificant.
It is not known how leptin is regulated. A strong correlation between plasma leptin and fasting insulin undoubtedly exists, but hyperleptinemia in both obese and lean humans is not likely the result of hyperinsulinemia [37]. A relationship between leptin and insulin dependent on sex or BMI was reported, but relationship between triglyceride concentrations and leptin was independent of sex, BMI, and insulin [18,24]. Hyperleptinemia, as an early sign of obesity, was closely linked to subcutaneous fat mass [39,40]. Percentage of body fat has been shown to be the strongest predictor of leptin levels even in lean women [41]. Leptin was highly correlated with percentage of body fat and with fat mass in adults irrespective of gender and age; however, the mean determinant of leptin plasma concentration in men and postmenopausal women was BMI, while in premenopausal women it was only the fat mass [42]. These findings contrast with our results showing minimal influence of BMI on HOMA IR in postmenopausal women.
All factors mentioned are connected with fat tissue: leptin and TNFa are directly produced chiefly by adipocytes, BMI growth is obviously accompanied by fat mass increase, and the typical hypertriglyceridemia associated with a decrease of HDL-cholesterol goes along with obesity and fat mass growth. The gender differences in circulating leptin were best explained by percentage of body fat  and -inversely -by lean body mass [25]. In both genders the intra-abdominal fat correlated with insulin resistance, while the subcutaneous fat correlated with circulating leptin [11,12]. In men obesity led to a prevalent increase of intra-abdominal fat, while in women of subcutaneous fat [13]. Influences of different compartments of adipose tissues could elucidate the variability of correlations between insulin resistance and high leptin concentrations in lean and obese subjects of both genders [43]. In our nonhyperlipemic postmenopausal women the content of subcutaneous fat mass might be higher than in non-hyperlipemic men of appropriate age, which indicated a higher serum concentration of leptin. However, the insulin resistance -related to intra-abdominal fat mass -did not differ from men. The significant increase of insulin resistance and leptin concentration in hyperlipemic men might reflect the growing content of both subcutaneous and intra-abdominal fat mass (see the significant increase of BMI). In hyperlipemic women the significant increase of insulin resistance accompanying only minimal insignificant increase of leptin could be caused by prevalent growing of intra-abdominal fat mass.
In elderly postmenopausal women, an association between leptin and plasma lipoprotein concentration was found which depended on adiposity [17], and inverse correlations between serum leptin and HDL-cholesterol were described [44]. In our study, insulin resistance in women seemed to be more notably than in men influenced by lipid disorders, i.e. positively by serum triglycerides and inversely by HDL-cholesterol. These findings might be important in considering the concept of treatment of insulin resistance-related disorders in postmenopausal women.
The significant role of TNFa in insulin resistance, caused by inhibiting the transduction of insulin signaling and by down-regulation of glucose transporter GLUT-4 and insulin receptor substrate-1, has been repeatedly confirmed [45][46][47][48]. Our results supported these findings unambiguously only in women, while in men TNFa seemed paradoxically to interfere with other factors -mainly BMI and leptin -in influencing insulin resistance, thus playing a quite different role. Previously it was found [46] that correlation between serum TNFa on the one side, and insulin, HOMA IR, serum triglycerids, respectively, on the other side, was substantially more significant in women than in men. Serum concentration of TNFa in patients with type 2 diabetes of both genders correlated only with the quantity of intra-abdominal fat compartment [50]. Visceral obesity correlated with plasmatic aldosterone and with insulin resistance only in premenopausal women, but not in men [51].
From all these data we might support our above mentioned conclusion -that rising of insulin resistance in hyperlipemic women was associated with an increase of intra-abdominal fat, because this fat mass in particular is a source of TNFa, which interfered with insulin sensitivity only in women. We came to this conclusion irrespective of the finding that the increase of serum TNFa in hyperlipemic women was statistically insignificant; results of Spearman's correlation (Table 3) and multiple regression analysis confirm a distinct role of this factor. In hyperlipemic men not only the serum concentration of TNFa has decreased instead of increasing, but according to multiple regression analysis it played a quite different role in influencing insulin sensitivity, interfering with factors that determined insulin resistance (leptin and BMI).
In the control group of men IgG anticardiolipin was inversely correlated to HOMA IR. The significance of this finding is not clear. These antibodies indicate vascular and thrombotic complications and oxidative modification of lipoproteins [52,53] and may represent an increased risk of atherogenic and inflammatory complications. In this case, however, their growing might be connected with an increase in insulin sensitivity. Anyway, ACL IgG evidently did not participate significantly in influencing the increase of insulin resistance associated with hyperlipidemia, although other anti-cardiolipin correlations could be masked by the relatively large inter-individual variations in this parameter.
Neither serum concentration of hFABP, a factor ensuring transmembrane transport and oxidative metabolisation of long-chain fatty acids [54,55], was significantly changed in hyperlipemic and insulin resistant subjects of both genders. This factor was very weakly associated only with HOMA IR in women (see Table 4D), indicating that enhanced metabolisation of fatty acids in cells might to some degree contribute to insulin resistance.

Conclusions
In postmenopausal women as well as in men of approximately equal age serum leptin plays a significant role as an important determinant of insulin resistance. In addition to this factor, in women the grade of insulin resistance is very considerably influenced by serum triglycerides, tumor necrosis factor alpha, and by decreased concentration of HDL-cholesterol, while in men only a mild influence of BMI and decreased HDL-cholesterol is observed. These findings are explained as a consequence of gender-related differences in adipose tissue composition and/or distribution in both normal-weight and over-weight subjects and should be taken into account in treatment of patients with metabolic risk factors of cardiovascular diseases.  Dr. Simona Gwozdziewiczová performed analysis of clinical and biochemical data and edited the manuscript.

List of abbreviations
Prof. Jirí Hrebícek initiated the study, participated in its design and coordination, and wrote and edited the manuscript.