Adipokine concentrations in nonobese women: a study of reproductive aging, body mass index, and menstrual cycle effects.

Approval was obtained from the University of Michigan Hospital's Institutional Review Board for a protocol involving analysis of new data using archived human blood samples. Fifty-nine healthy volunteers from the parent study ( Lukacs et al., 2006 ) were categorized into three study groups by age and menstrual status: 21 cycling young women (CY; 20–30 years), 19 cycling older women (CO; 40–52 years) and 19 women in the early postmenopause period (EPM; 40–52 years; mean 2.8 ± 0.5 years postmenopause; Soules et al., 2001 ). BMI for the participants ranged from 20 to 28 kg/m 2 . The women reported no history of smoking, chronic medical conditions (i.e., heart disease, diabetes, polycystic ovarian syndrome), or current use of oral contraceptives, hormone therapy, anabolic or corticosteroids; had not been pregnant or breast feeding in the past 6 months nor engaged in dieting or exercise of more than 5 hr/week; and reported infrequent (less than 1 drink/week) alcohol consumption. Plasma values for liver function, hemoglobin, hematocrit, glucose, testosterone, free testosterone, dehydroepiandosterone sulfate (DHEAS), and thyroid stimulating hormone were within normal female ranges. In EPM participants, screening values for follicular stimulating hormone (FSH) were ≥ 30 IU/L, and all women reported a final spontaneous menstrual period at least 12 months prior to study. Menstruating women had a history of menses every 21–45 days. Participants in the parent study ( Lukacs et al., 2006 ) completed serial blood sampling at the General Clinical Research Center (GCRC) at the University of Michigan Hospital Systems. Fasting samples included glucose, triglycerides, and cholesterol. Mean values of four hourly samples for FSH and two samples (9:00 am and 12 noon) for estradiol (E 2 ) and progesterone were used for analysis. Height and weight measurements were obtained via measurement to calculate BMI: weight (kilograms)/height (meters squared). CY and CO women were studied in the early follicular phase (cycle Day 4 ± 1; Day 1 = the first day of menses) of a menstrual cycle and returned to the GCRC for three additional blood samples at weekly intervals for hormone measures (E 2 , progesterone). Menstrual cycle length was determined from the first day of menstrual bleeding to the day before the next bleeding episode. Plasma samples were assayed for E 2 , progesterone, and FSH as previously described ( Lukacs et al., 2006 ). Androgens (total testosterone, free testosterone and DHEAS) were measured by the hospital laboratory as part of the endocrine screening test panel. Fasting plasma glucose (FPG) was determined by glucose oxidase method in the laboratories of the University of Michigan Hospital System (Ann Arbor, MI). The lower limit of detection (LOD) was 20mg/dl; inter-assay coefficient of variation (CV) was 2.6%. This study analyzed new data on plasma adiponectin, leptin and insulin concentrations from archived fasting baseline blood samples; AD concentrations were also measured from the archived weekly blood samples of menstruating women. AD and leptin values were used to calculate the ratio of AD to leptin (A/L ratio). Fasting plasma glucose (FPG) and fasting insulin (FI) concentrations were used to estimate insulin resistance by homeostasis model assessment (HOMA-IR): FI (µU/ml) × FPG (mmol/L) / 22.5 ( Matthews et al., 1985 ) with the Oxford HOMA calculator ( Levy, Matthews, & Hermans, 1998 ) program v2.2 ( http://www.dtu.ox.ac.uk/index.php?maindoc=/homa/index.php ). For the prospective assessment of the menstrual cycle, stored plasma was sufficient only to determine AD values. Plasma samples for AD, L and FI were measured in duplicate by double antibody radioimmunoassay (RIA) (Linco Diagnostics, Inc, St. Charles, MO). The LOD for insulin was 2 µU/ml, for leptin 0.5ng/ml and for AD 1 µg/ml. Intra- and inter-assay CVs were 3.2% and 4.3%, respectively, for insulin; 3.3% and 6.0%, respectively, for leptin; and 1.6% and 7.0%, respectively, for AD. For the purposes of statistical analysis, values below the assay LOD were set at the value of assay sensitivity. Descriptive data are presented as mean ± SEM. As hormone measures and physiologic variables were positively skewed, we used natural log-transformed values in the analysis and treated measures as continuous variables. Differences in AD concentrations and parameters of insulin sensitivity (FI, FPG, HOMA-IR, A/L ratio, lipids) between the three study groups were evaluated by one-way ANOVA with post hoc Tukey. Correlations among AD concentrations, age, BMI, sex steroids, and insulin sensitivity parameters were performed using Pearson r . Group differences in categorical variables were compared using Chi square or Fisher's exact test. Two-tailed p values ≤ 0.05 were regarded as statistically significant. To examine the influence of ovulation on AD concentrations, we included data from subjects ( n = 21) who completed all four serial blood draws during the study cycle and had documented evidence of ovulation (progesterone ≥ 3 ng/ml) on at least one of the weekly follow-up blood tests. Repeated measures ANOVA was used to examine changes in AD at the four time points: the early follicular (EF) baseline evaluation and three weekly (EF+1, EF+2, EF+3) follow-up time points in the same menstrual cycle. As menstrual cycle lengths varied and the weekly samples reflected different times within individuals, the weekly sample with the highest progesterone value and a rising E 2 level from baseline was used to anchor the midluteal (ML) phase value. AD concentrations from the EF and ML phases were compared using the t -test. All analyses were performed with SPSS software (version 14.0; SPSS, Chicago, IL).

Subjects identified themselves as Caucasian (80%), Hispanic (8%), Asian (7%) and African American (5%). Study groups did not differ by race (χ 2 = 3.07, p = 0.2). Table 1 presents clinical characteristics of the three study groups. CO and EPM subjects were similar in age (matched within 1–2 years of each other by design) and older than the CY group. Testosterone values differed as expected, with higher concentrations demonstrated in CY women. E 2 was similar in the CO and CY women, but mean FSH was higher ( p = .0001) in the CO group compared to the CY group. Using the criteria developed by the Stages of Reproductive Aging Workshop ( Soules et al., 2001 ), CO women were in the late reproductive stage while CY women were in the peak reproductive stage. The majority of the sample (69%) was of normal weight (BMI ≤ 25 kg/m 2 ). There were no differences in BMI among the study groups. Triglyceride concentrations were similar among the subjects but cholesterol was higher in the early postmenopause women compared to both CO and CY ( p = .001) women. There were no differences in insulin sensitivity values (FI, FPG, HOMA-IR, A/L ratio) among the three study groups (data not shown). BMI was negatively correlated with the A/L ratio ( r = −0.70; p < .0001), positively associated with FI ( r = 0.28; p = .03) and HOMA-IR ( r = 0.28; p = .03), and showed a trend for a positive relationship with FPG ( r = 0.24; p = .06). To evaluate the interaction of BMI (normal weight [NW]: BMI ≤ 25 kg/m 2 ; overweight [OW]: BMI > 25kg/m 2 ) and menopause status on insulin sensitivity parameters, subjects were stratified into four groups: premenopausal normal-weight (PRE-NW; n = 30), premenopausal overweight (PRE-OW; n = 10), early postmenopausal normal-weight (EPM-NW; n = 11) and early postmenopausal overweight (EPM-OW; n = 8). As group sizes were small, we used the Kruskall Wallis ANOVA with post hoc Mann Whitney U tests to examine differences among the four groups. By ANOVA, no differences in measures of FI, FPG or HOMA-IR were observed among the groups, although trends in the expected directions were apparent as indicated by p values of < .1 ( Table 2 ). A/L ratio values were lower (less insulin sensitivity) in the OW groups compared to the NW groups regardless of menopause status ( Table 2 and Figure 1 , top panel). By ANOVA, AD differed by study group, with the highest concentrations found in the EPM group (CY: 13.0 ± 0.9 µg/ml vs. CO: 14.0 ± 1.1 µg/ml vs. EPM: 17.7 ± 1.5 µg/ml; p = .05), while leptin values were similar across the groups (CY: 10.9 ± 1.1 µg/ml vs. CO: 10.5 ± 1.4 µg/ml vs. EPM: 13.1 ± 2.0 µg/ml; p = NS). When subjects were reclassified by menopause status and BMI, EPM-NW women demonstrated higher levels of AD than their overweight counterparts and both groups of premenopausal women ( Figure 1 , bottom panel). Lower AD levels were also observed in PRE-OW compared to PRE-NW women. Leptin concentrations were higher in both overweight groups regardless of stage of reproductive aging ( Figure 1 middle panel). AD levels were negatively correlated with BMI ( r = −0.34; p = .009), leptin ( r = −0.31; p = .02), and triglycerides ( r = −0.38; p = 0.003) and positively associated with the A/L ratio ( r = 0.71; p < .0001). A trend for an inverse relationship between AD and insulin ( r = −0.24; p = .06) and HOMA-IR ( r = −0.25; p = .06) was observed. Age was not directly correlated with AD ( r = 0.23; p = 0.8) but became significant when controlling for BMI ( r = 0.29; p < .03). Leptin was positively correlated with BMI ( r = 0.72; p < .0001), cholesterol ( r = 0.28; p = .03), triglycerides ( r = 0.43; p = .001), FPG ( r = 0.40; p = .002), insulin ( r = 0.38; p = 0.003), and HOMA-IR values ( r = 0.40; p = .002) but negatively associated with the A/L ratio ( r = −0.89; p < .0001). Leptin did not demonstrate an association with age ( r = 0.035; p = NS). Neither AD nor leptin concentrations were correlated with E 2 , testosterone or free testosterone levels. The subset of cycling women were 37.2 ± 2.7 years of age on average, of normal weight (mean BMI = 23.3 ± 0.6 kg/m 2 ) with a mean cycle length of 30.3 ± 0.9 days. By repeated measures ANOVA, mean E 2 and progesterone concentrations changed significantly as expected across the menstrual cycle, but no differences in AD secretion were detected ( Figure 2 ). When the mean EF and ML values were compared, E 2 and progesterone values again differed, but AD concentrations were similar (EF: 13.3 ± 0.9 µg/ml vs. ML: 13.1 ± 1.0 µg/ml; t = 0.33; p = .75). Neither weight (NW, OW) nor age (young, old) status demonstrated an interaction with AD across the cycle. AD concentrations were not correlated with E 2 or progesterone values across the cycle.

Using a cross-sectional, case-controlled approach, we did not observe an aging effect on AD secretion in non-obese midlife women. In older cycling women with adequate E 2 , AD values were similar to those of the younger group despite a mean age difference of more than 20 years. In contrast, early postmenopausal women of similar age as the CO group but with negligible estrogen levels demonstrated significantly higher AD concentrations than the young cycling women. When the confounder of high BMI was removed, this menopause effect on AD was heightened. Taken together, these findings support an inhibitory role for estrogen on AD secretion during the premenopausal years that is removed in early menopause. Our data suggest that an elevated BMI prevents the early postmenopausal AD rise. When the CY and CO women were combined into one premenopausal group and compared to early postmenopause women, with both groups further stratified by BMI as normal weight (NW) or overweight (OW), AD levels were higher in EPM-NW women compared to all other groups, demonstrating the dual effects of normal body weight and estrogen loss. In contrast, AD was similar between PRE and EPM overweight women, suggesting that high BMI inhibited the potential rise in AD early in the menopause transition. Our results extend those of earlier work in mostly obese women with the metabolic syndrome ( Chu et al., 2006 ; Hong et al., 2007 ) demonstrating that the adverse (suppressive) effect of body weight on AD secretion is present even in healthy, overweight women who do not meet the clinical definition of obesity. Most prior investigations have been cross sectional, and researchers have observed either decreased ( Chu et al., 2006 ; Sieminska et al., 2006 ) or unchanged ( Hong et al., 2007 ; Milewicz et al., 2005 ; Nishizawa et al., 2002 ) AD concentrations after menopause. Differences in sample populations such as study groups with known metabolic conditions ( Chu et al., 2006 ; Hong et al., 2007 ; Sieminska et al., 2006 ) or unaddressed confounders such as advanced age ( Hong et al., 2007 ; Sieminska et al., 2006 ) may explain the discrepancies. The failure to confirm the reproductive aging stage of participants with both E 2 and FSH measures ( Jurimae et al., 2005 ; Nishizawa et al., 2002 ) or control for the timing of adipokine sampling in premenopausal women ( Chu et al., 2006 ) may also account for differences in study results. The higher AD concentrations observed in our early postmenopausal group, however, are congruent with findings from a recent longitudinal study that demonstrated a significant but small increase in AD across the menopause transition in healthy women evaluated for changes in cardiovascular risk factors and fat mass measures ( Lee et al., 2009 ). In contrast, the few studies of AD secretion across the menstrual cycle demonstrate very consistent findings. An analysis of data in patients (mean age 33 years) with endometriosis and other gynecological complaints suggested that AD levels do not differ between the proliferative and secretory phase of a menstrual cycle, times when estrogen and progesterone are known to change dramatically ( Takemura et al., 2005 ). In two pilot studies of lean women (Dafopoulos, Sourlas, Kallitsaris, Pournaras, & Messinis, in press; Kleiblova, Springer, & Haluzik, 2006 ), no differences in AD values were observed across the cycle phases. Our findings in a larger sample of both young and older cycling women corroborate this earlier work and confirm the stability of AD secretion despite changes in the sex steroid milieu with ovulation. The paradoxical findings between the menopause and menstrual-cycle arms in our study suggest there is an inhibitory effect of estrogen on AD during the reproductive years that remains stable and unaffected by dramatic elevations in the sex steroid milieu with ovulation. In early menopause, as estrogen declines below early follicular phase levels, this inhibition is removed. In addition, the lack of a menstrual cycle effect fails to implicate progesterone in this ovarian inhibition and suggests that follicular-phase levels of estradiol are sufficient to maximally suppress AD secretion. Measures of insulin sensitivity were similar among the three study groups and did not deteriorate as expected after menopause. The preservation of insulin sensitivity in early menopause in conjunction with the rise in AD is counterintuitive, given the reported increases in central adiposity associated with estrogen loss ( Poehlman, Toth, & Gardner, 1995 ) and the inverse association of AD and insulin resistance with these fat depots ( Cnop et al., 2003 ). Although we did not assess body fat characteristics, it is possible that our postmenopausal subjects, who were early in the postmenopausal period (mean 2.8 years), had not yet developed significant enough increases in central adiposity to suppress AD secretion. A later (> 5 years) postmenopausal study group would provide the opportunity to examine the duration of menopause as a contributing factor to this interaction between estradiol, central adiposity, AD secretion and insulin resistance risk. In addition to incorporating years postmenopause as a study variable, future work should include measures of central and peripheral adiposity to clarify this relationship. Finally, our observation of a significantly blunted A/L ratio in both pre- and early-postmenopausal overweight groups, in the face of unaltered values of other measures of insulin sensitivity, supports emerging data for use of this ratio as a more sensitive marker of insulin sensitivity ( Inoue, Yano, Yamakado, Maehata, & Suzuki, 2006 ). Leptin concentrations were positively associated with BMI, insulin and HOMA-IR values in the present study, which corroborates previous reports ( Chu et al., 2006 ; Hong et al., 2007 ; Kontogianni, Dafni, Routsias, & Skopouli, 2004 ). The higher levels of leptin demonstrated in overweight women compared to their normal-weight counterparts, regardless of menopause status, confirm the findings of prior studies, well controlled for BMI and menopause status, ( Hadji et al., 2000 ; Haffner, Mykkanen, & Stern, 1997 ) that demonstrated that body size and not estrogen status is the major mediator of L secretion. The tight control for the confounders of age and BMI and the accurate characterization of E 2 status with serial samples used to calculate a mean value are major strengths of this study. The cross-sectional design precludes a determination of causality and the small number of minority women in the sample limits generalizability, especially to populations most at risk for the adverse effects of obesity. Without measures of adiposity, we could not account for the effects of body fat distribution on adipokine concentrations in conjunction with E 2 status. Our findings are limited to women in the early postmenopausal period and cannot be extended to those in the late (> 5years) postmenopausal period ( Soules et al., 2001 ). In some analyses, CY and CO women were categorized as premenopausal but were likely in different reproductive stages. While the cycling young women in this study were in the peak reproductive stage, the cycling older women who had elevations in FSH and regular menstrual cycles, as determined by menstrual diaries at screening, were most likely in either the late reproductive stage or on the cusp of the early menopausal transition where some menstrual-cycle length variability is occurring ( Soules et al., 2001 ). In summary, our findings support the following hypotheses: 1) in the absence of obesity, midlife women demonstrate minimal adverse aging effects on insulin sensitivity and adipokine secretion; 2) menopause is not a major mediator of leptin secretion; 3) even in non-obese women, an elevated BMI (25–30 kg/m 2 ) plays a key role in adverse effects on both leptin and AD secretion before and after menopause; 4) the elevation in AD in the early postmenopausal period, coupled with the lack of change in AD across the menstrual cycle, suggests that follicular phase concentrations of E 2 may be sufficient to maximally suppress AD secretion in normal-weight women. These findings also support the protective effects of a normal BMI in early menopause and substantiate the importance of addressing body-weight management in the clinical care of women across the lifespan. Future studies using larger, more diverse study populations are needed to confirm these findings, further define the interactive effects of menopause on women's health, and ultimately guide the development of targeted nursing interventions to promote health and reduce metabolic disease risk.