Change in Circulating Adiponectin in Advanced Old Age: Determinants and Impact on Physical Function and Mortality. The Cardiovascular Health Study All Stars Study

Abstract

Background.

Cross-sectional studies show that adiponectin is higher in older than in younger adults but long-term change in adiponectin, its determinants, and its relationship to functional decline or survival in the elderly population have not been evaluated.

Methods.

We investigated predictors of longitudinal change in adiponectin, and the association of this adipokine or its antecedent change with physical deterioration and all-cause mortality in 988 participants in a population-based study who completed examinations in 1996–1997 and 2005–2006, had serial adiponectin measurements and underwent follow-up through June 2009.

Results.

Adiponectin level rose significantly during follow-up, but the increase was smaller in blacks, was associated with declining weight or fasting glucose and, in men, lower albumin, and was affected by medications. Adiponectin was independently associated with greater physical decline, but the relationship for adiponectin change was driven by concomitant weight decrease. Both adiponectin and its change independently predicted mortality, even after adjustment for weight change. The association for adiponectin and mortality was observed in whites but not in blacks and only for levels in the upper range (hazard ratio = 1.85, 95% confidence interval = 1.36–2.52 per SD ≥ 20 mg/L), whereas that for adiponectin change was linear throughout in both racial groups (hazard ratio = 1.30, 95% confidence interval = 1.10–1.52 per SD).

Conclusions.

Adiponectin levels increase over time in long-lived adults and are associated with greater physical disability and mortality. Such increases may occur in response to age-related homeostatic dysregulation. Additional investigation is required to define the underlying mechanisms and whether this represents a marker or causal factor for mortality in this age group.

THE adipocyte-derived peptide adiponectin has emerged as a potentially crucial molecule to cardiometabolic disorders (1). In the laboratory, adiponectin exhibits insulin-sensitizing, anti-inflammatory, and antiatherogenic properties (1). Yet, although its metabolic benefits have been confirmed in clinical studies, the relationship of circulating adiponectin to cardiovascular disease (CVD) in humans has been inconsistent (2). Although a protective association has been documented in younger adults without prevalent CVD (3), this has not been reproduced in groups with manifest heart disease (4,5) or in older adults (2), where the relationship has in fact been directionally opposite. This paradoxical adverse association with CVD and mortality in older adults has been replicated in different settings (6–8), but its basis remains unclear.

It is well recognized from cross-sectional studies that circulating adiponectin is higher in older than in younger adults (9,10). Together with the observation that men with exceptional longevity harbor genetic variants associated with greater adiponectin levels (11), this has led to the proposition that the higher concentration with age could reflect a survivor bias (9,11). Contrariwise, the higher age-related levels and link to increased mortality suggest that resistance to adiponectin’s beneficial actions (12) or, alternatively, a response to adverse processes (13) or direct deleterious effects could be involved. Serial measurement of adiponectin in older adults would shed light on the basis for higher levels documented cross-sectionally.

An important aspect of aging is a decline in cognitive and physical faculties, which serves as a harbinger of mortality (14). Obesity, dysglycemia, and CVD have distinct associations with circulating adiponectin levels, and all are strong determinants of functional decline (15). Defining the relationship of adiponectin to functional impairment in old age could help to clarify its reported association with mortality.

To address these questions, we measured levels of adiponectin 9 years apart in a cohort surviving to advanced old age who underwent detailed assessment of functional status. Our objectives were to identify determinants of adiponectin change and to examine the association of adiponectin and its antecedent change with physical impairment and mortality in this older cohort.

METHODS

Study Population and Procedures

The All Stars Study is an ancillary investigation of the Cardiovascular Health Study (CHS) whose objective is to assess determinants of successful aging. Details of both CHS and CHS All Stars have been previously reported (15,16). Briefly, CHS is a population-based investigation of risk factors for CVD in adults aged 65 years or older comprising 5,888 participants recruited in 1989–1990 and 1992–1993 from four U.S. field centers. Of the 5,553 participants in 1992–1993, 1,677 (30.2%) were alive and participated in the All Stars follow-up examination in 2005–2006 (15). These visits were conducted in person in 68% of participants and included a blood sample. Blood specimens collected and stored in 2005–2006 were paired with stored samples from the 1996–1997 examination to assess analytes of potential importance to aging.

The 1996–1997 and 2005–2006 examinations entailed updating of participants’ social and medical histories. Participants underwent standardized blood pressure and anthropometric determinations, along with assessments of cognitive and physical function (16). Collection of fasting blood samples and laboratory measurements have been previously described (17,18).

Definitions of Risk Factors

Hypertension was defined by blood pressure greater than or equal to 140 mmHg systolic or greater than or equal to 90 mmHg diastolic or by self-report and antihypertensive treatment. Diabetes was defined as a fasting blood glucose greater than or equal to 126 mg/dL or hypoglycemic therapy. Functional impairment was defined as the presence of physical disability (reported difficulty with ≥1 activities of daily living [ADLs]) and/or cognitive impairment (a score <80 on the Modified Mini-Mental Status Examination), as described previously (15). CVD was defined as coronary or peripheral arterial disease, heart failure, stroke, or transient ischemic attack. Clinical CVD was ascertained in the 1996–1997 and 2005–2006 examinations combining the CHS questionnaire, medical-record review, and physician confirmation (19). As detailed elsewhere (20), follow-up surveillance and ascertainment entailed biannual interviews of participants through 2005–2006. Potential incident events and all deaths were investigated by review of medical records and death certificates. These were initially classified by local physicians and subsequently adjudicated by a CHS committee using standardized criteria (16,20).

Outcomes

The functional outcome of incident physical disability was based on two measures, ADL difficulty and mobility impairment, ascertained at semiannual telephone contacts after the 2005–2006 examination. The first measure was defined as self-reported difficulty in one of six ADLs (walking around the house, transferring, bathing, toileting, dressing, and eating). For the second measure, mobility impairment, participants were asked about difficulty with two activities: walking half a mile and walking up 10 stairsteps. Follow-up questions determined the extent of difficulty (some, a lot, and unable to do) or, for those reporting no difficulty, the degree of ease (very easy, somewhat easy, and not that easy) with these activities. The response for each activity was then scored on a scale from −2 to 3 from “very easy to do” to “unable to do.” The two scores were averaged to generate a mobility score with 11 values in half-point increments, with increasing values indicating greater difficulty.

All-cause mortality following the All-Stars visit was ascertained by telephone follow-up, proxy contact, and review of obituaries. Follow-up for incident physical disability and all-cause mortality was available through June 2009.

Measurement of Adiponectin

Adipokine testing was performed at the CHS Core Laboratory in 2008 on serum stored at −70°C since venipuncture in 1996–1997 and 2005–2006. Adiponectin was measured using an ELISA (R&D Systems, Minneapolis, MN) with paired samples from both time points run together. The minimum detectable adiponectin concentration was 0.246 ng/mL; intra-assay and interassay coefficients of variation were 2.5%–4.7% and 5.8%–6.9%, respectively.

Statistical Analysis

Adiponectin and its interval change were mildly positively skewed and varied by sex. We therefore examined levels of covariates according to sex-specific tertiles of adiponectin and tested differences using chi-square or analysis of variance tests for trend. Continuous measures of adiponectin level and change were used in all other analyses. Pearson correlation coefficients between adiponectin and body size, glucose, and C-reactive protein (CRP) were computed for men and women separately in 1996–1997 and 2005–2006. Potential determinants of change in adiponectin included these and other clinical variables and their interval change, assessed in multiple linear regression models. Positively skewed variables were log transformed.

Associations of outcomes with adiponectin level or its change were assessed with multivariable Cox models (incident ADL difficulty and mortality) and generalized estimating equations (mobility score). Modeling was done in three stages, initially adjusting for age, sex, and race; adding significant clinical covariates thereafter; and subsequently adjusting for interval weight change. All models were tested for effect modification by sex, race, CVD, and health status. Linearity was assessed by plotting Lowess smoothers. Where departures from linearity were observed, linear splines were used. All analyses were performed with STATA, version 11.0 (College Station, TX).

RESULTS

CHS All-Stars participants with available adiponectin measurements (n = 988) were more likely to be male, younger, have better functional status, and report good-to-excellent health compared with All-Stars participants (n = 689) who did not have adiponectin measured.

There was a significant association between use of angiotensin-receptor blockers in 1996–1997 (in eight women and one man) and interval decline in adiponectin from 1996–1997 to 2005–2006. Use of thiazolidinediones in 2005–2006 (in nine women and nine men) was associated with higher adiponectin that same year and greater interval increase since 1996–1997. Because only a small subset was receiving these medications, these participants were excluded from relevant analyses.

Adiponectin was higher in women than in men at the 2005–2006 examination (17.9 ± 7.5 mg/L vs 14.0 ± 6.9 mg/L, p < .001) but increased more in men than in women during the preceding 9 years (3.3 ± 5.0 mg/L vs 2.5 ± 4.7 mg/L, p = .02). As shown in Table 1, increasing sex-specific adiponectin tertiles at the 2005–2006 visit were significantly associated with older age, greater education, and alcohol consumption but fewer blacks or diabetics, and lower measures of body size. Higher adiponectin was also related to poorer health status but lower CRP and use of diabetes medications, beta-blockers, or statins. There were no significant cross-sectional associations between higher adiponectin tertiles and functional measures except for lower grip strength.

Table 1.

Associations of Adiponectin With Clinical Characteristics at the 2005–2006 Examination^*

	Adiponectin Sex-Specific Tertile†
Variables	Low	Medium	High	p Trend
Age, y	84.3 ± 3.1	85.3 ± 3.5	86.1 ± 3.9		<.001
Black race	83 (26.2)	36 (11.1)	33 (10.3)		<.001
Education, y	12.9 ± 2.9	13.3 ± 2.6	13.5 ± 2.7		.018
Income ≥$12,000	226 (83.4)	255 (91.1)	244 (87.5)		.16
BMI, kg/m2	28.2 ± 4.6	26.6 ± 4.4	25.2 ± 4.2		<.001
Waist, cm	103 ± 13.3	97.5 ± 12.6	92.7 ± 12.4		<.001
Weight, kg	74 ± 14	69 ± 14	65 ± 13		<.001
Hypertension	235 (74.1)	227 (70.5)	218 (68.1)		.095
CVD	130 (41.0)	122 (37.7)	137 (42.8)		.64
Diabetes	75 (23.7)	37 (11.4)	30 (9.4)		<.001
Fasting glucose, mg/dL	106 ± 31	97 ± 15	95 ± 22		<.001
Total cholesterol, mg/dL	179 ± 38	183 ± 38	182 ± 39		.40
Current smoking	10 (3.2)	11 (3.4)	8 (2.5)		.63
Alcohol, drinks/wk	1.3 ± 4.0	1.7 ± 4.4	2.3 ± 4.8		.003
Self-reported poor/fair health	54 (17.0)	71 (22.0)	80 (25.0)		.014
Creatinine‡, mg/dL	1.0 ± 0.4	1.0 ± 0.5	1.0 ± 0.7		.72
CRP‡, mg/L	2.8 ± 11.5	2.1 ± 5.5	1.9 ± 7.7		<.001
Medications
Insulin	19 (6.0)	4 (1.2)	4 (1.2)		<.001
Oral hypoglycemic	48 (15.1)	24 (7.4)	19 (5.9)		<.001
Beta-blocker	138 (43.5)	125 (38.6)	84 (26.3)		<.001
Angiotensin-receptor blocker	55 (17.4)	47 (14.5)	48 (15.0)		.42
Angiotensin-converting-enzyme inhibitor	79 (27.3)	68 (23.4)	64 (21.8)		.12
Statin	131 (41.3)	125 (38.6)	96 (30.0)		.003
Functional measures
Mobility difficulty	115 (36.4)	120 (37.0)	122 (38.1)		.65
ADL difficulty	98 (30.9)	91 (28.1)	107 (33.4)		.49
Gait speed, m/s	0.8 ± 0.2	0.8 ± 0.2	0.8 ± 0.2		.22
Grip strength, kg	23.7 ± 8.8	22.6 ± 9.1	20.2 ± 8.2		<.001
3MSE score	88.6 ± 10.0	89.2 ± 10.1	88.0 ± 12.9		.51
Digit symbol substitution score	31.3 ± 13.9	33.8 ± 12.5	32.6 ± 14.5		.26

Notes: 3MSE = Modified Mini-Mental Status Examination; ADL = activities of daily living; BMI = body mass index; CRP = C-reactive protein; CVD = cardiovascular disease.

^*Entries are mean ± SD or n (%).

^†Tertile cut-points were 13.6 and 21.0 mg/L in women and 10.1 and 15.8 mg/L in men.

^‡Geometric mean.

Between 1996–1997 and 2005–2006, body weight decreased by 4.1 ± 7.0 and 2.9 ± 6.2 kg in women and men, respectively, but glucose and CRP showed no significant changes. In univariable analyses, the concentration of adiponectin in 1996–1997 was negatively correlated in women and men with weight (r = −.35 and −.22, respectively, both p < .001), glucose (r = −.32 and −.27, both p < .001), and CRP (r = −.24 and −.19, both p < .004). In women, these correlations remained unchanged at the 2005–2006 examination for weight but weakened significantly for glucose (r = −.21, p < .001) and CRP (r = −.19, p < .001), whereas in men, the correlation of adiponectin with weight strengthened significantly at the later date (r = .−30, p < .001), remained unchanged for glucose (r = −.26, p < .001) and disappeared for CRP (r = −.03, p = .58).

Multivariable predictors of change in adiponectin between 1996–1997 and 2005–2006 are shown in Table 2. Older age was an independent determinant of increased change in women, but not men, whereas black women tended to have less increase in adiponectin than nonblack women. Changes in weight and fasting glucose remained leading inverse predictors of adiponectin change in both sexes, as was serum albumin in men, with increases linked to corresponding declines and vice versa. Beta-blocker use was associated with lower levels of adiponectin at each visit, reflected in coefficients of similar magnitude but opposite direction for 1996–1997 and 2005–2006. For participants receiving beta-blockers at both examinations, the effect on adiponectin change was neutral. In turn, use of angiotensin-receptor blockers at the 2005–2006 visit was associated with greater adiponectin increase in men. Results were similar when adjusted for adiponectin level in 1996–1997.

Table 2.

Multivariable Predictors of Change in Circulating Adiponectin Between 1996–1997 and 2005–2006

	Women (n = 614)		Men (n = 347)
Risk Factor	Coefficient (95% CI)	p	Coefficient (95% CI)	p
Age, per y	0.14 (0.04 to 0.24)	.004	−0.06 (−0.19 to 0.07)	.39
Black race	−1.09 (−2.02 to –0.16)	.021	−0.51 (−1.82 to 0.81)	.45
Weight change (1996–1997 to 2005–2006), per 6.8 kg	−1.28 (−1.62 to −0.95)	<.001	−1.87 (−2.37 to −1.37)	<.001
Glucose change (1996–1997 to 2005–2006), per 25 mg/dL	−1.25 (−1.65 to −0.85)	<.001	−0.58 (−0.98 to –0.18)	.005
Serum albumin (1996–1997), per 0.29 g/dL	—	—	−0.98 (−1.40 to −0.56)	<.001
Any beta-blocker (1996–1097)	1.62 (0.62 to 2.61)	.002	1.75 (0.37 to 3.12)	.013
Any beta-blocker (2005–2006)	−1.75 (−2.49 to −1.01)	<.001	−1.59 (−2.59 to −0.60)	.002
Any angiotensin-receptor blocker (2005–2006)	—	—	1.79 (0.34 to 3.23)	.015
R2	.20		.24

Note: CI = confidence interval.

During follow-up, 266 of 681 participants at risk developed new difficulty with ADLs. There was a significant positive association between adiponectin and incident ADL difficulty that persisted after multivariable adjustment that included prior weight change (hazard ratio = 1.32 per SD 95% confidence interval [1.14–1.34]). Although change in adiponectin was modestly associated with future ADL difficulty after controlling for significant covariates, this relationship was abolished once weight change was considered (Supplementary Table 1). The mean mobility score increased from 95% confidence interval −0.14 (−0.24 to −0.04) at the 2005–2006 visit to 95% confidence interval 0.48 (0.35–0.61) at 3-year follow-up. The annual rate of change in the mobility score at the mean value of adiponectin was a 0.21 point increase. Although adiponectin was not associated with a higher mean mobility score, it did have a significant, if small, effect on the rate of increase in the score (Supplementary Table 1). Specifically, each unit standard deviation increase in adiponectin led to a 0.03-point steeper annual rise in the mobility score. By contrast, adiponectin change was associated with the mean mobility score but did not influence its rate of change, yet the relationship was abrogated after adjustment for weight change. There were no significant interactions of adiponectin or adiponectin change with sex, race, or CVD in these analyses.

During the 3 years following the 2005–2006 examination, 215 participants died. Assessment for effect modification of adiponectin or adiponectin change with respect to mortality revealed a significant interaction only between adiponectin and race (p = .042). Mean adiponectin levels in 2005–2006 were lower for black (n = 153) than for white (n = 817) participants (13.5 ± 7.5 mg/L vs 17.0 ± 7.4 mg/L, p = .026). African Americans also differed in risk factor profile (Supplementary Table 2), notably having more hypertension and diabetes, poorer health, and higher CRP than whites. Scatterplot smoothers for the relation of adiponectin and death showed a nonlinear association for whites but no association for blacks. For whites, a model using linear splines with knots at 10 and 20 mg/L showed a linear relationship for values greater than or equal to 20 mg/L but no association for values less than 20 mg/L. Table 3 shows the adjusted analyses for mortality, with the association for adiponectin (but not for adiponectin change) stratified by race and modeled as standard deviation increase for levels greater than or equal to 20 mg/L and 0 for values less than 20 mg/L. The significant association between adiponectin greater than or equal to 20 mg/L and mortality persisted after adjustment for covariates, including weight change. Adiponectin change, in turn, significantly predicted incident mortality across the recorded range, with a 30% relative increase of death for every 5.0 mg/L greater increase in interval change. When the 2005–2006 adiponectin level and its antecedent change were considered jointly, both remained significantly associated with fatal events.

Table 3.

Relations of Adiponectin and Its Antecedent Change With Mortality

	Adiponectin (2005–2006)			Adiponectin Change (1996–1997 to 2005–2006)
Model	Race	HR per SD* ≥20 mg/L (95% CI)	p	HR per SD* (95% CI)	p
1	White	1.93 (1.47–2.55)	<.001	1.38 (1.21–1.58)	<.001
	Black	0.65 (.23–1.85)	.42
2	White	2.03 (1.50–2.75)	<.001	1.37 (1.18–1.59)	<.001
	Black	0.64 (0.21–1.89)	.41
3	White	1.85 (1.36–2.52)	<.001	1.30 (1.10–1.52)	.002
	Black	0.44 (0.15–1.32)	.14
4	White	1.60 (1.14–2.25)	.007	1.22 (1.03–1.45)	.024
	Black	0.38 (0.13–1.17)	.09

Notes: Model 1. Adjusted for age, sex, and race.

Model 2. Additionally adjusted for weight, diabetes, alcoholic drinks/week, CVD, fair/poor self-reported health, and ln(CRP).

Model 3. Additionally adjusted for weight change.

Model 4. Adiponectin and adiponectin change jointly included with covariates from Model 3.

CI = confidence interval; HR = hazard ratio; CVD = cardiovascular disease; CRP = C-reactive protein.

^*Per SD increase in adiponectin (7.5 mg/L) or change in Adiponectin (5.0 mg/L).

DISCUSSION

To our knowledge, this is the first report to date to assess long-term longitudinal change in adiponectin in a community-based cohort and to examine the relationship of both adiponectin and its interval change with incident physical decline and all-cause mortality in an elderly population. Our study demonstrates that adiponectin does in fact increase with age, a phenomenon that was more pronounced in men than in women. This interval increase in adiponectin was blunted in black women but was associated, independent of race, with decreasing weight or fasting glucose, use of specific medications, and, in men, lower albumin. This investigation also showed circulating adiponectin to predict a higher incidence of ADL difficulty, along with more rapid onset of mobility impairment. Yet, although interval increase in adiponectin was also associated with greater deterioration in these physical disability measures, the relationship was largely attributable to an accompanying decline in weight. In addition, the current study highlights a positive association between adiponectin and its interval change with future mortality independent of measured covariates, including interval change in weight, although the association for a single measurement of adiponectin was only observed in whites and only for levels greater than or equal to 20 mg/L.

The observation that adiponectin concentration is greater in older adults than their younger counterparts, as documented previously in several cross-sectional studies (9,10), has been proposed to represent a survival advantage in the context of adiponectin’s favorable cardiometabolic associations (9,11). Indeed, genetic polymorphisms linked to higher circulating adiponectin have been found to be more common in centenarians than in other cohorts (11). Yet several epidemiological studies in older adults have reported positive, not negative, associations between higher adiponectin and incident CVD or all-cause mortality (2,6–8). This finding is contrary to expectation based on results from laboratory studies and younger cohorts, where the adipokine exhibits a protective cardiometabolic profile (1,3). The resulting conundrum does not appear explainable by weight loss and sarcopenia or by heart failure or renal insufficiency, adverse conditions that raise adiponectin levels (2,7,8). This has led to the concept of adiponectin resistance as an explanation for both the higher adiponectin level with age and its adverse association with outcome (12).

Our findings based on serial measurement of adiponectin yield insights into this question. The correlation between adiponectin and fasting glucose did weaken at follow-up in women, but not in men, suggesting possible development of adiponectin resistance in female participants. The inverse relationship between change in adiponectin and change in fasting glucose in both sexes is inconsistent, however, with major resistance to adiponectin’s metabolic effects, wherein adiponectin and glucose would rise in tandem.

More generally, our results are compatible with the premise that the age-related increase in adiponectin and its positive association with subsequent physical disability and all-cause mortality reflect heightened adiponectin secretion in response to adverse processes that accompany aging. Lately, the notion that adiponectin may serve a general “housekeeping” function by facilitating phagocytosis of apoptotic cells by macrophages has gained currency (21). Such a role for adiponectin that goes beyond direct insulin-sensitizing and antiatherogenic high-affinity signaling through the “AdipoR1” and “AdipoR2” receptors is supported by: adiponectin’s 1,000-fold higher plasma concentration relative to other adipokines; discovery of binding of adiponectin to the calreticulin receptor on the macrophage surface (22); and recognition by adiponectin of “apoptotic cell-associated molecular patterns” on dead cell surfaces (21). Thus, through low-affinity interactions with dead cells and macrophages, the highly abundant peptide can promote opsonization, thereby serving an anti-inflammatory function (21). Such a role for adiponectin would explain the association with increased mortality observed here or in other studies of older adults and cohorts with prevalent CVD, in whom higher circulating adiponectin would act as a marker for greater homeostatic dysregulation and associated apoptosis.

The same concept would also apply to our physical disability findings. Moreover, although adipose tissue is the principal source of circulating adiponectin, skeletal muscle also has the capacity to secrete this adipokine (23). Interestingly, recent data show that adiponectin expression increases in rhabdomyocytes of heart failure patients, with concurrent downregulation of AdipoR1 and the receptor’s downstream pathways, indicative of adiponectin resistance (23). These findings raise the possibility that aging-associated sarcopenia might exhibit the same pattern of adiponectin expression and local resistance observed in heart failure–associated skeletal muscle wasting. To the degree that such heightened rhabdomyocyte expression may affect circulating adiponectin levels, the latter could reflect the severity of skeletal muscle homeostatic derangements.

Whatever the basis for the aging-related increase in adiponectin, this peptide could also have direct harmful effects—whether through AdipoR1/R2, T-cadherin (24), or complement activation (25)—with variation across tissues depending on the degree of tissue-specific resistance. Indeed, murine overexpression of adiponectin leads to diminished bone mineral density (26), and higher adiponectin has been linked to osteopenia at non-load-bearing sites independent of body mass index in women (27).

Given the association of adiponectin-raising polymorphisms with exceptional longevity (11), however, it is likely that the beneficial glycometabolic associations of adiponectin trump any such adverse effects, although it is possible that such adverse effects could be of greater consequence in susceptible older adults. The finding that the change in adiponectin level was associated with mortality (independent of its final level) suggests that, in contrast to a genetically determined higher lifelong adiponectin concentration, longitudinal increase in adiponectin with aging has net-negative implications for survival.

We could not demonstrate a similar relation of adiponectin with mortality for blacks as for whites. African American participants in our sample had more comorbidities than whites, and black women had a significantly lower interval increase in adiponectin than white women. Although our finding of lower adiponectin levels for black than for white elderly individuals is consistent with a previous report (8), available data do not support the notion that the adverse associations seen for adiponectin with CVD (2) and mortality (8) are attenuated in older African Americans. Furthermore, we found no evidence of an adiponectin interaction with race in relation to physical disability outcomes or for interval change in adiponectin and mortality. To the extent that effect-measure modification by race was present in our cohort, it likely underscores the complex multifaceted relations of adiponectin with glycometabolic and vascular factors and how the balance between its beneficial and harmful associations varies in different clinical contexts and populations. Whether the differences observed here reflect race-specific genetic variation warrants additional study.

Several limitations must be acknowledged. Participants with available adiponectin measurements reported better health status than the remaining All-Stars cohort, such that these results apply to generally healthier elders. Although there is a survival bias in this cohort living to advanced old age, assessment of incident physical disability and mortality after the last adiponectin measurement yields results that are internally valid. Even if the individuals studied herein were available by virtue of a survival advantage from higher lifelong levels of adiponectin, this does not negate the finding that a surge in adiponectin in late life has negative implications for further survival. Another limitation is the current lack of information on cause of death; analyses of cause-specific mortality will require longer follow-up. Moreover, fasting insulin was not available, but the observed associations of adiponectin with outcome were also independent of fasting glucose as a measure of prediabetes. Last, we did not undertake measurement of high–molecular weight adiponectin, a fraction that has been reported to have stronger glycometabolic associations (28) and warrants closer scrutiny.

In conclusion, the present study of adults surviving to advanced old age shows that adiponectin increases longitudinally with age, that both the magnitude of this increase and the resulting concentration independently predict all-cause mortality, and that the final level of adiponectin is also independently related to physical decline in this population. The mechanisms underlying adiponectin’s age-related increase and its adverse association with physical function and mortality require further study.

FUNDING

The CHS All Stars Study was supported by the National Institute on Aging AG-023629. CHS was supported by contract numbers N01-HC-85079 through N01-HC-85086, N01-HC-35129, N01 HC-15103, N01 HC-55222, N01-HC-75150, and N01-HC-45133 and grant number U01 HL080295 from the National Heart, Lung, and Blood Institute, with additional contribution from the National Institute of Neurological Disorders and Stroke (http://www.chs-nhlbi.org/pi.htm). Additional support was provided through R01 AG-15928, R01 AG-20098, and AG-027058 from the National Institute on Aging and R01 HL-075366 from the National Heart, Lung and Blood Institute, and the University of Pittsburgh Claude. D. Pepper Older Americans Independence Center by P30-AG-024827 (to A.B.N.) and by K23 HL-070854 and R01 HL-094555 from the National Heart, Lung and Blood Institute (to J.R.K.).

SUPPLEMENTARY DATA

Supplementary material can be found at: http://biomed.gerontologyjournals.org/