Adiponectin is an enigmatic protein released from fat cells and commonly considered to signal to peripheral tissues to improve insulin action. In line with its presumed effects, circulating adiponectin is not only lower in obese individuals but also correlates with insulin sensitivity and related metabolic phenotypes [e.g. lower triglyceride, higher HDL cholesterol, lower C-reactive protein (CRP) etc] and does so independently of body mass index (BMI). Likewise, in relatively simple prospective epidemiological studies, higher baseline adiponectin levels consistently associate with lower risk for type 2 diabetes independently of simple adiposity measures (1), and some (2), but not all (3), general population cohort studies also report its inverse association with incident cardiovascular events. So far so good, but a closer inspection of the literature rapidly reveals several apparently contradictory findings. Most striking amongst these is the observation that higher adiponectin levels in the elderly with and without established vascular disease associate with higher – not lower – risk for all-cause and cardiovascular mortality (4). Similar findings have been reported in younger cohorts undergoing coronary angiography (5) or those with chronic kidney disease (6).
In the light of these latter observations, and in the present issue of this journal, Forsblom and colleagues examined the association of adiponectin with 11-year risk of all-cause and cardiovascular mortality in the nationwide multicentre cohort of Finnish adults with type 1 diabetes (the FinnDiane study) (7). Of interest and in keeping with data from the present study, circulating adiponectin levels appear generally higher in patients with type 1 diabetics than in nondiabetic individuals. More importantly, Forsblom et al. find higher baseline adiponectin levels associated with higher all-cause and cardiovascular mortality in their cohort of patients with type 1 diabetics, an observation that held even when they adjusted for a comprehensive panel of potential confounding factors including: measures of renal dysfunction, measures of metabolic control, presence of existing cardiovascular disease and duration of diabetes. In addition, the link between higher baseline adiponectin and higher mortality was present also in patients with type 1 diabetics without overt nephropathy, further suggesting renal disease is an unlikely a major confounding factor underpinning these observations. In short, the findings from the present large and well-conducted study are noteworthy and, once again, somewhat puzzling. That said, the authors correctly admit that even their study cannot completely reassure against confounding by unmeasured ‘upstream’ factor(s), which not only increase adiponectin but also predict greater mortality.
Several questions arise from the present study (7) and prior overlapping observations (4–6). Most important amongst these is the potential nature of any link between higher circulating adiponectin levels and higher all-cause and cardiovascular mortality in a range of populations including patients with type 1 diabetes? As far as one can tell, adiponectin itself is unlikely to be a toxic agent to the vasculature although its molecular role in man (discussed further below) is not as clear as commonly appreciated. To get closer to a potential ‘missing link’, a more detailed inspection of the FinnDiane cohort would be useful. In this regard, it is noteworthy that overall death rate of one per 100 patient years is higher than the expected background death rate in a general population with a baseline age near 40 years. This is in keeping with known excess death rates in type 1 diabetes and fits with the present cohorts characteristics: an average duration of diabetes around 20 years, near 50% of the patients with hypertension, 50% with retinopathy and around a quarter with albumin excretion ratio in the microalbuminuria range or above (7). In other words, the present cohort exhibits several major vascular risk factors, yet only around 10% were on statin therapy, a figure that undoubtedly would be higher in the modern management of type 1 diabetes. Even though the authors adjusted for some of these risk factors and for established macrovascular disease, might there still be a potential for subclinical cardiac or vascular disease to confound their headline observation? The answer is almost certainly yes.
To address confounding by subclinical cardiac disease, we recently measured NT-proBNP levels – a marker of cardiovascular risk amongst those with CHD, with associations reported for both short-term outcomes in acute syndromes and long-term cardiovascular end-points in those with stable disease – in baseline samples in the aforementioned British Regional Heart Study (4). We examined to what extent adjustment for NT-proBNP (as a continuous measure) would help explain the previously reported positive associations of adiponectin with all-cause mortality and cardiovascular mortality in this elderly cohort. As expected, adiponectin concentration was inversely associated with several conventional cardiovascular disease (CVD) risk factors (8). However, perhaps somewhat against expectations, it was significantly and positively associated with NT-proBNP concentration. After adjustment for several vascular risk factors, including renal function and muscle mass, relative risks associated with a top third versus bottom third comparison of adiponectin concentration were 1.51 (1.02–2.23) for coronary heart disease, 1.67 (1.15–2.41) for CVD mortality and 1.41 (1.13–1.95) for all-cause mortality. Upon further adjustment for NT-proBNP, these relative risks attenuated to 1.31 (0.88–1.94), 1.31 (0.90–1.91) and 1.26 (1.01–1.59), respectively (8). In other words, adjustment for NT-proBNP markedly attenuated adiponectin’s association with mortality and vascular outcomes in asymptomatic elderly men, and it was the only measure to have any such effect. As NT-proBNP levels are primarily thought to reflect ventricular stress or ischaemia (silent or otherwise), our findings add some weight to the possibility that subclinical vascular disease acting via NT-proBNP or related peptides might raise circulating adiponectin levels. In this regard, it is notable that adiponectin and NT-proBNP (and BNP) levels correlate strongly and positively in those with and without heart failure leading us to speculate that natriuretic peptides might promote adiponectin secretion via by specific adipocyte membrane receptors (9). Regardless of exact mechanism(s), cohorts like FinnDiane would do well to add NT-proBNP levels to their baseline phenotyping to test the hypothesis set out here, namely that NT-proBNP as a marker of subclinical vascular disease partially explains the positive link between adiponectin and mortality risks. Of interest, prior small studies show NT-proBNP levels to be elevated in patients with type 1 diabetics and predictive of mortality (10). Thus, adding NT-proBNP to larger cohorts like FinnDiane would allow better testing of the predictive ability of NT-proBNP for incident vascular events in type 1 diabetes, an important question given emerging evidence in general population cohorts showing this measure may outperform CRP in enhancing the prediction of CVD events beyond traditional risk factors (11).
Finally, the reason behind somewhat higher adiponectin levels in all patients with type 1 diabetes (regardless of duration of disease) is worth commenting upon. A recent excellent review by Cook and Semple (12) helps in this regard; these authors show nicely that rather than adiponectin being an upstream mediator of insulin action in man, the reverse pathway, whereby human insulin resistance acting via hyperinsulinaemia lowers circulating adiponectin levels, appears more credible. They also show there is little direct evidence for adiponectin being an insulin sensitizing agent in man. Several lines of evidence are used to support their arguments including the observation that with beta-cell failure, adiponectin levels would predictably rise in type 1 diabetes (12). This review is also helpful in its discussion of the likely physiological role of adiponectin and in comparing the overlapping relationships of adiponectin and the hepatic-derived SHBG. The latter observations, in turn, might help add another potential dimension on emerging observations of a strong link between adiponectin and NT-proBNP because not only does NT-proBNP (and BNP) correlate positively with insulin sensitivity and inversely with BMI but a genetic polymorphism that regulates BNP (and thus NT-proBNP) levels is associated with a lower risk of diabetes (13). Collectively, these findings suggest that links between BNP and adiponectin might also be indirect and, in part, regulated via insulin action.
In summary, it appears as if our understanding of the factors that i. dictate circulating adiponectin levels and ii. determine adiponectin’s association with disease outcomes is very much in its infancy. Potentially important factors determining its levels in type 1 diabetes are depicted in Fig. 1. To advance knowledge further, some of the suggestions presented herein (including the potential to add BNP or NT-proBNP measurement to prospective cohorts with adiponectin levels already measured) should help. In addition, better use of genetic studies to potentially tease out causal associations together with more detailed molecular work will also be valuable (12). In the interim, it is clear that given adiponectin’s varying associations with disease outcomes (some apparently good and some bad), its use as a surrogate biomarker or in the clinical setting is potentially problematic. In the longer term, it appears as if this enigmatic adipokine will continue to baffle, and one suspects more surprises are still to follow.