Circulating levels of adiponectin are highly heritable, with more than 30-70% of its variability explained by genetics factors [17, 18]. A recent comprehensive linkage disequilibrium mapping revealed many SNPs in the adiponectin gene (ADIPOQ) were strongly associated with circulating adiponectin levels . In addition, an extensive bioinformatics analysis revealed the ADIPOQ region might be a high copy number variable region which potentially influences circulating adiponectin levels . Furthermore, ADIPOQ is a major gene influencing circulating adiponectin levels from the genome-wide perspective [18, 19]. The area of chromosome 3q27 where the ADIPOQ gene is located has been identified by genome-wide linkage studies (GWLSs) to be a susceptibility locus for risk for the type 2 diabetes (T2DM) , metabolic syndrome (MS)  and cardiovascular disease [22, 23]. However, this was not shown in recently published meta-analyses of genome-wide association studies (GWASs) [24–27]. The ADIPOQ gene is located nearby the IGF2BP2 (insulin-like growth factor 2 mRNA-binding protein 2) and AOMS1 (abdominal obesity-metabolic syndrome QTL 1) genes on chromosome 3q27 (based on Ensembl database). The other T2DM and MS susceptibility genes including HDLCQ5 (high density lipoprotein cholesterol level QTL 5) (on chromosome 3q24-q26), AGTR1 (angiotension receptor 1) (on chromosome 3q24), FGQTL6 (fasting plasma glucose level QTL 6) (on chromosome 3q21), RETNLB (resistin-like beta) (on chromosome 3q13.13), HYT7 (hypertension, essential, susceptibility to, 7) (on chromosome 3p14.1-q12.3), PPARG (peroxisome proliferator activated receptor gamma) (on chromosome 3p25.2) and ABHD5 (abhydrolase domain containing 5) (on chromosome 3p25.3-p24.3) may also in linkage disequilibrium with the ADIPOQ gene (based on Ensembl database). Thus, these genes may influence the functional mechanism and expression of the ADIPOQ gene including adiponectin levels.
Adiponectin has already been identified as a potential target for therapeutics to treat T2DM  and MS [29–31] in a series of clinical trials. Many existing drugs have been found to increase adiponectin levels, including statins (e.g. pravastatin, simvastatin, rosuvastatin and atorvastatin), angiotensin converting enzyme inhibitors and angiotensin receptor blockers (e.g. ramipril, quinapril, telmisartan, irbesartan and candesartan), β-adrenergic agonists, and thiazolidinediones (e.g. pioglitazone and rosiglitazone) . Other drugs that increase serum adiponectin levels were including non-statin anti-hyperlipidemic drugs (e.g. fenofibrate), non-TZD anti-diabetic drugs (e.g. acarbose and sulfonylurea glimepiride) and androgen blockers . A meta-analysis including 19 prospective studies had pointed out an increase in serum adiponectin levels in subjects undergoing treatment with thiazolidinediones (TZD) . Moreover, a systematic review including 33 clinical trials showed that exercise of varying prescription was able to increase serum adiponectin levels . However, a recent survey revealed the paradoxical findings regarding the role of adiponectin in human disease . According to the concept of the reversal epidemiology in the adiponectin physiology, adiponectin would behave as an insulin sensitizing and cardioprotective factor in the health state and as a wasting marker in the advanced states of disease .
Genetic variants in RETN (the resistin gene) have been examined by many groups, and it was estimated that up to 70% of the variation in circulating resistin levels could be explained by genetic factors . Moreover, recent fine-mapping of SNP studies which covering the full RETN gene revealed several SNPs of the RETN gene account for the high variablity of resistin levels . In the San Antonio Family Heart Study, the maximum linkage signal for the RETN expression was found on chromosome 19p13 (location of the RETN gene) . In addition, the RETN gene is located nearby the INSR (insulin receptor) and LDLR (low density lipoprotein receptor) genes on chromosome 19p13.2 (based on Ensembl database). This suggests that RETN expression may be cis-regulated, meaning there are variants in or near the RETN gene that influence the abundance of its mRNA . The other T2DM and MS susceptibility genes including ATHS (atherosclerosis susceptibility, lipoprotein associated) (on chromosome 19p13.3-p13.2), AKT2 (murine thymoma viral homolog-2) (on chromosome 19q13.2), FFAR 1 (free fatty acids receptor 1) (on chromosome 19q13.12), FFAR 2 (on chromosome 19q13.12) and FFAR 3 (on chromosome 19q13.12) may also in linkage disequilibrium with the RETN gene (based on Ensembl database). In the meta-analyses of genome-wide linkage studies (GWLSs), suggestive evidence of linkage was observed for LDL cholesterol [39, 40], apolipoprotein B , total cholesterol , and HDL cholesterol  on chromosome 19p13. These provide compelling evidence that the region of chromosome 19p13 harbor important determinants of lipid levels in individuals with T2DM . However, the RETN gene was not detected as a susceptibility locus for risk for the T2DM, MS and cardiovascular disease in recently published meta-analyses of genome-wide association studies (GWASs) [24–27]. Thus, whether the RETN gene modulates metabolic homeostasis independently or functions in concert with other causative genes in a haplotype block remains to be elucidated.
Compared to adiponectin, the effects of drugs treatment on resistin levels in patients with T2DM and MS is less described. However, a few clinical trials showed that the anti-diabetic (e.g. rosiglitazone) , anti-hypertensive (e.g. amlodipine)  and anti-dyslipidemic (e.g. pitavastatin)  drugs were able to reduce circulating resistin levels and may contribute to improving insulin action in patients with T2DM and MS. Recently, Koh et al. reported that amlodipine (a calcium channel blocker) therapies significantly decreased resistin levels greater than ramipril (an angiotensin-converting enzyme inhibitor) or candesartan (an angiotensin II receptor antagonist) therapies in patients with hypertension . In addition, resistin concentration decreased after long-term exercise training in overweight adolescents . A recent long-term follow-up study revealed elevated serum resistin levels were associated with higher rates of mortality and hospitalization for heart failure . However, serum resistin levels do not add prognostic information among high-risk persons with established coronary heart disease .
In consistent with our study, a significant inverse correlation between serum adiponectin and resistin levels has also been reported in the literatures [12, 13]. It has been reported that those with highest increases of adiponectin also displayed a trend towards a decline in resistin levels . Tuttolomondo et al. demonstrated that diabetic subjects with diabetic foot had higher resistin levels and lower adiponectin levels compared to diabetics without diabetic foot . Both hypoadiponectinemia and hyperresistinemia were also positively correlated with diabetes duration, hypertension, dyslipidemia, retinopathy, previous cerebrovascular disease (TIA/ischemic stroke), neuropathy, and diabetic foot grade . Furthermore, both hypoadiponectinemia and hyperresistinemia were associated with out-of-clinic hypertension  and may have prognostic significance for future cardiovascular events in patients with masked hypertension . Elevated resistin opposed to adiponectin plasma levels was proposed to be a strong predictive factor for the occurrence of major adverse cardiac events in patients with stable multivessel coronary artery disease over 1-year follow-up . A recent prospective longitudinal pilot trial revealed systemic therapy ameliorates endothelial cell function by increase adiponectin and decrease resistin levels in patients with plaque-type psoriasis . In addition, analysis with (18)F-fluorodeoxyglucose positron emission tomography revealed both adiponectin and resistin may be useful as biomarkers to reflect vascular inflammation . Thus, the balance of the opposite effects of adiponectin and resistin at the level of the endothelial cell may be an important determinant of endothelial dysfunction, and in turn the progress of atherosclerosis.
Several studies have illustrated the interaction between adipokines (including adiponectin and resistin) and adenosine 5' monophosphate-activated protein kinase (AMPK), and highlighted AMPK as a potential target for the development of tissue-specific AMPK modulators in the treatment of T2DM and MS . In patients with metabolic syndrome, thiazolidinediones (TZDs) including pioglitazone  and rosiglitazone  treatment markedly increased adiponectin and decreased resistin levels. These treatment effects on both adiponectin and resistin may further contribute to the AMPK activation exerted by TZDs . A recent randomized double blind clinical trial demonstrated that short-term treatment with losartan (an angiotensin II receptor antagonist drug) improved both adiponectin and resistin levels in hypertensive subjects . Furthermore, fenofibrate therapy improved both adiponectin and resistin levels, and may directly contribute to improving insulin sensitivity in hypertriglyceridemic patients . These may in turn exert detrimental and beneficial effects on glucose and lipid metabolism.
Crosstalk of adipokines including adiponectin and resistin at the expression level and/or sites of brain's central action may eventually lead to the development and perpetuation of T2DM and MS . The intricate interactions between adiponectin and resistin with catecholamines may play an integral role in metabolism . In addition, adipocyte-derived microvesicles mediated transport of adiponectin and resistin gene transcripts into macrophages and might play a role as a novel intercellular communication tool by transporting RNA in paracrine and possibly endocrine manners . Miyamoto et al. found that resistin may increase the susceptibility of metabolic syndrome by modulating adiponectin secretion from adipocytes . Resistin may enhance hepatic gluconeogenesis, presumably by antagonizing adiponectin, which inhibits enzymes involved in gluconeogenesis through AMPK activation . Furthermore, SNP-420C > G of the resistin gene was associated with lower circulating adiponectin levels in a Japanese cohort study . Thus, the transcriptional activity of the resistin gene may also influence circulating adiponectin levels.
Adiponectin  and resistin  hormones are considered significant root factors for the regulation of energy, glucose, and lipid homeostasis as well as insulin signalling pathway. Moreover, it has been reported that the overall structure of multimeric assembly of the resistin is similar to that of adiponectin . Both have been characterized as coiled-coil trimers that formed tail-to-tail hexamers through disulfide bonds near their amino termini . Furthermore, both of these hormones circulate in serum in two distinct assembly states . The comparable domain architecture of these two adipocyte-specific hormones, despite having diametrically opposed physiological effects, suggested a common regulatory mechanism in metabolic homeostasis .
The adiponectin-resistin (AR) index
It has been reported that the adiponectin-resistin ratio might be potentially useful in prediction of the future cardiovascular risk in women with the polycystic ovary syndrome . Moreover, changes of the relative proportion of adiponectin to resistin might play a more important role in hormonal disturbances in polycystic ovary syndrome than the absolute concentrations of these adipokines . In addition, mice under chronic variable stress and fed with a high-fat diet showed impaired glucose tolerance associated with low plasma adiponectin-resistin ratios . It seems that changes of circulating adiponectin and resistin levels may be the effect of their mutual interaction in adipose tissue. Thus, the AR index that included information on both serum adiponectin and resistin levels may has a more integrated and concentrated explanation than single measure of serum adiponectin and resistin levels in the present study.
Taking these studies together, adiponectin and resistin may be useful markers for insulin resistance and the variables that can integrate the abnormalities of the metabolic syndrome and cardiometabolic function. For this reason we attempted to estimate a threshold for the AR index for the identification of T2DM and MS. Although further studies may be necessary to confirm the efficacy of periodically measuring AR index in the management of insulin resistance, MS and T2DM, our study certainly highlights the potential for the AR index to move one step closer to becoming an established biomarker of the metabolic status. The evaluation of intervention strategies can be facilitated and strengthened by the use of the AR index that measure biological parameters of disease progression and therapeutic response. Routine assessment of the AR index may allow for a better understanding of the underlying disease conditions and optimization of anti-diabetic therapy targeting beyond simple glycemic control. Thus, the AR index has a potential for routinely available in general clinical practice and make a meaningful contribution to patient care.
The insulin resistance (IRAR) index
Given the complicated nature of the euglycemic hyperinsulinemic clamp technique and the potential dangers of hypoglycemia in some patients, alternatives have been sought to simplify the measurement of insulin resistance. In recent years, several markers have been proposed for the screening, diagnosis, and therapeutic monitoring of insulin resistance. However, all have problems that limit their use to research studies. None succeeds in integrating the global assessment of the metabolic abnormalities that may increase risk for developing type 2 diabetes (T2DM) and metabolic syndrome (MS).
Compared to other classical insulin resistance indexes, quantitative insulin sensitivity check index (QUICKI) was reported to have the advantage of being applicable to wider ranges of insulin sensitivity and more reproducible . It has been showed that QUICKI was among the most accurate and useful surrogate indexes for determining insulin sensitivity in humans . However, QUICKI and insulin action do not correlate highly, particularly in individuals with mildly insulin resistance, impaired glucose tolerance or elderly patients with poorly controlled T2DM . Moreover, QUICKI is less robust for early diagnosis of insulin resistance in persons without T2DM or MS . QUICKI has low sensitivity for detecting insulin resistance in lean individuals with beta cell dysfunction . QUICKI use fasting glucose levels in their calculations and fasting glucose levels are steady-state levels that are not a reflection of glucose utilization after a glucose load . Also, QUICKI reflects hepatic insulin resistance only, not insulin resistance at peripheral tissues .
The recently developed HOMA-AD was a more accurate indicator for assessing insulin resistance than the HOMA-IR . HOMA-AD is a modified version of homeostasis model assessment of insulin resistance (HOMA-IR) index which calculated from the product of serum insulin and plasma glucose levels divided by serum adiponectin levels . Modification of HOMA-IR with adiponectin levels resulted in an index exhibiting a good correlation with M-values even in diabetic patients with moderate hyperglycemia . In addition, Zaletel et al. showed that the adiponectin derived index correlated best with the euglycemic hyperinsulinemic clamp derived sensitivity index compared to other surrogate measures of insulin resistance including HOMA-IR, QUICKI, fasting glucose/insulin ratio or McAuley index . A recent electron spin resonance study revealed adiponectin might has a close correlation with rheological behavior and microcirculation in hypertension . Moreover, adiponectin may be a marker for global metabolic status including insulin resistance and metabolic syndrome .
It is well-established that insulin resistance in adipose tissue will lead to elevated serum resistin levels and reduced serum adiponectin levels . Lipolysis plays a role in the developing of insulin resistance in healthy subjects, with an estimated overall contribution of approximately 39% . Increased lipolysis in adipose tissues was associated with elevation of systemic free fatty acids and insulin resistance . Adiponectin in physiological concentrations inhibits spontaneous as well as catecholamine-induced lipolysis . Resistin induces lipolysis and re-esterification of triacylglycerol stores, and increases cholesteryl ester deposition, in human macrophages . Therefore, inclusion of adiponectin and resistin into the QUICKI formula can be beneficial and can increase its detection power by including those subjects with peripheral insulin resistance, especially in view of the following: increased fasting resistin levels  and reduced fasting adiponectin levels  could reflect insulin resistance earlier than hyperglycemia since lipolysis was more sensitive to insulin than glucose utilization; a small increase in resistin  and decrease in adiponectin levels  in healthy individuals were reported to induce insulin resistance; insulin resistance of lipolysis induced by adiponectin and resistin were suggested as explaining a large variation in insulin sensitivity of glucose disposal in lean individuals .
It has been well-reported that adiponectin  and resistin  are promising biomarkers of insulin resistance. A recent clinical trial revealed pioglitazone plus vildagliptin treatment improved both adiponectin and resistin levels and might effective in preserving beta-cell function, and in reducing insulin resistance and inflammatory state parameters in subjects with poorly controlled T2DM . Moreover, the imbalance in deleterious and protective adipokines including adiponectin and resistin plays pivotal roles in the development and progression of pancreatic beta-cell dysfunction under insulin-resistant conditions . Adiponectin and resistin levels were also strongly correlated with the key metabolic endpoints of T2DM and MS as well as insulin sensitivity in the present study (Table 3 and Table 4). Therefore, we generated a modified version of QUICKI, the novel IRAR index by taking account into adiponectin and resistin levels. The IRAR index clearly had the narrowest and most favourable distribution of residuals among the surrogate indexes (Table 6 and Figure 4). The IRAR index may has higher precision, consistency, reproducibility and robustness than classical surrogate indexes (Table 6 and Figure 4). Thus, we hypothesized that the IRAR index may be a more appropriate model of insulin sensitivity than other surrogate indexes in Malaysian men.
The reported values for the definition of insulin resistance vary widely. A World Health Organization (WHO) consensus group concluded that the insulin sensitivity index of the lowest 25% of a general population can be considered as an insulin resistance state . The European Group for the Study of Insulin Resistance take a more restricted view, defining insulin resistance as the insulin sensitivity index of the lowest 10% of a nonobese, nondiabetic, normotensive white population . Therefore, further investigation or replicate studies are required to validate the suggestive reference range or minimum cut-off values of insulin sensitivity for the IRAR index in Malaysian men. In addition, long-term prospective studies are required to determine the actual reference range or minimum cut-off values of the novel IRAR index for insulin resistance assessment in the general population. The novel IRAR indexes give an opportunity to implement and use of this index in the daily clinical practice for screening persons with increased risk of future development of T2DM and MS due to insulin resistance. It is also very useful for monitoring the diseases progression and therapeutic response. Furthermore, it will allow early treatment or delay the onset of long-term severe complications such as cardiovascular risk.
Strengths and limitations
This represent a first attempt to study the interaction effect of adiponectin and resistin in the modulation of the key metabolic endpoints of T2DM and MS. Our samples comprised of Malay, Chinese and Indian subjects from Malaysia, which represented a major segment of the Asian population. The covariates in term of ages and ethnicity were in homogeneity and were matched with the case-control groups (Table 1). Most potential confounders were carefully controlled for, which limits the possibility of residual confounding effect. Given the well-established difference in circulating adiponectin [8, 67] and resistin  levels between men and women, our samples were only comprised male subjects to avoid the confounding effect of gender. Clinical measurements were taken under standardized protocol and biomarkers were measured using assays with good precision (Additional file 1, Figure S1 and Additional file 1, Supplementary Methods). Although it has been reported that various definitions of MS hold different predictive powers in detecting pathological levels of key adipocytokines such as hypoadiponectinemia and hyperresistinemia, the International Diabetes Federation (IDF) definition is quantitatively more powerful than its counterparts in terms of prevalence [4, 78]. The IDF definition incorporates ethnicity by providing different criteria for the MS in different ethnic groups . Moreover, the IDF definition is the most updated and globally accepted definition for MS .
Nevertheless, this study had limitations. The findings apply mainly to Malaysian men and may not be widely generalizable because of the homogeneity of the study population. Since this is not a prospective study, this study may have reverse causation due to possible effects of T2DM and MS on adiponectin and resistin levels. The results were based on single measurements of the adipokines and therefore may not reflect long-term exposure to these hormones. Although adjusted for known confounding factors, residual confounders imperfectly measured or unmeasured cannot be excluded.
We measured total adiponectin and not the high molecular weight fraction, which has been proposed to have substantially more potent effects on hepatic insulin sensitivity compared with total adiponectin . However, a recent study showed that total and high-molecular-weight (HMW) adiponectin have similar utility for the identification of insulin resistance and metabolic disturbances . This suggested that total adiponectin levels may provide clinical information of the same diagnostic value as HMW adiponectin . It has been reported that the low molecular weight form of resistin displays significantly increased bioactivity . However, the high molecular weight hexamer of resistin is predominant in the human circulation . Moreover, the measurement of total adiponectin and resistin are better standardized, cheaper and more accessible than the high-molecular-weight adiponectin and the low-molecular-weight resistin. Thus, our findings may stimulate the use of adiponectin and resistin in clinical and epidemiological settings.
The euglycemic hyperinsulinemic clamp technique must be used as a gold standard index of insulin resistance to validate the robustness of the IRAR index before their widespread use. Further studies are required to investigate whether the superior predictive power of the AR and IRAR indexes demonstrated in the present study translates into a significant clinical benefit. The AR and IRAR indexes are unlikely to be cost-effective for short-time administration of these indexes. However, the AR and IRAR indexes are good indicator for long-term metabolic status. Thus, it will be very cost-effective for long-time administration since the frequency of monitoring the disease progression and therapeutic response periodically will be greatly reduced. A long-term prospective research is needed to reveal the predictive value of the AR and IRAR indexes for insulin resistance in association with T2DM and MS, and to find their optimal cutoff values for future risk assessment and disease prevention. Moreover, the normal AR and IRAR indexes range need to be established for each laboratory with an appropriate control group because of significant inter-laboratory variations in insulin and adipokines (including adiponectin and resistin) determinations and/or possible differences in various populations.