Physical and metabolic characteristics of the 144 participants are described in Table 2. Classification of ISO vs. IRO individuals according to quartiles of HEC results, whole body IS indices (SIisOGTT and Matsuda) and muscular IS index (Abdul-Ghani) are presented in Table 3. Using classification according to the HEC results, ISO and IRO individuals exhibited similar BMI (ISO: 32.4±4.3.4 kg/m², IRO: 33.9±3.4 kg/m²; p=0.19) and total FM (ISO: 38.9±7.3 kg, IRO: 41.3±7.4; p=0.16). Similar results were obtained with SIisOGTT, Matsuda Index and Abdul-Ghani muscle IR index. On the other hand, the IS classification based on HEC demonstrated that ISO individuals displayed significantly lower WC (ISO: 101.0±7.3cm, IRO: 108.1±8.9cm; p=0.001) and VAT area (ISO: 172.9±50.8cm², IRO: 222.5±59.5 cm²; p<0.001). Classification based on the SIisOGTT or the Matsuda Index quartiles showed similar results.

However, using the classification of IS according to the Abdul-Ghani muscle IR index, groups were comparable for WC (p=0.13) even though IRO individuals displayed significantly higher VAT area than ISO individuals (p<0.001).

Considering glucose homeostasis, fasting insulin and 2h-OGTT glucose and insulin levels were significantly lower in ISO vs IRO individuals defined with HEC results, SIisOGTT, Matsuda or Abdul-Ghani muscle IR indices. However, fasting glucose was lower in ISO individuals only when ISO was defined using SIisOGTT or Matusda index. ISO individuals displayed more favourable lipid and inflammatory profiles than IRO individuals. No significant difference was observed for systolic and diastolic blood pressure whatever the quartiles classification. Elevated alanine aminotransferase (ALT) values were observed in IRO individuals defined with HEC results or global IS indices. Finally, ISO individuals demonstrated significant lower FLI, LAP and VAI compared to IRO individuals whatever the quartiles classification.

Table 4 shows ISO vs. IRO classification according to liver IS surrogate indices. Using the HOMA-IR index, IS groups displayed similar FM but different BMI (p=0.047). Similar differences in cardiometabolic parameters between ISO and IRO individuals were observed with the Abdul-Ghani liver IS index and HEC results quartiles classifications except for fasting and 2h-OGTT glucose. As shown in the Table 5, individuals were then defined as ISO and IRO depending on the median cut-off value of HEC results, Matsuda index, SIisOGTT and Abdul-Ghani muscle and liver index. ISO and IRO individuals were matched for BMI, total FM as well as for WC whatever the median value classification but ISO individuals displayed significantly lower VAT area than IRO individuals (from p<0.05 to p<0.001). Cardiometabolic markers differences observed in quartile value classification (Table 3 and Table 4) remain significant between ISO and IRO individuals whatever the median value classification.

A ROC curve analysis was performed (Figure 1) to determine the ability of IS surrogate indices to discriminate ISO vs. IRO individuals using HEC as the reference method and the median GIR value as cut-off (median=11.59 mg/min/kg of LBM). Matsuda and SIisOGTT ROC AUCs were similar (0.73± 0.04 and 0.74± 0.04; p=0.57, respectively) and significantly higher than Abdul-Ghani liver and muscle indices AROCs(0.65± 0.05 and 0.63±0.05, respectively), which were equivalent (p=0.62). The best cut-off values were 2.5 for Matsuda index (sensitivity: 54.2% and specificity: 83.1%) and 0.26 for SIisOGTT (sensitivity: 70.8% and specificity: 64.8%).

The aim of the present investigation was to determine the ability of surrogate indices of IS to identify ISO vs. IRO individuals in a postmenopausal overweight and obese women population with similar BMI and total FM but with a large variation in IS according to the classification obtained using the gold standard HEC technique.

Comparison of ISO vs IRO individuals is useful to investigate cardiometabolic risk markers abnormalities related to IR regardless of the confounding effect of major differences in weight/adiposity. The relationship between body fat distribution and IS is well established. It was demonstrated in a sample of morbidly obese individuals (BMI = 45 ± 1.3 kg/m²) that independently of BMI and total FM, increased VAT area was associated with IRO obesity (Kloting et al. 2010). Accordingly, the LAP index was significantly higher in IRO than ISO individuals. Indeed, LAP has been demonstrated to be closely related to IR and reflect increase in WC and TG over time (Xia et al. 2012). We also found that IRO individuals presented significantly higher fat accumulation in the liver as estimated by FLI than ISO individuals.

Using HOMA-IR to define the quartile of IRO individuals, we observed that IRO and ISO individuals were not matched for BMI and total FM indicating that differences observed in cardiometabolic risk markers could be partly attributed to the difference in FM between groups. The HOMA-IR is a simple surrogate index of IR requiring only fasting insulin and glucose levels. Due to its simplicity, this index is widely used to estimate IR for research purposes and in clinical practice. Using a HOMA-IR cut-off value ≥2.5, Calori et al. (Calori et al. 2011) categorized ISO and IRO individuals with similar BMI. However, IRO individuals were older, and 28% had type 2 diabetes. In another study using the same HOMA-IR cut-off value, ISO and IRO individuals categorization leaded to significant different BMI between groups (Kuk and Ardern 2009). For both studies, neither sex nor FM measurements were considered to match the individuals. Discrepancies between results might be due to differences between studies samples and the absence of standardisation for insulin assay (Antuna-Puente et al. 2011). Categorization of ISO and IRO individuals using quartiles of Abdul-Ghani liver IS index values, leaded to similar BMI and body FM between groups. These anthropometric parameters are not included in the index formula. Thus, Abdul-Ghani liver IR index is probably a relevant liver IS surrogate index to identify hepatic IR among obese individuals independently of fat mass.

The Abdul-Ghani muscle IR index demonstrated a good ability to discriminate ISO vs. IRO groups for comparable BMI and total FM. Muscle is considered as the major site of insulin-stimulated glucose uptake and has an important contribution to GIR (Kahn and Flier 2000). We then speculated that the accumulation of ectopic lipids in muscle could have impaired insulin signalling and lead to IR (Kelley and Mandarino 2000).

The two whole body IS surrogate indices showed a reasonable ability to define ISO vs. IRO individuals matched for BMI and total FM. Using a ROC analysis, these IS indices exhibited better performance than Abdul-Ghani liver and the muscle IS indices to identify IRO individuals. The higher sensitivity and specificity of SIisOGTT could be due to the fact that this index has been developed and validated against the HEC results in 107 participants included in the present study cohort. However, the Matsuda index, which validation was determined in another cohort, showed similar results as the SIisOGTT. Moreover, it is interesting to note that both SIisOGTT and Matsuda index have been recently proposed as reliable index to predict IS in non-diabetic population (Pisprasert et al. 2012). In our study, IRO individuals demonstrated higher estimated liver fat accumulation than ISO individuals. A previous cross-sectional investigation showed that ISO individuals had less liver fat (direct measure) than IRO subjects and the two groups were distinguished on the basis of lipid accumulation in liver but not subcutaneous or visceral fat (Stefan et al. 2008). Moreover, Fabbrini and et al. (Fabbrini et al. 2009), reported that intrahepatic triglyceride content was associated to IR and increased TG secretion. Surprisingly, we found that liver IR indices were less effective than the muscle and whole body indices to differentiate these two groups. Our results could be explained by a possible disconnection between liver fat and IS. It should also be emphasized that we performed the HEC technique using relatively high dose of insulin. Thus, our measurement of whole body insulin probably reflected more skeletal muscle than liver glucose utilization. Indeed, HEC performed with classical low insulin rate (i.e. 40 MU/m2/min) are more likely to show liver IS.

The correlation of adipokines such as adiponectin, resistin and leptin with visceral obesity as well as IR is now well established. We could therefore speculate that the prediction of IS/IR by the surrogate indices of IR might be strengthened by adding those biomarkers in the algorithms. Indeed, a recent sub study using as sample of postmenopausal overweight and obese women from the MONET population has shown promising results by using indices integrating adipokines to estimate IR/IS (Vatier et al. 2017). However, future studies are needed on a wider scale and more diverse populations to validate those indices.

We acknowledge several limitations of our study. IS indices were used to discriminate ISO vs. IRO groups of individuals without allowing defining single individuals as IS or IR obese. Our sample was composed of non-diabetic, obese and overweight postmenopausal women, limiting our conclusions to this population. Indeed, we identified cut-off points for Mastsuda and SIisOGTT indices but they need to be confirmed by independent research teams in other cohort of subjects including men and using other reliable insulin measurement s kits allowing the surrogate indices calculation. Nevertheless, such biological tool is particularly interesting in non-diabetic overweight and obese subjects susceptible to be ISO or IRO. We performed HEC with high dose of insulin and therefore might be more consistent with muscle glucose disposal rate. However, many previous studies found similar excellent correlation between HEC tests with lower insulin infusion dose and the best surrogate indices highlighted in the present study (Patarrao et al. 2014). In addition, most indices used are based on stimulated glucose and insulin concentrations and thus results can be confounded by endogenous insulin secretion as well as variable insulin clearance. However, this is also the case for fasting indices whose do not permit to clearly make the difference in IR, IS and insulin clearance, which are all inter-related in insulin resistance subjects. Hepatic fat infiltration has been estimated and not directly measured. Moreover, it should be stressed that the cut-off numerical values of Matsuda and SlisOGTT indexes reported in this paper are valid only in our lab. Both Matsuda and SlisOGTT indices heavily depend on insulin concentration values, and unfortunately the insulin assay is still not standardized (Borai et al. 2010).Thus, validating cut-off values will need both the standardization of insulin assay as well as prospective cohorts with accepted cardiometabolic end-points. It should also be noticed that we do not have a comparison group of a non-obese non-diabetic healthy postmenopausal women. Thus whether the ISO group based on HEC technique displayed similar insulin sensitivity with such control group remains to be clarified. However, in a previous independent study among non obese non diabetic healthy individuals, authors found similar mean value of glucose infusion rates when compared to our ISO group (Pisprasert et al. 2012). Finally, the cross-sectional design of our study does not allow to determine any causal association between the IS status and related cardiometabolic risk markers independently of BMI and total FM. Longitudinal studies may improve this point in the future. Nevertheless, our results are strengthened by the use of gold standard HEC for IS measurement as well as CT-scan for VAT quantification.

In conclusion, our results confirmed that despite similar BMI and total FM, IRO patients exhibited more VAT accumulation and cardiometabolic risk markers than ISO individuals. Moreover, some surrogate indices of IS/IR are not valid index to identify ISO and IRO individuals because the presence of cardiometabolic risk factors need to be assessed independently of BMI and total fat mass. In our sample of postmenopausal obese and overweight women, despite some limitations, whole body surrogate indices of IS are more in line with the classification based on the golden-standard GIR results compared to regional IS surrogates indicators.