Limited predictive value of the IDF definition of metabolic syndrome for the diagnosis of insulin resistance measured with the oral minimal model

ARTICLE

Auteur(s) : E Ghanassia, E Raynaud de Mauverger, J-F Brun, C Fedou, J Mercier

Institut Ceramm, Hôpital Lapeyronie, CHRU de Montpellier

Article reçu le 10 Avril 2009, accepté le 19 Juin 2009

Insulin resistance (IR) is defined as a defect in insulin action at the level of insulin sensitive tissues or organs. Abdominal obesity and sedentarity, as well as puberty, pregnancy or aging [1], are some recognized factors of IR. Considered as the main pathological mechanism in the early stages of type 2 diabetes, it is also very common in non diabetic subjects, whose insulin secretion may adapt the increasing demand caused by IR. Insulin resistance syndrome is currently defined as a cluster of clinical and biological abnormalities, including identified syndromes as polycystic ovaries syndrome or non-alcoholic steatohepatitis, and whose prevalence is correlated to insulinaemia and inversely correlated to insulin sensitivity (S_I). S_I is a continuous and highly variable parameter among a given population [1]. There is currently no consensus on the value of S_I defining IR; however, according to most studies, the lowest quartile of S_I distribution in a given population is generally considered as insulin resistant enough to significantly develop abnormalities or syndromes characterized by IR [2].

Since the description of ‘syndrome X’ by Reaven [3], the use of the complex and costly euglycaemic hyperinsulinaemic clamp technique as the “gold standard” for the determination of S_I is poorly adapted to large epidemiological studies and everyday clinical practice. Many indexes, as the popular HOMA model, were proposed to estimate S_I. Unfortunately, they were proved to have several limitations, particularly when a defect in insulin secretion was associated [4, 5].

In order to easily diagnose IR, many definitions of Metabolic Syndrome (METS) were suggested but, in a public health perspective, they were aimed at preventing cardiovascular diseases (CVD) and type 2 diabetes (T2D). Many definitions have been proposed, from the original WHO to the most widely employed NCEP-ATP III criteria. They all included the main features of insulin resistance but they also differed in cut-off values and the respective importance of each item. In order to harmonise these various definitions, the International Diabetes Federation (IDF) published a worldwide consensus in 2005 [6]. Abdominal obesity became the main criterion with cut-off values of waist circumference depending on ethnic group and fasting glycaemia considered abnormal over 5.5 mmol/L.

Despite their many common features, all these evolutions made METS and Insulin Resistance Syndrome diverge into somewhat different entities. Unfortunately a confusion remains between those two as they appear to be considered interchangeable in many publications. Consequently, several authors, including Reaven himself, attempted to rehabilitate Insulin Resistance Syndrome, to distinguish it from METS, and to criticize the current use of the latter, sometimes announcing its next disappearance [7].

In a similar, though more temperate perspective, we hypothesized that the two most frequently used definitions were perhaps not systematically well correlated to IR and were likely to identify subjects whose insulin sensitivity is uncertain. This study was thus undertaken in order to compare the prevalence of METS according to NCEP ATP-III and IDF definitions and to determine their predictive values for the diagnosis of IR.

Methods

The WC was measured at the part of the trunk located midway between the lower costal margin (bottom of lower rib) and the iliac crest (top of pelvic bone) while the person was standing, with feet about 25-30 cm apart (10-12 in), without compressing any underlying soft tissues, at the end of a normal expiration.

Blood samplings were obtained after clinical examination and immediately centrifuged. Plasma glucose, cholesterol, triglycerides, HDL and LDL levels were determined with an Olympus 2700 Chemistry Analyzer, using routine well-standardized procedures. Plasma insulin levels was determined using RIA methods (kit Bi-Insuline IRMA; Schering CIS bio international, Gif-sur-Yvette, France).

We used the two most recent definitions of METS. According to NCEP-ATP III, subjects must meet at least three of the five following criteria: abdominal obesity (defined as WC ≥ 102 cm for Caucasian men and WC ≥ 88 cm for caucasian women), raised TG level (> 1.7 mmol/L) or specific treatment for this lipid abnormality; reduced HDL cholesterol (< 1 mmol/L in men and < 1.3 mmol/L in women) or specific treatment for this lipid abnormality; systolic blood pressure ≥ 130 mmHg and/or diastolic blood pressure ≥ 85 mmHg) or treatment of previously diagnosed hypertension; raised fasting plasma glucose (≥ 6.1 mmol/L) or previously diagnosed type 2 diabetes.

According to International Diabetes Federation (IDF), subjects must have abdominal obesity (defined as WC ≥ 94 cm for Caucasian men and WC ≥ 80 cm for Caucasian women, with ethnicity specific values for other groups) associated with any two of the following factors: raised TG level (> 1.7 mmol/L) or specific treatment for this lipid abnormality; reduced HDL cholesterol (< 1 mmol/L in men and < 1.3 mmol/L in women) or specific treatment for this lipid abnormality; raised blood pressure (systolic blood pressure ≥ 130 mmHg and/or diastolic blood pressure ≥ 85 mmHg) or treatment of previously diagnosed hypertension; raised fasting plasma glucose (≥ 5.6 mmol/L) or previously diagnosed type 2 diabetes.

Subjects had their breakfast after clinical examination and blood sampling collection. It included bread (80 g), butter (10 g), jam (20 g), milk (80 mL), sugar (10 g) and instant coffee (2.5 g). Caloric intake was around 2070 kg (9.1% proteins; 27.5% lipids; 63.4% carbohydrates). Mean duration of food intake was 6 min. Blood samplings for plasma glucose and insulin were collected twice before breakfast and 15, 30, 60, 90, 120, 150, 180, 210 and 240 min after the beginning of the meal.

Caumo [8] extends Bergman’s minimal model computation to the analysis of meal test, such as this Standardized Hyperglucidic Breakfast (SHB), and provides an evaluation of S_I that can actually be calculated with a Microsoft Excel workbook according to the formulae published by its authors in their original paper [8]. It is based on the analysis of changes in plasma glucose and insulin concentration measured after the SHB. SI is given by the “oral minimal model” which is actually Bergman’s one with simply another term called Ra _OGTT added to the first equation. Model equations are thus [8]:

Where G is plasma glucose concentration, I is plasma insulin concentration, suffix “b” denotes basal values, X is insulin action on glucose production and disposal, V is distribution volume, and S_G, p₂, and p₃ are model parameters. Specially, S_G is the fractional (i.e., per unit distribution volume) glucose effectiveness, which measures glucose ability per se to promote glucose disposal and inhibit glucose production; p₂ is the rate constant describing the dynamics of insulin action; p₃ is the parameter governing the magnitude of insulin action. Interestingly, these two equations can be simplified, allowing calculating S_I with a quite simple area under the curve formula:

where G is plasma glucose concentration, ΔG and ΔI are glucose and insulin concentrations above basal, respectively, AUC denotes the area under the curve; GE (also called S_G) is glucose effectiveness (min^-1 x 10^-2); D_OGTT is the dose of ingested glucose per unit of body weight (mg.kg^-1); and f is the fraction of ingested glucose that actually appears in the systemic circulation. When glucose falls below basal, a slightly different formula needs to be used (we refer to equation n°7 in Caumo et al. [8]). Calculations of S_I require insertion of values for S_G and f. Here we used the value of glucose effectiveness given by our previously validated formula S_G =2.921 e ^-0.185(G₆₀^-G₀⁾ [9]. Besides, as in Caumo’s paper, the value for f is set as f=0.8.

Caumo’s model is the extension to oral glucose tolerance tests of Bergman’s equations initially developped for the intravenous glucose tolerance test (IVGTT). It has been shown to be suitable for calculating S_I during this hyperglucidic breakfast test [9].

Control values with this technique are available on large databases and on 537 controls we defined the threshold for insulin resistance as the limit of lower quartile i.e., S_I < 3.27 min^-1.μU^-1.ml^-1.

Normality of each continuous variable’s distribution was assessed using the Shapiro-Wilk test. When normality was established, results were given as mean ± SD whereas they were given as median in the opposite case. Results for discontinuous and qualitative variables were given as frequency and expressed in percents. For both METS definitions, we determined their sensitivity, specificity, positive and negative predictive value for the diagnosis of IR. All calculations were performed with the software StatBox Pro 6.0.

Univariate analysis of correlation between discontinuous variables was performed using the chi-square test corrected with Yates’ continuity test. Concerning continuous variables, if normality was established, we used the Student’s t test. If normality was not established, we used the non-parametric Mann-Whitney’s U test. Significance was set at p <0.05.

The specificity was calculated as the number of truly negative subjects divided by the sum of false positive and true negative. The positive predictive value was calculated as the number of truly positive subjects divided by the sum of true positive and false positive. The negative predictive value was calculated as the number of truly negative subjects divided by the sum of true negative and false negative ones. The likelihood ratio of a positive test result (LR+) is sensitivity divided by 1- specificity. The likelihood ratio of a negative test result (LR-) is 1- sensitivity divided by specificity [10].

Results

Prevalence of METS was 37.4% according to NCEP and 40% according to IDF. This difference was not significant, even after adjustment for gender.

89 subjects did not meet the NCEP and IDF criteria and 55 subjects met the criteria for both definitions. Consequently, concordance rate was 96%. 6 subjects met the criteria for only one definition: 5 subjects for the IDF criteria and one only for the NCEP criteria. On figure 1 are shown for example glucose and insulin profiles according to number of METS IDF criteria: this figure only shows trend to more severe impairment of glucose tolerance, as the METS score increases, but individual values of glucose and insulin have been computed with the Oral Minimal Model in order to calculate S_I.

There was no difference in S_I levels between genders. According to the definition we used for IR, 25% of subjects were considered insulin resistant (table 1) according to the cut-off value (upper limit of the lower quartile of S_I) which was 3.27 min^-1.μU^-1.ml^-1 in the control sample. Using NCEP as well as IDF definition, positive predictive value (PPV) for the diagnosis of IR increased with the number of criteria met (table 2). Sensitivity, specificity, positive predictive value PPV, negative predictive value (NPV) and likelihood ratios are reported in table 3.
Table 1 Subjects characteristics and prevalence of each criterion, of metabolic syndrome according to NCEP and IDF definition and of insulin resistance.

Discussion

Few studies investigated the same topic and their results are heterogeneous. In a cohort of 3000 Australian Caucasian subjects, Adams et al. report a prevalence of 22.8% (IDF) vs 15.7% (NCEP) with the lowest concordance between men and older women [11]. On the opposite, Guerrero-Romero et al. investigated 700 Mexican subjects [12] aged 30 to 64 and report a prevalence of 22.3% (IDF) vs. 22.6% (NCEP). The most important study was undertaken by Assmann et al. [13] who used the data from three cohorts (DHS, NHANES, PROCAM). The lowest concordance rate was also found between men but when ethnic origin was considered, concordance was over 95% fort DHS Caucasian women and NHANES Caucasian subjects. On the whole, our results are in agreement with these reports, illustrating such a concordance between these definitions that one might wonder if a new definition was really useful.

Our second goal was to determine the predictive power of METS for the diagnosis of IR quantified with a laboratory measurement. To assess insulin resistance, we used a standardized breakfast test with minimal model analysis as previously published [9]. This recently introduced approach has been validated in comparison with various well recognized and more sophisticated techniques such as the glucose clamp [8, 14-16] and the minimal model analysis of an intravenous glucose tolerance test [9], and even more with sophisticated tracer experiments [17]. Compared to OGTT or IVGTT, it is appealing because of its simplicity and the fact that this is a fully physiologic stimulus avoiding artefactual hypoglycaemia, and that it can be employed quite easily in large samples of patients. Since we had a large database of previous breakfast tests in control subjects, athletes and patients since 1988, we could define on 537 controls a cut-off value defining insulin resistance. The threshold for insulin resistance was therefore defined here as the upper limit of lower quartile i.e., SI < 3.27 min^-1.μU-1.ml^-1.

In this study we do not investigate insulin secretion, although it is interesting to study it in connection with insulin resistance. It is well known that insulinemia increases for compensating insulin resistance, in order to keep the product insulin sensitivity x insulin secretion constant, so that the relationship between insulin secretion and insulin sensitivity physiologically displays an hyperbolic shape [16]. This homeostatic loop is often disrupted in patients with glucoregulatory disorders, explaining the poor reliability of surrogates like the homeostasis model assessment in patients whose glucose tolerance is impaired [18]. In this study we only focus on insulin resistance.

Compared to this measurement, in our sample of patients, both definitions of the metabolic syndrome appear to have quite the same sensitivity, specificity or positive and negative predictive value, although there may be a slight tendency toward a superiority for the IDF definition. Sensitivity of both definitions was poor (NCEP vs. IDF, 56% vs 63%) while specificity was somewhat higher (NCEP vs IDF, 68% vs 69%).

To our knowledge, our study is the first to investigate the predictive power of the IDF definition as the only available data deal with the NCEP definition and yield very different results depending on the population studied. Egan et al. report results comparable to ours with sensitivity of 56.6% and specificity of 73.2% [19]. In contrast, other studies of north-american and hispano-american populations [20-23] showed a lower sensitivity (42 to 61%) and a higher specificity (83.5 to 94%). This can be explained by the fact that cut-off values used in the NCEP definition were determined among the same ethnic groups. Moreover, IR was calculated using various techniques, including HOMA-IR, euglycaemic hyperinsulinemic clamp or Bergman’s Minimal Model after IVGTT, and caution for pooling such results obtained with different techniques is surely required. Notwithstanding, in all these studies, as well as in ours, METS sensitivity is quite poor, proving it to be inadequate as a screening or diagnosis test for IR.

Whatever the definition used, PPV was very low (NCEP vs IDF, 37.5% vs 40%). However, it tended to increase with the number of criteria met, especially with NCEP definition. When all the criteria were met, the latter had a higher PPV (NCEP vs IDF, 60% vs 50%). It must also be noticed that the most frequent combinations of NCEP criteria have the lowest PPV (7 – 33.3%) whereas those of IDF criteria have the highest PPV (60-100%). Other studies report higher values, with a PPV higher than 70% in two of them [20-22].

By contrast, predictive negative value of both definitions of METs for diagnosing insulin resistance appears to be somewhat better (81.9 and 84.5% for respectively IDF and NCEP definitions. This means that individuals lacking METs are unlikely to be insulin resistant, and thus that screening them with this clinical evaluation may be useful for excluding patients with normal values of SI. As seen on table 3, the calculated likelihood ratio for negative results is 0.651 with the NCEP definition and 0.55 with the IDF definition. According to Bayes’ theorem states the pre-test odds of disease multiplied by the likelihood ratio yields the post-test odds of disease [24], i.e., if we assume that 25% of a population is insulin resistant, an individual in whom clinical scoring does not detect a METs has only an chance of 22% (NCEP definition) or 18% (IDF definition) of being insulin resistant. Therefore, to some extent, a negative clinical score of METs may have some usefulness for excluding the diagnosis of insulin resistance.

One can wonder whether it is important to have accurate diagnostic tools for insulin resistance as defined above on biological grounds, i.e. a low value of SI. It has been emphasized that SI is a continuous variable without any well-defined pathological threshold. The concept of IRS was proposed in order to unify apparently independent diseases around a common mechanism in an understanding and, consequently, therapeutic purpose [1]. This unifying concept has been challenged over the last years but is still supported by a large and ever growing body of evidence [25]. Moreover, regarding recent advances in the use of insulin-sensitizing strategies (physical activity) or drugs (metformin) in the treatment or prevention of T2D or polycystic ovaries syndrome [23], it is likely that a simple way for determining IR in patients may be, if it were available, of some clinical interest. Nevertheless, the measurement of IR is currently restricted to clinical research, as it remains not demonstrated that it could improve the management of IR-associated diseases. Other studies have already demonstrated that NCEP criteria were unable to correctly predict IR. Our study gives the same conclusion and provides the first evidence in the literature that the IDF definition has rather the same level of accuracy for determining the level of SI as the NCEP definition.

Conclusion

We conclude that, whatever the definition used, if one aims at diagnosing insulin resistance, METS is a rather inadequate tool, yet it may be of some help, with caution, for a population screening. Its only interest at the individual scale is to provide a rather low probability of being insulin resistant when criteria are not met, especially using the IDF definition. All this means that laboratory measurements (preferably a dynamic test) are mandatory for the diagnosis of insulin resistance.

References

2 Yeni-Komshian H, Carantoni M, Abbasi F, Reaven GM. Relationship between several surrogate estimates of insulin resistance and quantification of insulin-mediated glucose disposal in 490 healthy nondiabetic volunteers. Diabetes Care 2000; 23: 171-5.

3 Banting lecture. Role of insulin resistance in human disease. Diabetes 1998; 37: 1595-607.

4 Malita FM, Karelis AD, St-Pierre DH, Garrel D, Bastard JP, Tardif A, et al. Surrogate indexes vs. euglycaemic-hyperinsulinemic clamp as an indicator of insulin resistance and cardiovascular risk factors in overweight and obese postmenopausal women. Diabetes Metab 2006; 32: 251-5.

5 Wallace TM, Levy JC, Matthews DR. Use and abuse of HOMA modeling. Diabetes Care 2004; 27: 1487-95.

6 Alberti KG, Zimmet P, Shaw J. IDF Epidemiology Task Force Consensus Group. The metabolic syndrome, a new worldwide definition. Lancet 2005; 366: 1059-62.

7 Reaven GM. The metabolic syndrome: requiescat in pace. Clin Chem 2005; 51: 931-8.

8 Caumo A, Bergman RN, Cobelli C. Insulin sensitivity from meal tolerance tests in normal subjects: a minimal model index. J Clin Endocrinol Metab 2000; 85: 4396-402.

9 Aloulou I, Brun JF, Mercier J. Evaluation of insulin sensitivity and glucose effectiveness during a standardized breakfast test: comparison with the minimal model analysis of an intravenous glucose tolerance test. Metabolism 2006; 55: 676-90.

10 Deeks JJ, Altman DG. Diagnostic tests 4: likelihood ratios. Brit Med J 2004; 329: 168-9.

11 Adams RJ, Appleton S, Wilson DH, Taylor AW, Dal Grande E, Chittleborough C, et al. Population comparison of two clinical approaches to the metabolic syndrome: implications of the new International Diabetes Federation consensus definition. Diabetes Care 2005; 28: 2777-9.

12 Guerrero-Romero F, Rodriguez-Moran M. Concordance between the 2005 International Diabetes Federation definition for diagnosing metabolic syndrome with the National Cholesterol Education Program Adult Treatment Panel III and the World Health Organization definitions. Diabetes Care 2005; 28: 2588-9.

13 Assmann G, Guerra R, Fox G, Cullen P, Schulte H, Willett D, et al. Harmonizing the definition of the metabolic syndrome: comparison of the criteria of the Adult Treatment Panel III and the International Diabetes Federation in United States American and European populations. Am J Cardiol 2007; 15: 541-8.

14 Dalla Man C, Yarasheski KE, Caumo A, Robertson H, Toffolo G, Polonsky KS, et al. Insulin sensitivity by oral glucose minimal models: validation against clamp. Am J Physiol Endocrinol Metab 2005; 289: 954-9.

15 Dalla Man C, Caumo A, Basu R, Rizza R, Toffolo G, Cobelli C. Measurement of selective effect of insulin on glucose disposal from labeled glucose oral test minimal model. Am J Physiol Endocrinol Metab 2005; 289: 909-14.

16 Cobelli C, Toffolo G, Dalla Man C, Campioni M, Denti P, Caumo A, et al. Assessment of beta-cell function in humans, simultaneously with insulin sensitivity and hepatic extraction, from intravenous and oral glucose tests. Am J Physiol Endocrinol Metab 2007; 293: 1-15.

17 Man D, Caumo A, Basu R, Rizza R, Toffolo G, Cobelli C. Minimal model estimation of glucose absorption and insulin sensitivity from oral test: validation with a tracer method. Am J Physiol Endocrinol Metab 2004; 287: 637-43.

18 Brun JF, Raynaud E, Mercier J. Homeostasis model assessment and related simplified evaluations of insulin sensitivity from fasting insulin and glucose: no need for log transformation but beware of the limits of validity. Diabetes Care 2000; 23: 1037-8.

19 Egan BM, Papademetriou V, Wofford M, Calhoun D, Fernandes J, Riehle JE, et al. TROPHY Sub-study Investigative Team. Metabolic syndrome and insulin resistance in the TROPHY sub-study: contrasting views in patients with high-normal blood pressure. Am J Hypertens 2005; 18: 3-12.

20 Onat A, Hergenc G, Turkmen S, Yazici M, Sari I, Can G. Discordance between insulin resistance and metabolic syndrome: features and associated cardiovascular risk in adults with normal glucose regulation. Metabolism 2006; 55: 445-52.

21 Sierra-Johnson J, Johnson BD, Allison TG, Bailey KR, Schwartz GL, Turner ST. Correspondence between the adult treatment panel III criteria for metabolic syndrome and insulin resistance. Diabetes Care 2006; 29: 668-72.

22 Cheal KL, Abbasi F, Lamendola C, McLaughlin T, Reaven GM, Ford ES. Relationship to insulin resistance of the adult treatment panel III diagnostic criteria for identification of the metabolic syndrome. Diabetes 2004; 53: 1195-200.

23 Meyer C, McGrath BP, Teede HJ. Effects of medical therapy on insulin resistance and the cardiovascular system in polycystic ovary syndrome. Diabetes Care 2007; 30: 471-8.

24 Akobeng AK. Understanding diagnostic tests 2: likelihood ratios, pre- and post-test probabilities and their use in clinical practice. Acta Paediatr 2007; 96: 487-91.

25 Kashyap SR, DeFronzo RA. The insulin resistance syndrome: physiological considerations. Diab Vasc Dis Res 2007; 4: 13-9.