Validity and Reproducibility of HOMA-IR, 1/HOMA-IR, QUICKI and Mcauley’S Indices in Patients with Hypertension and Type II Diabetes

Journal of Human Hypertension (2007) 21, 709–716 & 2007 Nature Publishing Group All rights reserved 0950-9240/07 $30.00 www.nature.com/jhh ORIGINAL ARTICLE Validity and reproducibility of HOMA-IR, 1/HOMA-IR, QUICKI and McAuley’s indices in patients with hypertension and type II diabetes

PA Sarafidis1,2, AN Lasaridis1, PM Nilsson3, MI Pikilidou1, PC Stafilas1, A Kanaki1, K Kazakos1, J Yovos1 and GL Bakris2 11st Department of Medicine, AHEPA University Hospital, Aristotle University, Thessaloniki, Greece; 2Hypertension/Clinical Research Center, Department of Preventive Medicine, Rush University Medical Center, Chicago, IL, USA and 3Department of Medicine, Malmo¨ University Hospital, Lund University, Malmo¨, Sweden

The aim of this study was to evaluate the validity and subjects’ body weight or fat-free mass and every other reliability of homeostasis model assessment-insulin clamp-derived index. 1/HOMA-IR and QUICKI indices resistance (HOMA-IR) index, its reciprocal (1/HOMA-IR), were positively correlated with the M-value (r ¼ 0.342, quantitative insulin sensitivity check index (QUICKI) and Po0.05 and r ¼ 0.456, Po0.01, respectively) and the rest McAuley’s index in hypertensive diabetic patients. In 78 clamp indices. McAuley’s index generally presented less patients with hypertension and type II diabetes glucose, strong correlations (r ¼ 0.317, Po0.05 with M-value). In insulin and triglyceride levels were determined after a multivariate analysis, HOMA-IR was the best fit of clamp- 12-h fast to calculate these indices, and insulin sensi- derived IS. Coefficients of variation between the two tivity (IS) was measured with the hyperinsulinemic visits were 23.5% for HOMA-IR, 19.2% for 1/HOMA-IR, euglycemic clamp technique. Two weeks later, subjects 7.8% for QUICKI and 15.1% for McAuley’s index. In had again their glucose, insulin and triglycerides conclusion, HOMA-IR, 1/HOMA-IR and QUICKI are valid measured. Simple and multiple linear regression analy- estimates of clamp-derived IS in patients with hyperten- sis were applied to assess the validity of these indices sion and type II diabetes, whereas the validity of compared to clamp IS and coefficients of variation McAuley’s index needs further evaluation. QUICKI between the two visits were estimated to assess their displayed better reproducibility than the other indices. reproducibility. HOMA-IR index was strongly and in- Journal of Human Hypertension (2007) 21, 709–716; versely correlated with the basic IS clamp index, the M- doi:10.1038/sj.jhh.1002201; published online 19 April 2007 value (r ¼À0.572, Po0.001), M-value normalized with

Keywords: insulin resistance; HOMA-IR; QUICKI; McAuley’s index

Introduction disturbances.2 In addition, both IR and compensatory hyperinsulinemia have been independently The term ‘metabolic’ or ‘insulin-resistance’ syn- associated in longitudinal studies with increased drome refers to a clustering in the same individual risk for CVD.3,4 of disorders that represent risk factors for cardio- In certain individuals, IR can precede the devel- vascular disease (CVD), such as impaired glucose opment of type II diabetes for many years, even tolerance (IGT) or type II diabetes, hypertension, 5,6 1 decades. During this long course compensatory dyslipidaemia and obesity. Insulin resistance (IR) hyperinsulinemia can have harmful consequences was originally proposed to be the central disorder of on many tissues not presenting IR and, through the syndrome, causally related with the other various mechanisms, contribute to the development of other components of the syndrome, that is 7,8 Correspondence: Dr PA Sarafidis, 1st Department of Medicine, hypertension or hypertriglyceridemia. Thus, a AHEPA University Hospital, Aristotle University, St Kiriakidi 1, proportion of subjects with the metabolic syndrome 546 36, Thessaloniki, Greece. can exhibit hypertension long before the develop- E-mail: [email protected] ment of overt type II diabetes. Sources of support: This paper was not supported by any source and represents an original effort on our part. Several methods have been proposed for assess- Received 4 August 2006; revised 16 December 2006; accepted 19 ment of IR, or its reciprocal variable, insulin December 2006; published online 19 April 2007 sensitivity (IS), in humans. Among them the Surrogates of insulin sensitivity in hypertension PA Sarafidis et al 710 hyperinsulinemic euglycemic clamp technique, de- days for the evaluation and five of them did not scribed by DeFronzo et al.9 in 1979, represents undergo the clamp due to difficulties in achieving currently the ‘gold standard’.10 However, due to its stable intravenous access. Therefore, the final study many technical requirements, the clamp technique, group included 78 Caucasian patients (36 men and as well as several other methods for IS measure- 42 women) with complete data set, which were ment, is difficult to apply in everyday clinical included in the analysis. practice or large-scale epidemiologic studies. This is why during the past 20 years numerous simple indices have been developed for IS measurement. Study protocol Some of these indices derive from an oral-glucose Participants were admitted to the Clinical Research tolerance test, whereas others are based only on Laboratory of our Department at 0700 after a 12-h fasting glucose and insulin values,1 and this simpli- overnight fast and without having received their city in obtaining the latter surrogates was the morning hypoglycemic or antihypertensive medica- obvious reason for their wider use. tion. Blood samples were drawn to determine the Among these indices, the most well-known is levels of fasting plasma glucose, insulin and

perhaps the homeostasis model assessment-insulin triglycerides and HbA1c. Fasting glucose and insulin resistance (HOMA-IR) index,11 whereas the recipro- values were used to estimate HOMA-IR, 1/HOMA-IR cal of HOMA-IR, as well as a more recently proposed and QUICKI indices and insulin and triglycerides to index, the quantitative insulin sensitivity check estimate the McAuley’s index, as described below. index (QUICKI),12 are increasingly used. The valid- Subjects had also measured their body weight and ity of these estimates in relation to the clamp height to calculate body mass index (BMI). The body technique has been examined in several studies, in composition of each participant was then analysed a wide variety of populations.13–20 However, data on by bioelectrical impedance analysis with the use of their validity in hypertensive populations are gen- the Bodystat1500 device (Bodystat Ltd, Douglas, Isle erally limited. Recently, McAuley et al.21 proposed of Man, British Isles) to estimate the fat-free mass. another index, which uses fasting insulin and Bioelectrical impedance analysis has been shown to triglyceride values, to be a strong predictor of be a simple, valid and reliable method to evaluate clamp-derived IS, but this index was not adequately body composition.23,24 After these procedures, sub- further investigated. In addition, data on the jects had their IS determined with the clamp reproducibility of these surrogate indexes are quite technique. limited11,13,19,22 compared to the data on validity. Two weeks after the first visit, participants Therefore, the aim of the present study was to returned at the Department, again after a 12-h fast determine the comparative validity and the repro- and without having received morning medications ducibility of HOMA-IR, 1/HOMA-IR, QUICKI and to give blood samples for the determination of McAuley’s index in a population consisting of fasting plasma glucose, insulin and triglyceride patients that have initially developed hypertension levels (visit 2). From these values, all surrogate and subsequently type II diabetes. indices were again calculated to evaluate their reproducibility between different time points. All subjects were strictly advised to keep their diet Methods habits and physical activity unchanged during these 2 weeks to avoid major changes in background IR. Subjects Among patients attending the hypertension or diabetes outpatient clinic of the 1st Department of Assessments Medicine, AHEPA Hospital, Thessaloniki, Greece, Hyperinsulinemic euglycemic clamp. The hyper- we selected those with both hypertension and type insulinemic euglycemic clamp was performed as II diabetes that had diagnosis of hypertension before described previously.9,10 In brief, two intravenous the diagnosis of diabetes. We excluded patients infusion lines were placed, one into an antecubital receiving insulin treatment, having glycated hemo- vein for the infusion of insulin and glucose and the globin (HbA1c) 410%, history of secondary hyper- other into a hand or wrist vein by retrograde tension, serum creatinine levels 42 mg/dl, history cannulation for frequent blood sampling. After a of myocardial infarction or unstable angina within 10-min priming infusion, insulin infusion was held the past 6 months, heart failure NYHA class III–IV, constant at 0.6 nmol*mÀ2*minÀ1 for the rest 110 min. malignancy or any other condition with poor Blood glucose concentration was determined every prognosis. From this population we randomly 5 min and it was clamped at the euglycemic level selected 100 individuals, using a computer-gener- (5 mmol/l) by infusion of variable amounts of a 20% ated list. Of these, six were not possible to be dextrose solution. The total body glucose disposal contacted and eight refused to participate. The rate (M-value), the basic clamp-derived IS index, remaining 86 subjects volunteered for the study, was the average value of the glucose infusion providing informed consent after information. Three rate during the final 40 min of the 120-min study of those subjects did not show up at the scheduled (steady-state). As different studies have used various

Journal of Human Hypertension Surrogates of insulin sensitivity in hypertension PA Sarafidis et al 711 sub-indices derived from the clamp, the M-value visit 1 and visit 2 and x¯ the pooled mean value for was normalized with body weight (Mbw) and fat-free the two visits. We also performed comparisons 10 mass (Mffm) for reasons of comparison. The mean of between the mean values of visit 1 and visit 2 with the three plasma insulin measurements obtained at the use of Student’s t-test for paired variables or 80, 100 and 120 min of the study represented steady- Wilcoxon’s Signed ranks test, depending on normal- state insulin and was used to standardize M-values, ity of distribution, and calculated correlation coeffi- thus forming the respective M/I indices. Metabolic cients between those two occasions. P-value levels clearance rate (MCR) of plasma glucose, the last o0.05 (two-tailed) were considered statistically clamp-derived index, was obtained by dividing the significant. respective M-values by the mean steady-state plasma glucose concentrations, and was again further normalized with body weight and fat-free mass.9,10 Results Estimation of surrogate indices. HOMA-IR index The demographic and clinical characteristics of the was calculated according to the formula: HOMA- total study population, men and women, as well as IR ¼ fasting glucose in mmol/l*fasting insulin in baseline values of the surrogate indices examined mU/ml/22.5.11 The reciprocal of HOMA-IR was and the various indices derived from the hyperin- calculated from the type: 1/HOMA-IR ¼ 1/[HOMA- sulinemic euglycemic clamp are presented in IR]. QUICKI was estimated according to the formula Table 1. by Katz et al.12: QUICKI ¼ 1/[log(fasting insulin in Correlation analysis demonstrated that HOMA-IR mU/ml) þ log(fasting glucose in mg/dl)]. Accord- was strongly and inversely correlated with the M ingly, McAuley’s index was calculated from the (r ¼À0.572, Po0.001) (Figure 1a), M/I (r ¼À0.768, formula: McAuley’s index ¼ exp[2.63À0.28 ln (fast- Po0.001) and MCR (r ¼À0.576, Po0.001) values of ing insulin in mU/ml)À0.31 ln (fasting triglycerides the clamp, as well as all the above values normal- in mmol/l)].21 ized with body weight or fat-free mass (Table 2). 1/HOMA-IR index, was significantly but positively Biochemical analyses. Plasma glucose and trigly- correlated with M-value (r ¼ 0.342, Po0.05) and the cerides were measured with standard laboratory rest clamp-derived indexes (Figure 1b and Table 2). methods in Roche/Hitachi 912 automatic analyser The QUICKI, which also reflects IS, was again (Roche Diagnostics, Basel, Switzerland). Plasma positively correlated with M-value (r ¼ 0.456, insulin was determined by radioimmunoassay (Dia- Po0.01), and with all the rest clamp indexes Sorin, Saluggia, Italy). HbA1c was measured with (Figure 1c and Table 2). Both 1/HOMA-IR and high-performance liquid chromatography (Menarini QUICKI exhibited somehow lower correlation coef- Diagnostics, Florence, Italy). Glucose measurements ficients with M, but higher with M/I values than the during the clamp were performed with HemoCue HOMA-IR index. McAuley’s index also demon- B-glucose analyser (Hemocue AB, A¨ ngelholm, strated significant correlations with M (r ¼ 0.317, Sweden). This analyser was found accurate in Po0.05), M/I (r ¼ 0.546, Po0.001) and MCR indexes comparison to standard laboratory methods.25 In (r ¼ 0.546, Po0.05). However, it was not signifi- previous works of our group, HemoCue rendered a cantly correlated with Mffm and MCRffm indices and correlation coefficient of r ¼ 0.98 and a within-run its correlation coefficients were lower than all the variation o1.5%.26 surrogate indices examined (Figure 1d and Table 2). To explore mutual confounding of the relation- ships between each surrogated index and clamp- Statistical analysis derived IS from age, sex and BMI, as well as to Statistical analysis was performed using the Statis- estimate the best fit for clamp-IS, we performed tical Package for Social Sciences 13 software (SPSS multiple regression analysis. In all the models PC Inc., Chicago, IL, USA). All data are expressed as developed, the M-value was included as a depen- mean7s.d. The validity of each surrogate index was dent variable and age, sex and BMI as common- assessed with calculation of Pearson’s r correlation independent variables. HOMA-IR (model 1), coefficients between each surrogate index and the 1/HOMA-IR (model 2), QUICKI (model 3) or McAu- indexes of the clamp, for reasons of comparison. ley’s index (model 4) were each entered as an Multiple linear regression analysis was performed to additional independent variable in respective mod- determine whether age, sex and BMI influenced the els. Age did not affect any of the observed associa- association between each surrogate index and tions and further analyses were performed after clamp-derived M-value, as well as to explore the excluding it. Both the associations between HOMA- estimated best fit for clamp-derived IS. The repro- IR or QUICKI and the M-value were strengthened ducibility of each index was assessed with the use of after adjustment for sex and BMI, as indicated by the the coefficient of variance (CV) between its values at coefficients of determination (Table 3). The associa- visit 1 and visit 2. The CV waspffiffiffi calculated according tion between 1/HOMA-IR and the M-value was to the formula: CV ¼ðs:d:= 2ÞÃ100=x¯ where s.d. is only affected by sex and adjustment for this factor the standard deviation of the difference between also strengthened the association, whereas the

Journal of Human Hypertension Surrogates of insulin sensitivity in hypertension PA Sarafidis et al 712 Table 1 Baseline characteristics of the subjects (m7s.d.)

All subjects Men Women

N 78 36 42 Age (years) 61.678.1 62.578.4 60.977.5 Duration of DM (years) 8.378.4 9.078.5 7.776.7 Duration of hypertension (years) 14.1710.8 15.779.3 12.8710.6 Weight (kg) 75.579.0 78.278.8 73.378.9 BMI (kg/m2) 27.772.9 25.971.6 29.173.0 Waist to hip ratio 0.9370.07 0.9570.06 0.9070.07 Glucose (mmol/l) 8.4771.86 8.6771.67 8.3172.02 Insulin (mU/ml) 14.574.9 13.776.2 15.073.5 7 7 7 HbA1c (%) 7.70 1.08 7.92 0.99 7.52 1.13 Triglycerides (mmol/l) 1.5670.61 1.5570.60 1.5770.63 Systolic blood pressure (mm Hg) 141.979.6 144.6711.2 139.778.1 Diastolic blood pressure (mm Hg) 81.776.8 83.378.2 79.676.1 HOMA-IR 5.3771.87 5.2372.21 5.4971.59 1/HOMA-IR 0.21470.098 0.23270.130 0.19870.061 QUICKI 0.30370.015 0.30670.019 0.30170.012 McAuley’s index 1.78870.199 1.81370.230 1.76770.172 M (mmol*minÀ1) 2.7970.83 3.0170.97 2.6170.66 m À1 À1 7 7 7 Mbw ( mol*min *kg ) 37.0 10.4 38.6 12.1 35.7 8.8 m À1 À1 7 7 7 Mffm ( mol*min *kg ) 57.1 16.0 53.6 18.1 59.9 13.7 M/I (mmol*minÀ1*nmolÀ1*l) 2.8370.90 3.0771.14 2.6370.60 m À1 À1 À1 7 7 7 Mbw/I ( mol*min *kg *nmol *l) 37.9 13.4 40.4 18.1 35.9 7.7 m À1 À1 À1 7 7 7 Mffm/I ( mol*min *kg *nmol *l) 59.1 24.8 57.6 35.1 60.3 11.7 MCR (l minÀ1) 0.5670.17 0.6070.19 0.5270.13 * À1 À1 7 7 7 MCRbw (ml min kg ) 7.42 2.09 7.75 2.43 7.16 1.77 * À1* À1 7 7 7 MCRffm (ml*min *kg ) 11.4 3.2 10.7 3.6 12.0 2.8

Abbreviations: BMI, body mass index; HbA1c, glycated hemoglobin; HOMA-IR, homeostasis model assessment-insulin resistance index; M, Mbw, and Mffm total body glucose metabolism in absolute units and normalized with body weight and fat-free mass, respectively; M/I, Mbw/I and Mffm/I, the same indices divided by the steady-state insulin concentration; MCR, MCRbw and MCRffm, metabolic clearance rates, obtained by dividing the respective M-values by the mean steady-state plasma glucose concentrations; QUICKI, quantitative insulin sensitivity check index. a r = -0.572, P< 0.001 b r = 0.342, P< 0.05

4.0 4.0

3.0 3.0

2.0 2.0 M (mmol/min) M (mmol/min)

1.0 1.0

0.0 2.0 4.0 6.0 8.0 10.0 12.0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 HOMA-IR 1/HOMA-IR c r = 0.456, P<0.01 d r = 0.317, P < 0.05

4.0 4.0

3.0 3.0

2.0 2.0 M (mmol/min) M (mmol/min)

1.0 1.0

0.1 0.2 0.3 0.4 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 QUICKI McAuley's index Figure 1 Correlations between HOMA-IR (a), 1/HOMA-IR (b), QUICKI (c) or McAuley’s index (d) and the M-value of the clamp (r ¼À0.572, Po0.001, r ¼ 0.342, Po0.05, r ¼ 0.456, Po0.01, and r ¼ 0.317, Po0.05, respectively).

Journal of Human Hypertension Surrogates of insulin sensitivity in hypertension PA Sarafidis et al 713 association between McAuley’s index and M-value 0.001 0.001 0.001 was affected by neither sex nor BMI. As shown by ffm o o o the adjusted R2 values (Table 3), the HOMA-IR

MCR model explained 51.4% of the variability of the M-

0.676 value, the 1/HOMA-IR model 24.7% and the À QUICKI model 36.6% of it. McAuley’s index explained only the 18.1% of the M-value variability 0.05 0.259 0.102 0.001 0.612 0.05 0.559 0.001 and this was not significant. bw o o o o In regards to reproducibility, comparison of the mean values of each index between visit 1 and visit

0.348 2 yielded non-significant results (data not shown). À The CVs for the values of the surrogate indexes between visit 1 and visit 2 were 23.5% for HOMA- 0.05 0.390 0.01 0.590 0.05 0.228 0.001 IR, 19.2% for 1/HOMA-IR, 7.8% for QUICKI and o o o o 15.1% for McAuley’s index. The correlation coeffi-

MCR MCR cients between the values of the two measurement , metabolic clearance rates, obtained by dividing the

0.576 occasions were r ¼ 0.367, Po0.05; r ¼ 0.409, ffm À Po0.05; r ¼ 0.782, Po0.01; and r ¼ 0.545, Po0.05, respectively. 0.01 0.324 0.001 0.459 0.001 0.343 0.001 I / and MCR o o o o ffm bw M

0.678 Discussion À This study was designed to determine the validity and the reproducibility of HOMA-IR, 1/HOMA-IR, 0.001 0.453 0.001 0.781 0.001 0.841 0.001 /I o o o o QUICKI and McAuley’s index in a population of bw

M patients that have developed hypertension before type II diabetes. The reasons for choosing such 0.784 total body glucose metabolism in absolute units and normalized with body weight and fat-free

À patients were that this type of subjects is (a) quite ffm

M common in clinical practice and (b) closer to the natural course of the metabolic syndrome, thus it and 0.001 0.573 0.001 0.823 0.001 0.822 0.001 o o o o bw could be different in terms of degree of IR from the M M/I

, total hypertensive or the total type II diabetic

M population, which contains subjects with various

0.768 27

À levels of IR. Indeed, previous data suggest that individuals with both hypertension and type II diabetes have higher IR than normotensive type II 0.001 0.793 0.001 0.778 0.001 diabetic patients.28,29 Our main finding was that o o o ffm HOMA-IR, 1/HOMA-IR and QUICKI indices were M strongly and significantly correlated with all the

0.674 indices derived from clamp technique in this À population. These results further prove the validity of these measures to assess IS. McAuley’s index 0.05 0.252 0.112 0.546 0.001 0.611 0.01 0.559 0.001 presented significant, but less strong, correlations o o o o bw with most of clamp indices. Multivariate analysis revealed that HOMA-IR was the best fit among these , the same indices divided by the steady-state insulin concentration; MCR, MCR 0.708 I

/ indices of clamp-derived IS. However, HOMA-IR, 1/ À

ffm HOMA-IR and McAuley’s indices presented con- M siderably high CVs, a finding questioning their 0.05 0.385 0.01 0.589 0.05 0.478 0.001 reproducibility. o o o o

/I and 11

MM Since the description of HOMA-IR, several bw

M studies reported it to be a valid estimate of clamp- , rPrPrPrPrPrPrPrPrP

0.572 derived IS in various types of subjects, that is healthy À M/I lean individuals,14,16,30,31 obese subjects,14–16,31 subjects with IGT,16 type II diabetes,13–16 polycystic 16 17 -values by the mean steady-state plasma glucose concentrations; QUICKI, quantitative insulin sensitivity check index. ovarian syndrome (PCOS), pregnant women, as M 18,32 Correlation coefficients between the surrogate indices of insulin resistance and the various clamp-derived indices well as in children and adolescents. However, not all studies confirm the validity of HOMA-IR in certain subject categories, like elderly subjects or 33,34 McAuley’s index 0.317 QUICKI 0.456 respective mass, respectively; 1/HOMA-IR 0.342 Table 2 Abbreviations: HOMA-IR, homeostasis model assessment-insulin resistance index; HOMA-IR women with PCOS.

Journal of Human Hypertension Surrogates of insulin sensitivity in hypertension PA Sarafidis et al 714 Table 3 Coefficients of determination in multiple regression analyses models including surrogate indices of insulin sensitivity, sex and BMI as independent variables and M-value as dependent variable

Model 1 Model 2 Model 3 Model 4

b P b P b P b P

Sex À0.454 o0.01 À0.381 o0.05 À0.398 o0.05 À0.355 NS BMI 0.442 o0.01 0.365 NS 0.422 o0.05 0.261 NS HOMA-IR À0.691 o0.001 — — — — — — 1/HOMA-IR — — 0.433 o0.05 — — — — QUICKI — — — — 0.575 o0.001 — — McAuley’s index — — — — — — 0.302 NS R2 0.514 Po0.001 0.247 Po0.05 0.366 Po0.01 0.181 P ¼ 0.127

Abbreviations: BMI, body mass index; HOMA-IR, homeostasis model assessment-insulin resistance index; QUICKI, quantitative insulin sensitivity check index.

Studies examining the validity of HOMA-IR in normotriglyceridemic. It is not known whether this patients with hypertension are much less common. discrepancy in our population could have affected Bonora et al.14 have previously evaluated HOMA-IR the performance of this particular index. In any case, in 115 subjects with various degrees of glucose further studies seem necessary to assess its true tolerance. In the subgroup of hypertensive patients, validity. HOMA-IR displayed a strong correlation We have also developed multiple regression (r ¼À0.762) with IS represented with the Mffm index. analysis models including M-value as dependent In the only other study so far assessing the accuracy variable to determine which of the indices is the of HOMA-IR in patients with hypertension, Lansang strongest predictor of clamp-derived IS. The first et al.35 have found HOMA-IR to display again a model including sex, BMI and HOMA-IR as inde- strong correlation (r ¼À0.64) with M-value in 27 pendent variables predicted 51.4% of the variability hypertensives. Our study further supports these in difference in M-value with a formula to estimate findings, as it showed comparable correlation the M-value ¼ 0.568À0.746*sex þ 0.125*BMIÀ0.305* coefficients in a much larger sample of patients. HOMA-IR, where sex ¼ 0 for men and 1 for women. QUICKI was also found to be a valid surrogate of 1/HOMA-IR and QUICKI were less strong predictors clamp-derived IS in a similarly broad category of of clamp-derived IS, something different from subject types,12,16–20,30–32 although there were again previous findings in type II diabetic patients, where some studies with diverse findings.33,34 In the only both these indices explained about 55% of IS study on accuracy of QUICKI in hypertensives, it variability.19 McAuley’s index was not an indepen- was shown to be also highly correlated with IS dent predictor of M-value variability in this represented with Mbw/I index (r ¼ 0.84) in 27 population. patients with essential hypertension without dia- In regards to reproducibility, although Matthews betes.36 The validity of 1/HOMA-IR was also et al.11 observed a low precision of the HOMA-IR evaluated in normal and type II diabetic sub- model (CV of 31%) and suggested that this could jects,19,20 but only the previous study examined it limit its use, much less studies examined this in hypertensive patients, showing similar good aspect13,38 compared with the validity of the method. correlations (r ¼ 0.82).36 Again, our findings are in In the original description of QUICKI, Katz et al.12 total agreement with these results, as we observed supported that their index exerted high reproduci- almost identical correlations between these indices bility. However, they assessed reproducibility by and Mbw/I index. comparing the association of each of two separated McAuley et al.21 originally reported that their QUICKI estimations with clamp-derived IS, instead index correlated better with clamp-derived IS than a of determining the CV and the association of the two series of surrogate indexes, including HOMA-IR, measurements. Subsequent studies showed CVs of insulin-to-glucose ratio and Bennett index in the 11.7% for HOMA-IR,13 11.4% for 1/HOMA-IR and general population. A recent study also showed that 2% for QUICKI19 in type II diabetic patients, whereas, McAuley’s index had better correlations with to our knowledge, no study so far has examined the clamp-derived IS than both HOMA-IR and QUICKI reproducibility of the McAuley’s index. In the in 51 stable renal transplant recipients.37 Our present study, the reproducibility of these surrogate findings do not confirm those results, since McAu- indexes is lower than in the above studies, and for ley’s index was the worse predictor of IS among the HOMA-IR is closer to the estimation of Matthews surrogate measures examined. It has to be noted, et al.11 In a mixed population of healthy lean, obese however, that in our population the mean triglycer- and type II diabetic individuals, Mather et al.22 ide level is in the normal range, in contrast to reported a CV of 51% for HOMA and 5% for QUICKI. glucose levels, whereas in the study from McAuley A more detailed study also suggested an impor- et al.21 subjects are both normoglycemic and tant intraindividual variation of HOMA-IR, which

Journal of Human Hypertension Surrogates of insulin sensitivity in hypertension PA Sarafidis et al 715 was significantly higher in patients with type II aimed to examine this question in patients devel- diabetes than nondiabetic individuals.38 This could oping hypertension before type II diabetes, as reflect the intraindividual variation of glucose and discussed above. insulin values, which is also greater in patients with In conclusion, this study suggests that HOMA-IR, type II diabetes compared to normoglycemic sub- 1/HOMA-IR and QUICKI indices are valid estimates jects.39 The critical question is whether this biologic of IS in patients with both hypertension and type II variation of glucose and insulin transferred to diabetes. The McAuley’s index is less strongly surrogate indices limit their ability to give reliable correlated with clamp-derived IS and should rather estimates of IS. Studies on the reproducibility of not be used instead of the above indices. HOMA-IR clamp technique are also few, and have shown CVs seemed the best estimate of clamp-derived IS, but ranging from 5.840 to 15%.41 If the CV of the clamp is QUICKI displayed a much better reproducibility. On also high, this could reflect the true intraindividual this basis, a judgment should be made on the variation of IS, thus the surrogate indexes would just preferable use of these indices in this type of mirror a physiological phenomenon. However, if the patients, until future studies elucidate the precision reproducibility of the clamp method is around 5%, of HOMA-IR and 1/HOMA-IR. the surrogate indexes would need to be improved. In any case, the reproducibility of the clamp and the surrogate indices is a field that needs further References investigation for stable conclusions to be made. This study has also some limitations. First, the 1 Sarafidis PA, Nilsson PM. The metabolic syndrome: a subjects included were recruited from a population glance at its history. J Hypertens 2006; 24: 621–626. attending Outpatient Clinics of a University Hospi- 2 Reaven GM. Banting lecture 1988. Role of insulin tal. To what extent these patients differ from similar resistance in human disease. Diabetes 1988; 37: patients referring to community hospitals, health 1595–1607. centres or private physician offices cannot be 3 Hanley AJ, Williams K, Stern MP, Haffner SM. Home- ostasis model assessment of insulin resistance in established. Second, history of hypertension and relation to the incidence of cardiovascular disease: diabetes was obtained from existing databases of the the San Antonio Heart Study. Diabetes Care 2002; 25: above Clinics. Although such data are collected 1177–1184. systematically, history of diseases is self-reported 4 Pyorala M, Miettinen H, Laakso M, Pyorala K. Plasma and, thus, recall bias could potentially exist. insulin and all-cause, cardiovascular, and noncardio- Further, it could be argued that it is not known to vascular mortality: the 22-year follow-up results of the what extent these data apply to the general hyper- Helsinki Policemen Study. Diabetes Care 2000; 23: tensive population, since our subjects had also type 1097–1102. II diabetes and vice versa. However, we particularly 5 Beck-Nielsen H, Groop LC. Metabolic and genetic characterization of prediabetic states. Sequence of events leading to non-insulin-dependent diabetes mellitus. J Clin Invest 1994; 94: 1714–1721. 6 Turner NC, Clapham JC. Insulin resistance, impaired What is known about this topic: glucose tolerance and non-insulin-dependent diabetes, K Within the context of the metabolic syndrome, insulin pathologic mechanisms and treatment: current status resistance and hyperinsulinemia can contribute towards the and therapeutic possibilities. Prog Drug Res 1998; 51: development of hypertension long before the onset of type II 7,8 33–94. diabetes. Patients with both hypertension and type II 7 McFarlane SI, Banerji M, Sowers JR. Insulin resistance diabetes are more insulin-resistant that those with only type II diabetes.28,29 and cardiovascular disease. J Clin Endocrinol Metab K Several studies in a variety of populations have shown that 2001; 86: 713–718. indices like HOMA-IR or QUICKI are valid surrogate markers 8 Sarafidis PA, Lasaridis AN. Actions of PPARg agonists for insulin sensitivity.13–20 explaining a possible blood pressure lowering effect. K Data on the reliability of the above indices,13,19,22 as well as Am J Hypertens 2006; 19: 646–653. their validity in hypertensive populations11,13,19,22 are very 9 DeFronzo RA, Tobin JD, Andres R. Glucose limited. clamp technique: a method for quantifying insulin secretion and resistance. Am J Physiol 1979; 237: What this study adds: 214–223. K In patients who developed hypertension before type II diabetes HOMA-IR, 1/HOMA-IR and QUICKI were strongly 10 Ferrannini E, Mari A. How to measure insulin correlated with insulin sensitivity determined with the sensitivity. J Hypertens 1998; 16: 895–906. hyperinsulinemic euglycemic clamp. Thus, these indexes are 11 Matthews DR, Hosker JP, Rudenski AS, Naylor BA, valid estimates of insulin sensitivity in this population. Treacher DF, Turner RC. Homeostasis model assess- K In contrast to QUICKI, HOMA-IR and 1/HOMA-IR presented ment: insulin resistance and beta-cell function from considerably high coefficients of variation, a finding fasting plasma glucose and insulin concentrations in questioning their reproducibility. man. Diabetologia 1985; 28: 412–419. K McAuley’s index showed less strong correlations with clamp- 12 Katz A, Nambi SS, Mather K, Baron AD, Follmann DA, derived insulin sensitivity and intermediate reproducibility Sullivan G et al. Quantitative insulin sensitivity check compared with the rest indices. Based on these data, this index is not superior from older surrogate measures of insulin index: a simple, accurate method for assessing insulin sensitivity and its use should be further examined. sensitivity in humans. J Clin Endocrinol Metab 2000; 85: 2402–2410.

Journal of Human Hypertension Surrogates of insulin sensitivity in hypertension PA Sarafidis et al 716 13 Emoto M, Nishizawa Y, Maekawa K, Hiura Y, Kanda H, 27 World Health Organization. Definition Diagnosis and Kawagishi T et al. Homeostasis model assessment as a Classification of Diabetes Mellitus and its Complica- clinical index of insulin resistance in type 2 diabetic tions. Report of a WHO Consultation. Part 1: diagnosis patients treated with sulfonylureas. Diabetes Care and classification of diabetes mellitus. World Health 1999; 22: 818–822. Organization: Geneva, 1999. 14 Bonora E, Targher G, Alberiche M, Bonadonna RC, 28 Laakso M, Sarlund H, Mykkanen L. Essential hyper- Saggiani F, Zenere MB et al. Homeostasis model tension and insulin resistance in non-insulin-depen- assessment closely mirrors the glucose clamp techni- dent diabetes. Eur J Clin Invest 1989; 19: 518–526. que in the assessment of insulin sensitivity: studies in 29 Kashiwabara H, Inaba M, Maruno Y, Morita T, Awata T, subjects with various degrees of glucose tolerance and Negishi K et al. Insulin levels during fasting and the insulin sensitivity. Diabetes Care 2000; 23: 57–63. glucose tolerance test and Homa’s index predict 15 Ikeda Y, Suehiro T, Nakamura T, Kumon Y, Hashimoto subsequent development of hypertension. J Hypertens K. Clinical significance of the insulin resistance index 2000; 18: 83–88. as assessed by homeostasis model assessment. Endocr 30 Cheng C, Campbell KL, Kushner H, Falkner BE. J 2001; 48: 81–86. Correlation of oral glucose tolerance test-derived 16 Bastard JP, Rabasa-Lhoret R, Maachi M, Ducluzeau PH, estimates of insulin sensitivity with insulin clamp Andreelli F, Vidal H et al. What kind of simple fasting measurements in an African-American cohort. Meta- index should be used to estimate insulin sensitivity in bolism 2004; 53: 1107–1112. humans? Diabetes Metab 2003; 29: 285–288. 31 Straczkowski M, Stepien A, Kowalska I, Kinalska I. 17 Kirwan JP, Huston-Presley L, Kalhan SC, Catalano PM. Comparison of simple indices of insulin sensitivity Clinically useful estimates of insulin sensitivity during using the euglycemic hyperinsulinemic clamp techni- pregnancy: validation studies in women with normal que. Med Sci Monit 2004; 10: 480–484. glucose tolerance and gestational diabetes mellitus. 32 Conwell LS, Trost SG, Brown WJ, Batch JA. Indexes of Diabetes Care 2001; 24: 1602–1607. insulin resistance and secretion in obese children and 18 Gungor N, Saad R, Janosky J, Arslanian S. Validation of adolescents: a validation study. Diabetes Care 2004; surrogate estimates of insulin sensitivity and insulin 27: 314–319. secretion in children and adolescents. J Pediatr 2004; 33 Katsuki A, Sumida Y, Urakawa H, Gabazza EC, 144: 47–55. Murashima S, Morioka K et al. Neither homeostasis 19 Yokoyama H, Emoto M, Fujiwara S, Motoyama K, model assessment nor quantitative insulin sensitivity Morioka T, Komatsu M et al. Quantitative insulin check index can predict insulin resistance in elderly sensitivity check index and the reciprocal index of patients with poorly controlled type 2 diabetes homeostasis model assessment in normal range weight mellitus. J Clin Endocrinol Metab 2002; 87: 5332–5335. and moderately obese type 2 diabetic patients. Dia- 34 Diamanti-Kandarakis E, Kouli C, Alexandraki K, Spina betes Care 2003; 26: 2426–2432. G. Failure of mathematical indices to accurately assess 20 Yokoyama H, Emoto M, Fujiwara S, Motoyama K, insulin resistance in lean, overweight, or obese women Morioka T, Komatsu M et al. Quantitative insulin with polycystic ovary syndrome. J Clin Endocrinol sensitivity check index and the reciprocal index of Metab 2004; 89: 1273–1276. homeostasis model assessment are useful indexes of 35 Lansang MC, Williams GH, Carroll JS. Correlation insulin resistance in type 2 diabetic patients with wide between the glucose clamp technique and the home- range of fasting plasma glucose. J Clin Endocrinol ostasis model assessment in hypertension. Am J Metab 2004; 89: 1481–1484. Hypertens 2001; 14: 51–53. 21 McAuley KA, Williams SM, Mann JI, Walker RJ, Lewis- 36 Chen H, Sullivan G, Yue LQ, Katz A, Quon MJ. QUICKI Barned NJ, Temple LA et al. Diagnosing insulin is a useful index of insulin sensitivity in subjects with resistance in the general population. Diabetes Care hypertension. Am J Physiol Endocrinol Metab 2003; 2001; 24: 460–464. 284: 804–812. 22 Mather KJ, Hunt AE, Steinberg HO, Paradisi G, Hook G, 37 Oterdoom LH, de Vries AP, van Son WJ, van der Heide Katz A et al. Repeatability characteristics of simple indices JJ, Ploeg RJ, Gansevoort RT et al. Validation of insulin of insulin resistance: implications for research applica- resistance indexes in a stable renal transplant populations. J Clin Endocrinol Metab 2001; 86: 5457–5464. tion. Diabetes Care 2005; 28: 2424–2429. 23 Houtkooper LB, Lohman TG, Going SB, Howell WH. 38 Jayagopal V, Kilpatrick ES, Jennings PE, Hepburn DA, Why bioelectrical impedance analysis should be used Atkin SL. Biological variation of homeostasis model for estimating adiposity. Am J Clin Nutr 1996; assessment-derived insulin resistance in type 2 dia- 64(Suppl): 436S–448S. betes. Diabetes Care 2002; 25: 2022–2025. 24 NIH Consensus statement. Bioelectrical impedance 39 Mooy JM, Grootenhuis PA, de VH, Kostense PJ, Popp- analysis in body composition measurement. National Snijders C, Bouter LM et al. Intra-individual variation Institutes of Health Technology Assessment Confer- of glucose, specific insulin and proinsulin concentra- ence Statement. December 12–14, 1994. Nutrition tions measured by two oral glucose tolerance tests in a 1996; 12: 749–762. general Caucasian population: the Hoorn Study. 25 Voss EM, Cembrowski GS. Performance characteristics Diabetologia 1996; 39: 298–305. of the HemoCue B-glucose analyzer using whole-blood 40 Soop M, Nygren J, Brismar K, Thorell A, Ljungqvist O. The samples. Arch Pathol Lab Med 1993; 117: 711–713. hyperinsulinaemic–euglycaemic glucose clamp: reprodu- 26 Sarafidis PA, Lasaridis AN, Nilsson PM, Pagkalos EM, cibility and metabolic effects of prolonged insulin infusion Hitoglou-Makedou AD, Pliakos CI et al. Ambulatory in healthy subjects. Clin Sci (London) 2000; 98: 367–374. blood pressure reduction after rosiglitazone treatment 41 Bokemark L, Froden A, Attvall S, Wikstrand J, in patients with type 2 diabetes and hypertension Fagerberg B. The euglycemic hyperinsulinemic clamp correlates with insulin sensitivity increase. J examination: variability and reproducibility. Scand J Hypertens 2004; 22: 1769–1777. Clin Lab Invest 2000; 60: 27–36.

Journal of Human Hypertension