Emerging Treatments and T e c h n o l o g i e s ORIGINAL AR T I C L E

Homeostasis Model Assessment Closely Mirrors the Glucose Clamp Technique in the Assessment of Insulin Se n s i t i v i t y Studies in subjects with various degrees of glucose tolerance and insulin se n s i t i v i t y

ENZO BONORA, MD, PHD FRANCESCA SAGGIANI, MD sulin sensitivity (7). However, this method GIOVANNI TARGHER, MD MARINA B. ZENERE, MD is laborious, expensive, and theref o r e un- MARIA ALBERICHE, MD TIZIANO MONAUNI, MD suitable for large-scale or epidemiological RICCARDO C. BONADONNA, MD MICHELE MUGGEO, MD studies. Several alternative methods to evaluate insulin sensitivity have been pro- posed during the last two decades (8–13), but, although generally less complex and less troublesome than the glucose clamp OB J E C T I V E — To evaluate whether the homeostasis model assessment (HOMA) is a rel i - technique, none of them is as simple as is able surrogate measure of in vivo insulin sensitivity in humans. ne c e s s a r y in large-scale studies involving RESEARCH DESIGN AND METHODS — In the present study, we compared insulin hu n d r eds or thousands of subjects. sensitivity as assessed by a 4-h euglycemic ( 5 mmol/l) hyperinsulinemic ( 300 pmol/l) clamp Homeostasis model assessment (HOMA) with HOMA in 115 subjects with various degrees of glucose tolerance and insulin sensitivity. of insulin sensitivity was proposed about 10 years ago as a simple and inexpensive alter- RE S U LT S — We found a strong correlation between clamp-measured total glucose disposal native to more sophisticated techniques and HOMA-estimated insulin sensitivity (r = 0.820, P 0.0001), with no substantial dif- (14). Such a method derives an estimate of fe r ences between men (r = 0.800) and women (r = 0.796), younger (aged 50 years, r = insulin sensitivity from the mathematical 0.832) and older (r = 0.800) subjects, nonobese (BMI 27 kg/m2, r = 0.800) and obese modeling of fasting plasma glucose and in- (r = 0.765) subjects, nondiabetic (r = 0.754) and diabetic (r = 0.695) subjects, and nor- sulin concentrations. Although the HOMA motensive ( r = 0.786) and hypertensive (r = 0.762) subjects. Also, we found good agree - has been recently used in several clinical ment between the two methods in the categorization of subjects according to insulin sensitiv- ity (weighted k = 0.63). and epidemiological studies (15–24), it has not been definitely validated. Indeed, few CO N C L U S I O N S — We conclude that the HOMA can be reliably used in large-scale or data are available that compare insulin sen- epidemiological studies in which only a fasting blood sample is available to assess in- sitivity estimated by the HOMA with that sulin sensitivity. me a s u r ed by the glucose clamp technique (1 2 , 2 5 , 2 6 ) . Diabetes Care 23 :5 7 –63, 2000 In the present study, we perf o rm e d euglycemic hyperinsulinemic clamp stud- ies in combination with tracer glucose in- everal studies consistently demon- investigational and clinical relevance in fusion to measure insulin-stimulated glu- strated that insulin resistance is a identifying subjects at high risk of type 2 cose disposal in 115 individuals with Ss t rong predictor of type 2 diabetes diabetes and/or cardiovascular disease. various degrees of glucose tolerance and (1,2). More rec e n t l y , insulin resistance has Insulin resistance can be measured by insulin sensitivity and compared the been shown to be associated with prev a - using the glucose clamp technique (6), clamp studies with the estimate of insulin lent atheros c l e r osis (3–5). Thus, the rec o g - which is reg a r ded as the ref e r ence method sensitivity derived by the HOMA. nition of insulin resistance seems to have for an accurate assessment of in vivo in- RESEARCH DESIGN AND ME T H O D S F rom the Division of Endocrinology and Metabolic Diseases, University of Ve rona Medical School, Ve rona, Italy. Ad d r ess correspondence and reprint requests to Prof. Enzo Bonora, Endocrinologia e Malattie del Me- Subjects tabolismo, Ospedale Civile Maggiore, Piazzale Stefani, 1, I-37126 Ver ona, Italy. E-mail: [email protected]. The study included 115 subjects who all Received for publication 22 March 1999 and accepted in revised form 13 September 1999. un d e r went a glucose clamp with the for- Ab b re v i a t i o n s : CV , coefficient of variation; GIR, glucose infusion rate; HOMA, homeostasis model as- sessment; IVGTT, intravenous glucose tolerance test; TGD, total glucose disposal. mat described below in our laboratory in A table elsewhere in this issue shows conventional and Système International (SI) units and conversion 1995 and 1996. A total of 53 subjects with factors for many substances. type 2 diabetes was rec r uited among those

DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 57 HOMA in the assessment of insulin sensitivity

Table 1—Main clinical characteristics of subjects under study ef ficient of variation [CV]) was 5% in all subjects. At 2 h after the beginning of the Su b j e c t s glucose clamp, a prime constant infusion of 3-[3H ] -D-glucose was initiated at the No n d i a b e t i c Type 2 diabetic rate of 0.45 µCi/min and was continued Sex (M/F) 12 / 5 0 37 / 1 6 until the end of the study. The prime dose Age (years) 41.3 ± 1.4 (19 to 63) 55.6 ± 1.0 (31 to 67) of labeled glucose was calculated by divid- Body weight (kg) 76.9 ± 2.2 (53 to 115) 75.9 ± 1.4 (52 to 104) ing the glucose pool (plasma glucose con- BMI (kg/m2) 27.8 ± 0.7 (19 to 45) 27.7 ± 0.6 (20 to 51) centration glucose distribution volume Waist-to-hip ratio 0.88 ± 0.01 (0.73 to 1.04) 0.97 ± 0.01 (0.83 to 1.01) assumed to be 25% of body weight) by Fasting insulin (pmol/l) 63 ± 4 (21 to 182) 96 ± 7 (14 to 224) the product of 1.1 by the glucose infusion Fasting glucose (mmol/l) 5.0 ± 0.05 (4.3 to 6.0) 9.72 ± 0.35 (4.4 to 16.2) rate (GIR) from 100 to 120 min of the

Hb A 1c (%) — 6.6 ± 0.2 (4.2 to 9.0) study and then multiplying the result by HOMA score 2.06 ± 0.14 (0.7 to 6.5) 5.98 ± 0.48 (1.1 to 13.9) the tracer infusion rate. GIR was multi- TGD during clamp (µmol 34.6 ± 1.6 (17.4 to 75.9) 21.0 ± 1.1 (7.8 to 52.5) plied by 1.1 to take into account the ex- mi n –1 kg –1 fa t - f r ee mass) pected 10% average increase in GIR from Glucose rate of appearance 0.92 ± 0.50 ( 6.39 to 8.28) 4.1 ± 0.7 ( 5.4 to 13.4) 100–120 to 180–240 min of the glucose during clamp (µmol mi n –1 clamp. As previously rep o r ted (29), with kg –1 fa t - f r ee mass) this methodological approach, a steady Hy p e r tension (%) ( 160/95 25 . 8 56 . 6 state of tritiated glucose specific activity is mmHg or trea t m e n t ) obtained from 180 to 240 min of the Obesity (%) 41 . 9 37 . 7 clamp. In fact, in the entire group we ex- Data are means ± SEM (range) or %. amined, mean specific activity at 180 min and at 240 min averaged 843 ± 32 and 821 ± 33 dpm/µmol, res p e c t i v e l y , with a regularly attending the Diabetes Clinic at was perf o rmed in combination with CV of 2.5 ± 0.3% from 180 to 240 min. the University of Ver ona who were willing 3- [ 3H] - D-glucose infusion, as prev i o u s l y During this period, blood was drawn to participate in the study. Of these pa- rep o r ted in detail (28), to assess total glu- ev e r y 10 min to measure plasma levels of tients, 10 were treated with diet only, and cose disposal (TGD) accurately. glucose, serum insulin, and plasma triti- 43 were taking oral hypoglycemic agents All studies began at 8:00 A.M. Sub- ated glucose specific activity. Insulin-me- (22 were taking sulfonylureas, 21 were jects were admitted to the hospital after a diated TGD rate was calculated by divid- taking sulfonylureas plus metformin). Pa- 10- to 12-h overnight fast. Briefly, two ing the 3-[3H] - D-glucose infusion rate by tients taking insulin were excluded from Teflon cannulas were inserted: one was the steady-state 3-[3H] - D-glucose specific the study, and 62 nondiabetic subjects in s e r ted into an antecubital vein for infu- ac t i v i t y . More details have been rep o rt e d w e re re c ruited by an advertisement. All sion of insulin, glucose (20% dextrose), and el s e w h e r e (28). pa r ticipants underwent a physical exami- 3- [ 3H] - D-glucose, and the other was in- nation and routine blood chemistry evalu- se r ted into a contralateral heated ( 60 ° C ) Analytical determinations ation. None of them had a history of re- hand vein for arterialized blood sampling. Plasma glucose was measured by using the cent acute illness or clinical evidence of After baseline blood collections for glu- glucose oxidase method on a Beckman ca rd i o v a s c u l a r , kidney, liver, or endocrine cose and insulin determinations, a prime Glucose Analyzer (Fullerton, CA). Serum diseases. Body composition was measured constant (20 mU mi n 1 m 2 body sur- insulin was measured by using a double- by using bioimpedance analysis (27). face area) insulin infusion was started and antibody radioimmunoassay without Main clinical features of the study subjects continued for the subsequent 240 min. cro s s - r eactivity with proinsulin or split- a re shown in Table 1. All subjects gave The insulin prime consisted of two subse- pr oinsulin products (Linco Research, St. their written informed consent to parti c i - quent 5-min periods during which in- Louis, MO). The intra- and interassay CVs pate in the study. The protocol was ap- sulin was infused at the rate of 80 and 40 of serum insulin were 2.9 and 4.7%, re- pr oved by the Ethical Committee of the mU mi n 1 m 2, res p e c t i v e l y . sp e c t i v e l y . Plasma 3-[3H] - D-glucose spe- Ver ona City Hospital. In nondiabetic subjects, serum insulin cific activity was determined as prev i o u s l y in c r eased from an average basal value of described in detail elsewhere (28). Glucose clamp 76 ± 10 to a mean concentration of 292 ± The study consisted of a 4-h euglycemic 38 pmol/l in the last hour of the clamp, HOMA of insulin resistance hyperinsulinemic clamp. The clamp was wh e r eas the values were 108 ± 14 and 304 The estimate of insulin resistance by ca r ried out as originally described by De- ± 42 pmol/l, res p e c t i v e l y , in the diabetic HOMA score was calculated with the for- Fr onzo et al. (6), with the exception of the subjects. Plasma glucose was clamped at mula: fasting serum insulin (µU/ml) amount of insulin infused, which was 5 mmol/l by a variable glucose infusion. fasting plasma glucose (mmol/l)/22.5, as lower in the present study than that used In diabetic subjects, plasma glucose was described by Matthews and coworkers in the pioneering article by DeFronzo et al. left to drop until euglycemia was rea c h e d (12). With such a method, high HOMA The insulin amount was lowered to (generally within 120 min) and then was sc o r es denote low insulin sensitivity (in- achieve serum insulin levels re s e m b l i n g maintained at that level. In the last hour of sulin resistance). The CVs of HOMA those encountered after a meal. The clamp the clamp, plasma glucose variability (co- sc o r es were 9.4 ± 0.7 and 7.8 ± 0.6%, re-

58 DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 Bonora and Associates

metric correlation (Rs = 0.855, P 0.0001). The degree of linear correl a t i o n between log HOMA scores and log TGD rates was very similar between men and women ( 0.800 vs. 0.796), younger ( 50 years) and older ( 0.832 vs. 0.800) subjects, nonobese (BMI 2 7 kg / m 2) and obese ( 0.800 vs. 0. 7 6 5 ) subjects, nondiabetic and diabetic ( 0.745 vs. 0.695) subjects, and nor- motensive and hypertensive ( 0.786 vs. 0.762) subjects. Thus, the results were consistent within all categories of interes t . The agreement in the categorization was good according to insulin sensitivity when subjects were stratified into quintiles of HOMA scores and TGD rates (weighted k = 0.63) (Table 2). In parti c u l a r , when Figure 1—Scatterplot of TGD rates (µmol min 1 kg 1 fat-free mass) in the 4th h of a euglycemic considering that the categories of higher ( 5 mmol/l) hyperinsulinemic ( 300 pmol/l) clamp and HOMA scores in 115 subjects. , Nondia - insulin sensitivity corresponded to quintile betic subjects; , type 2 diabetic subjects. Linear regression: r = 0.627, P 0.0001; polynomial re - V of TGD and to quintile I of HOMA and gression: r = 0.743, P 0.0001. IR, insulin resistance. that the categories of greater insulin res i s - tance corresponded to quintile I of TGD and to quintile V of HOMA, we found sp e c t i v e l y , in 20 nondiabetic subjects and With log-transformed data, the correl a t i o n sa m e - c a t e g o r y agreement in 46% of sub- 20 diabetic individuals in whom this pa- c o e fficient between HOMA scores and jects, one-quintile disagreement in 49% of rameter was measured three times in the TGD rates was 0.820 (P 0.0001) (Fig. subjects, and two-quintile disagreement in fasting state at 5-min intervals. The CVs of 2). The correlation between log HOMA only 5% of subjects. Three- or four-q u i n t i l e HOMA scores were 13.8 and 11.2%, re- sc o r es and log GIR during the clamp was di s a g r eement never occurred . sp e c t i v e l y , in a single nondiabetic subject v i rtually identical (r = 0.801, P and in a single diabetic subject in whom 0.0001). Thus, from a statistical view- CO N C L U S I O N S — Insulin res i s t a n c e this parameter was measured in the morn- point, 65% of the variability of insulin plays a major role in the development of ing for 5 consecutive days. sensitivity assessed by the glucose clamp type 2 diabetes (1,2,30) and may also be technique could be accounted for by involved in atherogenesis (3–5). Thus, the Statistical analysis HOMA. The nonparametric Spearm a n ’s assessment of insulin sensitivity has be- HOMA scores and TGD rates during the rank correlation was similar to the para- come a frequent need for clinical investi- glucose clamp were log-transformed to ap p r oximate a normal distribution. Linear and polynomial reg r essions, Pearson’s sim- ple correlations, and Spearma n ’ s rank cor- relations between HOMA scores and TGD rates were computed. HOMA scores and TGD rates in each individual were strati- fied into quintiles, and Cohen’s k co e ff i - cient of agreement (weighted k) was com- puted. Weights for agreement were set at 0.00 (full disagreement), 0.25, 0.50. 0.75, and 1.00 (full agreement). All data are means ± SEM.

RE S U LT S — The entire group exhib- ited a strong inverse correlation between TGD rates during the glucose clamp and HOMA scores. With raw data, linear re- gr ession analysis gave a Pearson’s correl a - tion coefficient of 0.627 (P 0. 0 0 0 1 ) , but the scatterplot was skewed hyperboli- F i g u re 2—Simple correlation between log-transformed TGD rates (µmol m i n 1 k g 1 fa t - f re e cally (Fig. 1). A curvilinear fitting (polyno- mass) in the 4th hour of a euglycemic ( 5 mmol/l) hyperinsulinemic ( 300 pmol/l) clamp and log- mial reg r ession) of the data gave a correl a - transformed HOMA scores in 115 subjects. , Nondiabetic subjects; , type 2 diabetic subjects. IR, tion coefficient of 0.743 (P 0. 0 0 0 1 ) . insulin resistance.

DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 59 HOMA in the assessment of insulin sensitivity

Table 2—A g reement in the categorization of subjects according to insulin sensitivity, as In our study, the glucose clamp measured by the glucose clamp and as estimated by the HOMA methodology was combined with the glu- cose tracer dilution technique, and the HOMA quintiles study was perf o rmed at euglycemia in both nondiabetic and diabetic subjects. TGD quintiles I II II I IV V Thus, the experimental conditions were I 0 0 1 7 15 the same in all subjects, which thereby al- II 0 1 7 9 6 lowed us to gather a comparable measure- II I 0 7 8 6 2 ment of TGD during insulin infusion. In IV 9 7 6 1 0 this context, HOMA scores and TGD rates V 14 8 1 0 0 we r e strongly correlated, and the res u l t s Subjects were stratified into quintiles of TGD rates and HOMA scores. Number of observations in each quin- we r e consistent in the various subgrou p s tile is rep o r ted. Weighted k co e f ficient of agreement = 0.63. Agreement was based on the assumption that max- that we examined (men vs. women, older imum insulin sensitivity corresponded to quintile I of HOMA and quintile V of TGD, and minimum insulin vs. younger subjects, obese vs. nonobese sensitivity corresponded to quintile V of HOMA and quintile I of TGD. subjects, diabetic vs. nondiabetic subjects, and hypertensive vs. normotensive sub- gators and epidemiologists and may also An attractive approach to estimate in- jects). Taken together, these results sup- be useful for clinicians. sulin sensitivity (or insulin res i s t a n c e ) po r t the use of the HOMA as a surrog a t e In recent years, a new clinical syn- seems to be the HOMA, which was devel- index of insulin sensitivity in humans. Of dr ome has been described that features the oped by Matthews et al. (14) with com- course, this conclusion relies on the as- clustering of several metabolic, hemody- pu t e r -aided modeling of fasting glucose and sumption that TGD during the clamp is namic, and hemocoagulative disord e r s insulin concentrations. These authors the ref e r ence measure of insulin sensitivity. (31,32). Although this syndrome has been found that the HOMA-based insulin res i s - The HOMA score does not measure given many names, its most popular name tance score was strongly correlated with in- the amount of glucose metabolized per is “insulin resistance syndrome.” This sulin sensitivity assessed by the glucose unit of body weight or lean body mass name is justified by the idea that insulin clamp technique in both nondiabetic and during whole-body insulinization; rather, resistance is the major common denomi- diabetic subjects (r = 0.83 and 0.92, re- the HOMA score explores the sponta- nator of the abnormalities involved in the spectively) (14). However, validation of the neous homeostatic characteristics of a sy n d r ome. Unfortu n a t e l y , the vast major- HOMA was carried out in only a few sub- metabolic system by inferring what de- ity of data contributing to our knowledge jects (12 nondiabetic and 11 diabetic), and g ree of insulin sensitivity is compatible about this syndrome is not based on the the glucose clamp studies were not per- with these homeostatic characteristics. tr ue measurement of insulin res i s t a n c e . fo r med in conjunction with glucose tracer Ne v e r theless, the HOMA ranks individu- During the last 20 years, the assess- infusion, so glucose disposal could not be als similarly to the glucose clamp tech- ment of in vivo insulin sensitivity in hu- assessed accurately. Indeed, endogenous nique. In fact, in a large number of indi- mans has been frequently based on the use glucose production is not always com- viduals with various degrees of glucose of the glucose clamp technique (6). This pletely suppressed by physiological hyper- tolerance and insulin resistance, we found technique is considered the “gold stan- insulinemia, especially in diabetic subjects a strong correlation between the insulin da r d” (7), although it does not mirror the (28–30), and the GIR, which maintains eu- sensitivity values generated by the two physiological condition of continuously glycemia during the glucose clamp, can un- tests. In this re g a rd, the HOMA score changing glucose and insulin levels (with de r estimate the exact rate of glucose dis- seems to be as good a predictor of clamp- mutual control of the hormone on the sub- posal. In another validation study, Anderson de t e r mined insulin sensitivity as the short strate and vice versa) and of differing in- et al. (25) compared the HOMA with the insulin tolerance test (11) or the intra- sulin exposure in the liver and peripheral glucose clamp technique in a rel a t i v e l y venous glucose tolerance test (IVGTT) an- tissues. In this reg a r d, one may conclude gr eater number of subjects (n = 55), half of alyzed with the minimal model (34–37), that no method will ever be capable of whom had type 2 diabetes, and found a the method often indicated as the best al- tr uly measuring insulin sensitivity, but the weaker correlation between the two mea- te r native to the glucose clamp technique glucose clamp technique is the method su r es (r = 0.40). However, these authors (3,38,39). HOMA precision (rep ro d u c i b i l - with the fewest drawbacks, and it yields re- pe rf o r med isoglycemic and not euglycemic it y of the measure) seems to be compara- sults closest to the real measure. Unfortu - clamp studies, which thereby led to an ble to the glucose clamp technique and na t e l y , this technique is not suited for large - ov e r estimation of insulin sensitivity in displays a CV (10–15%) that is similar to scale or epidemiological studies because of hy p e r glycemic individuals because of the that observed with the glucose clamp (40; its complexity and high cost. Alterna t i v e mass effect of glucose (33). In the only R.C.B., unpublished observations). Un- methods have been proposed and used in other validation study we are aware of, qu e s t i o n a b l y , the HOMA is inferior to the clinical investigations, but none of them is Emoto et al. (26) compared HOMA and the glucose clamp technique in terms of its adequate for studies involving hundreds or glucose clamp technique in 80 type 2 dia- accuracy in assessing insulin sensitivity, thousands of subjects. Indeed, all of these betic subjects and found a good rel a t i o n - but the trade-off for this limitation lies in alternative methods include injections ship between the two measures of insulin the ease with which large numbers of sub- and/or infusions of hormones, drugs, or sensitivity (r = 0.725, P 0.001). Unfor- jects can be examined with a single glu- substrates as well as drawing several timed tu n a t e l y , no isotopic evaluation of TGD was cose and insulin measurement in the fast- blood samples (8–13). pursued in this study. ing state.

60 DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 Bonora and Associates

One could argue that we have con- substrate-1 in type 2 diabetic subjects that should be done with great caution. For the ducted clamp studies at relatively low in- resulted in high HOMA scores. Finally, same reason, the use of HOMA in the clin- sulin concentrations. Thus, the question Kumar et al. (20) re p o rted an impro v e- ical setting to identify insulin-resistant sub- arises whether the agreement of HOMA-es- ment in HOMA-estimated insulin resis- jects is not recommended because a cutoff timated and clamp-measured insulin sensi- tance after troglitazone treatment of type 2 value for insulin resistance estimated with tivity persists when the clamp experiment diabetic subjects in agreement with trog l i - the HOMA would not be easily defined. Of is perfo r med at higher insulin concentra- tazone effects observed in glucose clamp course, standardization of the insulin assay tions. During the years, we have collected a studies (45). Thus, the HOMA can unravel may circumvent these problems. Unfortu - la r ge number of clamp studies carried out insulin resistance, whether it is associated na t e l y , standardization of the insulin assay at 40 or 100 mU mi n 1 m 2 su rf a c e with type 2 diabetes or pre–type 2 dia- is far from becoming a rea l i t y . area. We have recently computed the betes, or with a disruption of insulin sig- One could ask whether a physiological HOMA scores in subjects undergoing these naling at the molecular level. basis underlies the strong relationship we studies and have found strong relation- Although definite proof is lacking, it ob s e r ved between HOMA scores and ships with TGD rates during the clamp. seems reasonable that HOMA cannot be clamp TGD rates. Indeed, the HOMA is a For example, in a group 19 subjects with used in patients with type 1 diabetes or in parameter that essentially explores the abil- various degrees of insulin sensitivity and patients with type 2 diabetes receiving in- ity of insulin to restrain hepatic glucose glucose tolerance undergoing a 40-mU sulin treatment because assessing the spon- pr oduction in the fasting state because mi n 1 m 2 su r face area clamp (11), we taneous homeostatic characteristics of the basal insulin has a substantial effect on he- found a correlation coefficient of 0. 8 8 1 . metabolic system in these individuals is patic glucose production (50) but a quanti- In another series of 47 studies in which the not possible. However, the data above re- tatively poor (if any) effect on peripheral clamp was carried out at supraphysiologi- p o rting on the consistency of HOMA glucose disposal (51). On the contrary, cal insulin concentrations (100 mU sc o r es in type 2 diabetic subjects exam- clamp TGD is a function mainly of periph- mi n 1 m 2 su r face area) (41–43), the co- ined at spontaneous fasting hypergl y c e m i a eral responses to higher insulin concentra- efficient of correlation between HOMA and at insulin-induced euglycemia could tions (28–30). We hypothesize that HOMA sc o r es and TGD rates was 0. 7 6 2 . challenge this belief. Further studies are sc o r es and clamp TGD rates are stron g l y One could also argue that the HOMA needed to clarify this aspect. Regardl e s s , co r related because, in virtually all clinical sc o r e is calculated in the fasting condition, in s u l i n - t r eated diabetic subjects rep re s e n t conditions that are characterized by pe- wh e r eas the TGD rate in diabetic subjects only a limited fraction of the diabetic pop- ripheral insulin resistance, hepatic insulin was calculated at euglycemia and not at ulation (46). Specific studies are needed resistance is also evident (28,30,32). This is fasting ambient hyperglycemia. Thus, one reg a r ding the capability of HOMA to as- pr obably because of the fact that peripheral may wonder whether the two measure- sess insulin sensitivity in particular type 2 insulin resistance of glucose metabolism is ments are comparable. We addressed this diabetes phenotypes such as subjects with generally associated with an impaired in- issue by comparing HOMA scores in five poor insulin secretion and normal insulin su l i n - s u p p r essed lipolysis (41,43). Indeed, subjects with type 2 diabetes whom we sensitivity or with severe insulin res i s t a n c e the insulin-signaling mechanisms that con- had examined both in the spontaneous and exhausted -c e l l s . tr ol glucose metabolism in the skeletal fasting state (10.2 ± 1.1 mmol/l) and a few In the three previous validation studies muscle and lipid metabolism in adipose tis- days later after an overnight infusion of (12,25,26), serum insulin was measured by sue are somewhat common (52,53) and are low-dose insulin to achieve euglycemia using a standard (not human insulin-spe- thought to be disrupted in insulin-res i s t a n t the next morning (4.8 ± 0.4 mmol/l) (44). cific) radioimmunoassay. In our study, we conditions (53). One consequence of the The mean HOMA scores were quite simi- used an insulin-specific radioimmunoassay deranged insulin signaling in adipose tissue lar (6.98 ± 1.88 at hyperglycemia and 5.89 with no significant cros s - r eactivity with is an exaggerated lipolysis that leads to an ± 2.17 at euglycemia, NS), and the mea- pr oinsulin or split-proinsulin prod u c t s , in c r eased flux of free fatty acids from su r es on the two separate occasions were th e r eby minimizing the interfe r ence ex- adipocytes to the liver, where they con- st r ongly correlated (r = 0.889). er ted on the HOMA score by raised plasma tribute to the diminished ability of insulin Fu r ther support for more widesprea d pr oinsulin levels, such as those possibly en- to suppress hepatic glucose prod u c t i o n use of the HOMA comes from several re- co u n t e r ed in diabetic subjects (47,48). (54). Thus, that a parameter addressing es- cent studies. For example, Haffner et al. HOMA is based on measuring plasma sentially hepatic insulin resistance (HOMA (18) found that insulin resistance esti- glucose and serum/plasma insulin. Al- sc o r e) is strongly related to a parameter de- mated by the HOMA predicts the develop- though the plasma glucose assay is known scribing to a greater extent peripheral in- ment of type 2 diabetes in Mexicans, which to be highly rep r oducible in diffe r ent labo- sulin resistance (clamp TGD rate) is not is similar to the observations made in Pima ratories, insulin assay can vary consider- su r p r i s i n g . Indians (2) and in Caucasians (1) with the ab l y , especially if antibodies cros s - re a c t i n g In conclusion, our data suggest that the glucose clamp technique and the IVGTT with proinsulin or split-proinsulin prod - HOMA is a valuable alternative to more so- analyzed with minimal model, res p e c t i v e l y . ucts are used (49). The use of specific anti- phisticated techniques in the evaluation of Fu rt h e rm o r e, Clement et al. (17) found insulin antibodies is becoming common, in vivo insulin sensitivity in humans. The that glucokinase-deficient subjects are in- and fewer discrepancies among laborato- HOMA seems to be specifically suited to sulin resistant either when scored by the ries are expected to occur in the future. At la r ge-scale studies in which only fasting HOMA or when evaluated by the glucose present, however, comparing HOMA blood samples are available. Neverth e l e s s , clamp technique, yet Zhang et al. (24) sc o r es generated in diffe r ent laboratories comparing HOMA scores obtained in dif- found a mutation of the insulin rec e p t o r that used diffe r ent insulin assay materials fe r ent studies cannot be done unless the in-

DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 61 HOMA in the assessment of insulin sensitivity sulin assay is standardized. Standardi z a t i o n ti v i t y . Am J Physiol 236:E667–E677, 1979 BMJ 313:1040–1044, 1996 of the insulin assay is also a prer equisite for 11 . Bonora E, Moghetti P, Zancanaro C, 22 . Hu m p h r eys P, McCarthy M, Tuomilehto J, in t r oducing the HOMA in clinical practice. Cigolini M, Querena M, Cacciatori V, Tuo m i l e h t o - W olf E, Stratton I, Morgan R, C o rgnati A, Muggeo M: Estimates of in Rees A, Owens D, Stengard J, Nissinen A, vivo insulin action in man: comparison of Hitman G, Tur ner RC, O’Rahilly S: Chro- insulin tolerance tests and euglycemic and mosome 4q locus associated with insulin A c k n o w l e d g m e n t s — This study was sup- h y p e rglycemic glucose clamp studies. resistance in Pima Indians: studies in three ported by grants from the Italian National Re- J Clin Endocrinol Metab 68:374–378, 1989 Eu r opean NIDDM populations. Di a b e t e s s e a rch Council and the Italian Ministry of 12 . Galvin P, Wa rd G, Walters J, Pestell R, 43:800–804, 1994 University and Scientific and Te c h n o l o g i c a l Koschmann M, Vaag A, Martin I, Best JD, 23 . Yamada K, Yuan X, Ishiyama S, Koyama K, Research. Al f o r d F: A simple method for quantita- Ichikawa F, Koyanagi A, Koyama W, Non- The authors gratefully acknowledge the tion of insulin sensitivity and insulin re- aka K: Association between Ala54Thr sub- skillful technical assistance of Monica Zardi n i lease from an intravenous glucose toler- stitution of the fatty acid-binding protein 2 and Federica Moschetta. ance test. Diabet Med 9:921–928, 1992 gene with insulin resistance and intra-ab- 13 . Piatti PM, Monti LD, Caumo A, Santam- dominal fat thickness in Japanese men. br ogio G, Magni F, Galli Kienle M, Costa S, Di a b e t o l o g i a 40:706–710, 1997 Re f e re n c e s Po n t i r oli A, Alberti KGMM, Pozza G: The 24 . Zhang Y, Wat N, Stratton IM, Wa rre n - 1. M a rtin BC, Wa rram JH, Krolewski AS, continuous low dose insulin infusion test: Pe r ry MG, Orho M, Groop L, Tur ner RC: Be r gman RN, Soeldner JS, Kahn CR: Role a simplified and accurate method for the UKPDS 19: Heterogeneity in NIDDM: sep- of glucose and insulin resistance in devel- evaluation of insulin sensitivity and insulin arate contributions of IRS-1 and beta-3 opment of type 2 diabetes mellitus: res u l t s s e c retion in population studies. J Clin ad re n e rg i c - r eceptor mutations to insulin of a 25-year follow-up study. La n c e t 34 0 : Endocrinol Metab 1:34–40, 1995 resistance and obesity respectively with no 925–929, 1992 14 . Matthews DR, Hosker JP, Rudenski AS, evidence for glycogen synthase gene muta- 2. Lillioja S, Mott DM, Spraul M, Ferra r o R, Naylor BA, Treacher DF, Tu rner RC: tions. Di a b e t o l o g i a 39:1505–1511, 1996 Foley JE, Ravussin E, Knowler WC, Ben- Homeostasis model assessment: insulin re- 25 . Anderson RL, Hamman RF, Savage PJ, nett PH, Bogardus C: Insulin re s i s t a n c e sistance and -cell function from fasting Saad MF, Laws A, Kades WW, Sands RE, and insulin secret o r y dysfunction as pre- plasma glucose and insulin concentrations Cefalu W for the Insulin Resistance Ather- cursors of non-insulin-dependent diabetes in man. Di a b e t o l o g i a 28:412–419, 1985 os c l e r osis Study: Exploration of simple in- mellitus: prospective studies of Pima Indi- 15 . United Kingdom Prospective Diabetes sulin sensitivity measures derived fro m an s . N Engl J Med 329:1988–1992, 1993 Study Group: UK Prospective Diabetes f requently sampled intravenous glucose 3. Ho w a r d G, O’Leary DH, Zaccaro D, Haffn e r St u d y . V. Characteristics of newly pres e n t - tolerance (FSIGT) tests: the Insulin Resis- S, Rewers M, Hamman R, Selby JV, Saad ing type 2 diabetic patients: estimated in- tance Atheros c l e r osis Study. Am J Epidemiol MF , Savage P, Bergman R: Insulin sensitivity sulin sensitivity and islet beta-cell function: 142:724–732, 1995 and atheros c l e r osis: the Insulin Resistance mu l t i c e n t r e study. Diabet Med 5: 4 4 4 – 4 4 8 , 26 . Emoto M, Nishizawa Y, Maekawa K, Hiura At h e ro s c l e r osis Study (IRAS) Investigators. 19 8 8 Y, Kanda H, Kawagishi T, Shoji T, Okuno Y, Ci rc u l a t i o n 93:1809–1817, 1996 16 . Ha f fner SM, Miettinen H, Stern MP: The Morii H: Homeostasis model assessment as 4. Bonora E, Tessari R, Micciolo R, Zenere M, homeostasis model in the San Antonio a clinical index of insulin resistance in type Tar gher G, Padovani R, Falezza G, Muggeo He a r t Study. Diabetes Care 20 : 1 0 8 7 – 1 0 9 2 , 2 diabetic patients treated with sulfonyl- M: Intimal-medial thickness of the carot i d 19 9 7 ur eas. Diabetes Care 22:818–822, 1999 art e r y in nondiabetic and non-insulin-de- 17 . Clement K, Pueyo ME, Vax i l l a i r e M, Rako- 27 . Lukaski HC, Johnson PE, Bolonchuk WW, pendent diabetic subjects: re l a t i o n s h i p toambinina B, Thuillier F, Passa P, Frog u e l Lykken GI: Assessment of fat-free mass with insulin resistance. Diabetes Care 20 : P, Robert JJ, Velho G: Assessment of insulin using bioelectrical impedance measure- 627–631, 1997 sensitivity in glucokinase-deficient sub- ments of the human body. Am J Clin Nutr 5. Yip J, Facchini F, Reaven GM: Resistance to jects. Di a b e t o l o g i a 39:82–90, 1996 41:810–817, 1985 insulin-mediated glucose disposal as a pre- 18 . H a ffner SM, Gonzales C, Miettinen H, 28 . Bonora E, Del Prato S, Bonadonna RC, dictor of cardiovascular disease. J Clin En - Kennedy E, Stern MP: A pro s p e c t i v e Gulli G, Solini A, Shank ML, Ghiatas AA, docrinol Metab 88:2773–2776, 1998 analysis of the HOMA model: the Mexico Lancaster JL, Kilcoyne RF, Alyassin AM, 6. De F r onzo RA, Tobin JD, Andres R: Glu- City Diabetes Study. Diabetes Care 1 9 : De F r onzo RA: Total body fat content and cose clamp technique: a method for quan- 1138–1141, 1996 fat topography are associated diffe re n t l y tifying insulin secretion and resistance. Am 19 . St e w a r t MW, Humphriss DB, Berrish TS, with in vivo glucose metabolism in J Physiol 237:E214–E223, 1979 B a rriocanal LA, Trajano LR, Albert i nonobese and obese nondiabetic women. 7. American Diabetes Association: Consensus KGMM, Walker M: Features of syndrom e Di a b e t e s 41:1151–1159, 1992 Development Conference on Insulin Resis- X in first-degree relatives of NIDDM pa- 29 . Bonadonna RC, Del Prato S, Bonora E, tance. Diabetes Care 21:310–314, 1998 tients. Diabetes Care 18:1020–1022, 1995 Gulli G, Solini A, DeFronzo RA: Effects of 8. Shen SW, Reaven GM, Farquhar JW: Com- 20 . Kumar S, Boulton AJM, Beck-Nielsen H, physiological hyperinsulinemia on the in- parison of impedance to insulin-mediated B e rthezene F, Muggeo M, Persson B, tracellular metabolic partition of plasma glucose uptake in normal subjects and in Spinas GA, Donoghue S, Lettis S, Stewart- glucose. Am J Physiol 2 6 5 : E 9 4 3 – E 9 5 3 , subjects with latent diabetes. J Clin Invest Long P for the Troglitazone Study Grou p : 19 9 3 49:2151–2160, 1970 Troglitazone, an insulin action enhancer, 30 . Del Prato S, Bonadonna RC, Bonora E, 9. Harano Y, Ohgaku S, Hidaka H, Haneda im p r oves metabolic control in NIDDM pa- Gulli G, Solini A, Shank M, DeFronzo RA: K, Kikkawa R, Shigeta Y, Abe H: Glucose, tients. Di a b e t o l o g i a 39:701–709, 1996 Characterization of cellular defects of in- insulin and somatostatin infusion for the 21 . Kiechl S, Willeit S, Poewe S, Egger G, sulin action in type 2 (non-insulin-depen- de t e r mination of insulin sensitivity. J Clin Ob e r hollenzer F, Muggeo M, Bonora E: In- dent) diabetes mellitus. J Clin Invest 9 1 : Endocrinol Metab 45:1124–1127, 1977 sulin sensitivity and regular alcohol con- 484–494, 1993 10 . Be r gman RN, Ider YZ, Bowden CR, Cobelli sumption: large, prospective, cro s s - s e c- 31 . Reaven GM: Banting Lecture 1988: role of C: Quantitative estimation of insulin sensi- tional population study (Bruneck Study). insulin resistance in human disease. Di a -

62 DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 Bonora and Associates

be t e s 37:1595–1607, 1988 tiveness in a population-based sample of M, Iwamoto Y: Serum proinsulin levels at 32 . De F r onzo RA, Ferrannini E: Insulin res i s - 380 young healthy Caucasians: analysis of fasting and after oral glucose load in patients tance: a multifaceted syndrome res p o n s i - the impact of gender, body fat, physical fit- with type 2 (non-insulin-dependent) dia- ble for NIDDM, obesity, hypert e n s i o n , ness, and life-style factors. J Clin Invest betes mellitus. Di a b e t o l o g i a 31 : 3 5 5 – 3 6 0 , dyslipidemia and athero s c l e rotic card i o- 98:1195–1209, 1996 19 8 8 vascular disease. Diabetes Care 1 4 : 1 7 3 – 40 . Scheen AJ, Paquot N, Castillo MJ, Lefebvre 48 . Phillips DIW, Clark PM, Hales CN, Os- 194, 1991 PJ: How to measure insulin action in vivo. mond C: Understanding oral glucose tol- 33 . De F r onzo RA, Ferrannini E: Influence of Diabetes Metab Rev 10:151–188, 1994 erance: comparison of glucose or insulin plasma glucose and insulin concentration 41 . Gr oop LC, Bonadonna RC, Del Prato S, me a s u r ements during the oral glucose tol- on plasma glucose clearance in man. Di a - Retheiser K, Zyck K, Ferrannini E, De- erance test with specific measurements of betes 31:683–688, 1982 Fr onzo RA: Glucose and free fatty acid me- insulin resistance and insulin secre t i o n . 34 . Saad MF, Anderson RL, Laws A, Wat a n a b e tabolism in non-insulin-dependent diabetes Diabet Med 11:286–292, 1994 RM, Kades WW, Chen YDI, Sands RE, Pei mellitus: evidence for multiple sites of in- 49 . Robbins DC, Andersen L, Bowsher R, D, Savage PJ, Bergman RN for the Insulin sulin resistance. J Clin Invest 84 : 2 0 5 – 2 1 3 , Chance R, Dinesen B, Frank B, Gingerich Resistance Atheros c l e r osis Study: A com- 19 8 9 R, Goldstein D, Widemeyer HM, Haffn e r parison between the minimal model and 42 . Bonadonna RC, Groop LC, Shank M, De- S, Hales CN, Jarrett L, Polonsky K, Porte the glucose clamp in the assessment of in- Fr onzo RA: Dose-dependent effect of in- D, Skyler J, Webb G, Kallagher K: Report sulin sensitivity across the spectrum of glu- sulin on plasma free fatty acid turno v e r of the American Diabetes Association’s cose tolerance. Diabetes 43:1114–1121, and oxidation in humans. Am J Physiol Task Force on Standardization of the In- 19 9 4 259:E736–E750, 1990 sulin Assay. Di a b e t e s 45:242–256, 1996 35 . Saad MF, Steil GM, Kades WW, Ayad MF, 43 . G roop LC, Salorante C, Shank M, 50 . Del Prato S, Gulli G, Bonadonna RC: Effe c t Elsewafy WA, Boyadjian R, Jinagouda SD, Bonadonna RC, Ferrannini E, DeFron z o of hyperglycemia on glucose metabolism: B e rgman RN: Diff e rences between the RA: The role of free fatty acid metabolism reg u l a t o r y role of basal plasma insulin lev- tolbutamide-boosted and the insulin- in the pathogenesis of insulin resistance in els. Di a b e t e s 41:188A, 1992 modified minimal model protocols. Di a - obesity and noninsulin-dependent dia- 51 . Del Prato S, Riccio A, Vigili de Kreu t z e n - be t e s 46:1167–1171, 1997 betes mellitus. J Clin Endocrinol Metab 72 : b e rg S, Dorella M, Tiengo A, DeFro n z o 36 . Be a r d JC, Bergman RN, War d WK, Porte 96–107, 1991 RA: Basal plasma insulin levels exert a DP Jr: The insulin sensitivity index in non- 44 . Bonadonna RC, Del Prato S, Bonora E, qualitative but not a quantitative effect on diabetic man: correlation between clamp- Saccomani MP, Gulli G, Natali A, Frascerra glucose-mediated glucose uptake. Am J derived and IVGTT-derived values. Di a - S, Pecori N, Ferrannini E, Bier D, Cobelli Ph y s i o l 268:E1089–1095, 1995 be t e s 35:362–369, 1986 C, DeFronzo RA: Roles of glucose trans- 52 . Zierath JR, Livingston JN, Thorne A, 37 . Be r gman RN, Prager R, Volund A, Olefsky po r t and glucose phosphorylation in mus- Bolinder J, Reynisdottir S, Lonnqvist F, JM: Equivalence of the insulin sensitivity cle insulin resistance of NIDDM. Di a b e t e s Ar ner P: Regional diffe r ences in insulin in- index in man derived by the minimal 45:915–925, 1996 hibition of non-esterified fatty acid rel e a s e model method and the euglycemic glucose 45 . Suter SL, Nolan JJ, Wallace P, Gumbiner fr om human adipocytes: relation to insulin clamp. J Clin Invest 79:790–800, 1987 B, Olefsky JM: Metabolic effects of new receptor phosphorylation and intracellular 38 . Ha f fner SM, Howard G, Mayer E, Bergm a n oral hypoglycemic agent CS-045 in signaling through the insulin receptor sub- RN, Savage PJ, Rewers M, Mykkanen L, NIDDM subjects. Diabetes Care 15 : 1 9 3 – strate-1 pathway. D i a b e t o l o g i a 4 1 : 1 3 4 3 – Ka r ter AJ, Hamman R, Saad MF: Insulin 203, 1992 1354, 1998 sensitivity and acute insulin response in 46 . Muggeo M, Verlato G, Bonora E, Bressan F, 53 . Virkamaki A, Ueki K, Kahn CR: Prot e i n – African-Americans, non-Hispanic whites, G i rotto S, Corbellini M, Gemma ML, pr otein interaction in insulin signaling and and Hispanics with NIDDM: the Insulin Moghetti P, Zenere M, Cacciatori V, Zop- the molecular mechanisms of insulin res i s - Resistance Atheros c l e r osis Study. Di a b e t e s pini G, de Marco R: The Ver ona Diabetes tance. J Clin Invest 103:931–943, 1999 46:63–69, 1997 Study: a population-based survey on 54 . Rebrin K, Steil GM, Mittleman SD, Bergm a n 39 . Clausen JO, Borch-Johnsen K, Ibsen H, known diabetes mellitus prevalence and RN: Causal linkage between insulin sup- B e rgman RN, Hougaard P, Winther K, 5-year all-cause morta l i t y . Di a b e t o l o g i a 38 : pr ession of lipolysis and suppression of liver Pedersen O: Insulin sensitivity index, 18–325, 1995 glucose output in dogs. J Clin Invest 93 : acute insulin response, and glucose effe c - 47 . Yoshioka N, Kuzuya T, Matsuda A, Tan i g u c h i 741–749, 1996

DIABETES CARE, VOLUME 23, NUMBER 1, JANUARY 2000 63