International Journal of Obesity (1999) 23, 1307±1313 ß 1999 Stockton Press All rights reserved 0307±0565/99 $15.00 http://www.stockton-press.co.uk/ijo Effect of obesity and insulin resistance on resting and glucose-induced thermogenesis in man

S Camastra1, E Bonora, S Del Prato, K Rett, M Weck and E Ferrannini1* on behalf of EGIR (European Group for the Study of Insulin Resistance){

1Department of Internal Medicine, and Metabolism Unit of the CNR Institute of Clinical Physiology, University of Pisa School of Medicine, Pisa, Italy

OBJECTIVE: to assess the impact of obesity and insulin sensitivity on resting (REE) and glucose-induced thermo- genesis (GIT). DESIGN: Data from 322 studies carried out in non-diabetic subjects of either gender, covering a wide range of age (18 ± 80 y) and body mass index (BMI, 18 ± 50 kg=m2). MEASUREMENTS: Insulin sensitivity and thermogenesis were measured by combining the euglycaemic insulin clamp technique with indirect calorimetry. RESULTS: REE was inversely related to age (P ˆ 0.001) and the respiratory quotient (P ˆ 0.03), and positively related to BMI, lean body mass (LBM), fat mass, and percentage fat mass (all P < 0.0001). In a multiple regression model, LBM- adjusted REE was estimated to decline by 9% between 18 and 80 y, independently of obesity and insulin sensitivity. In contrast, GIT was strongly associated with insulin sensitivity (P < 0.0001) but not with gender, age or BMI. By multiple regression analysis, GIT was linearly related to insulin sensitivity after controlling for gender, age, BMI and steady- state plasma insulin levels. Furthermore, both of the main components of insulin-mediated glucose disposal (glucose oxidation and glycogen synthesis) correlated with GIT independently of one another. In the subset of subjects (n ˆ 89) in whom waist-to-hip ratio (WHR) measurements were available, GIT was inversely associated with WHR (P < 0.001 after adjustment by gender, age, BMI, insulin sensitivity and steady-state plasma insulin concentration). In this model, a signi®cant interaction between WHR and gender indicated a stronger adverse effect on GIT of a high WHR in women than in men. CONCLUSIONS: In healthy humans, age, lean mass and respiratory quotient are the main independent determinants of resting thermogenesis. In contrast, insulin sensitivity and, to a lesser extent, abdominal obesity are the principal factors controlling glucose-induced thermogenesis.

Keywords: thermogenesis; insulin resistance; obesity; diet-induced thermogenesis; energy expenditure; body mass; fat distribution

Introduction between obesity and impaired DIT has received con- siderable attention over the last decade, but results have been controversial. Pittet et al 1 and, later, Golay The ingestion or infusion of nutrients is associated et al 2 demonstrated that the thermic response to the with an increase in metabolic rate, termed diet- ingestion of glucose is reduced in obese compared induced thermogenesis (DIT). The association with lean subjects. This observation was supported by other studies, which found DIT to be decreased in obese individuals also in response to a mixed meal.3±6 *Correspondence: E Ferrannini, CNR Institute of Clinical Consequently, it was postulated that this defect might Physiology, Via Savi 8, 56126 Pisa, Italy. E-mail: [email protected]. be involved in the aetiology of obesity or in the { H Beck-Nielsen (University of Odense, Denmark), P Bell maintenance of the obese state. However, other stu- (University of Belfast, UK), E Bonora (University of Verona, Italy), dies failed to ®nd a decreased thermic effect of a B Capaldo (Federico II University, Naples, Italy), P Cavallo-Perin 7±9 (University of Turin, Italy), S Del Prato (University of Padova, mixed meal in obese subjects. These contrasting Italy), E Ferrannini (CNR Institute of Clinical Physiology, Pisa, results have led investigators to try and determine Italy), D Fliser (University of Heidelberg, Germany), A Golay which feature of overweight could be linked with the (University of Geneva, Switzerland), LC Groop (Lund University, 10 Sweden), S Jacob (Stadtklinik, Baden-Baden, Germany), M termogenic defect. Segal et al demonstrated that Laakso (University of Kuopio, Finland), N Lalic (University of glucose-induced thermogenesis (GIT) is reduced in Belgrade, Yugoslavia), G Mingrone (Catholic University, Rome, young men in proportion to the increase in percentage Italy), A Mitrakou (University of Athens, Greece), G Paolisso (University of Naples II, Napoli, Italy), K Rett (University of body fat even at the same total body weight, suggest- MuÈ nchen, Germany), U Smith (University of GoÈ teborg, Sweden), ing a role of body composition in the impaired GIT of M Weck (Kreischa, Germany) and H Yki-JaÈ rvinen (University of the obese. On the other hand, other studies have Helsinki, Finland). Received 17 February 1999; revised 25 May 1999; accepted focussed on insulin resistance as a cause of defective 30 June 1999 thermogenesis.11,12 As insulin resistance is more Insulin resistance and thermogenesis S Camastra et al 1308 frequent in obese than lean persons, the separate insulin infusion rate of 1 mU=min per kg of body impact of obesity and insulin resistance on GIT has weight (7 pmol=kg). Brie¯y, polyethylene cannulas been dif®cult to evaluate. In studies combining the were inserted into an antecubital vein (for the infusion euglycaemic insulin clamp technique with indirect of glucose and insulin) and retrogradely into a wrist calorimetry in 32 young men, Segal et al 12 concluded vein heated at 60C in a hot box or a heating pad (for that a blunted GIT is rather related to insulin resis- intermittent blood sampling of arterialised venous tance than obesity per se. Whether this ®nding also blood). At time zero, a primed-constant infusion of applies to women, whether it is in¯uenced by age, and regular insulin was begun, and continued for 120 min. to what extent it may be modulated by body fat Four minutes into the insulin infusion, an exogenous distribution has not been determined. glucose infusion was started, and adjusted every 5 ± In the present work, we addressed these questions 10 min in order to maintain plasma glucose within by using the EGIR (European Group for the Study of  10% of its baseline value. Blood samples were Insulin Resistance) database, a collection of insulin obtained at timed intervals in the fasting state and clamp studies carried out in non-diabetic, normo- during the clamp, for the measurement of plasma tensive Europeans.13 From this database, 322 studies glucose and insulin concentrations. combining indirect calorimetry with a standard insulin Indirect calorimetry was performed with the use of clamp were analysed with the aim of establishing the a computerised, continuous, open-circuit system with general physiological determinants of both resting and a canopy (using the Metabolic Measurement Cart glucose-induced thermogenesis in healthy humans. Horizon, SensorMedics, CA, or the DeltaTrac, Sen- sorMedics, Helsinki, Finland).

Materials and methods Analytical procedures Plasma glucose was measured by the glucose oxidase method. Plasma insulin concentrations were measured Subjects by radioimmunoassay. Twenty clinical research centres in Europe agreed to pool their available clamp studies on the condition that the study subjects met the following criteria: (a) Data analysis no clinical or laboratory evidence of cardiac, renal, Lean body mass (LBM) was calculated by Hume's 16 liver or endocrine disease; (b) a fasting plasma glu- formula; fat mass was then obtained as the differ- cose concentration < 6.7 mmol=l and normal glucose ence between body weight and LBM. In two-thirds of tolerance by WHO criteria;14 (c) normal blood pres- the study subjects (n ˆ 213), a direct measurement of sure ( < 160=95 mmHg); (d) no recent change LBM, performed in eight centres by electrical bioim- (  10%) in body weight; and (e) no current medica- pedance (or the equivalent tritiated water techni- 17 tion. In nine centres (two in Scandinavia, four in que), was available. Since the measured and central Europe, and three in Italy) the clamp studies estimated LBM values were highly correlated were combined with indirect calorimetry, yielding a (r ˆ 0.84, P < 0.0001), the estimated LBM value total of 322 cases, which form the database for the was used for the entire dataset to normalise metabolic present report. At each centre, the protocol was and thermogenic rates per unit of lean tissue. In reviewed and approved by the local Ethics Commit- addition, the pattern of results was virtually tee, and informed consent was obtained from all unchanged when, instead of expressing REE and subjects before their participation. GIT per kg of LBM, LBM itself was used as a covariate of absolute energy expenditure values. Insulin action was expressed at the total glucose Protocol disposal rate during steady-state euglycaemic hyper- The minimum of information required for each case insulinaemia. With the insulin dose used in the current was age, anthropometric variables, and fasting and study, hepatic glucose output has been previously steady-state ( ˆ ®nal 40 min of a 2 h clamp, see below) shown to be fully suppressed in old as well as plasma glucose and insulin measurements. Height was young subjects.18,19 Therefore, total glucose disposal measured to the nearest centimetre, weight to the was calculated from the exogenous glucose infusion nearest kilogram. Body mass index (BMI) was calcu- rate during the last 40 min of the 2 h clamp after lated as the weight divided by the square height. The correction for changes in glucose concentration in a waist-to-hip circumference ratio (WHR), which was total distribution volume of 250 ml=kg. Total glucose available in a subset of 89 subjects in ®ve centres, was disposal was normalised per kg of LBM. Energy determined by measuring the waist circumference at expenditure, respiratory quotient (RQ) and net sub- the narrowest part of the torso, and the hip circum- strate oxidation rates were calculated from gas ference in a horizontal plane at the level of the exchange measurements, as previously described.20 maximal extension of the buttocks. Resting energy expenditure (REE) and clamp energy Insulin action was measured in all subjects by the expenditure were taken to be the mean values during euglycaemic insulin clamp technique15 using an 40 min of baseline and the ®nal 40 min of the 2 h Insulin resistance and thermogenesis S Camastra et al 1309 clamp, respectively. Glucose-induced thermogenesis non-obese group. The fasting RQ did not differ between (GIT) was calculated as the difference between clamp groups. In the whole dataset, REE was inversely related energy expenditure and REE.11 Insulin-stimulated to age (P ˆ 0.001), insulin sensitivity (P ˆ 0.05), and glucose oxidation was calculated as the difference RQ (P ˆ 0.03), and positively related to all indices of between the total (last 40 min of the clamp) and body mass: BMI (r ˆ 0.63), LBM (r ˆ 0.81), fat mass fasting rate of glucose oxidation. Glucose storage (r ˆ 0.70), and percentage fat mass (r ˆ 0.58) (all was calculated by subtracting total glucose oxidation P < 0.0001). LBM alone explained 65% of the varia- from total glucose uptake. bility in REE. When REE was expressed as a function of For the purpose of the present analysis, obesity was height, weight and age, the following relationships were de®ned as the upper quartile of the BMI distribution obtained: REE (kcal=day) ˆ (1103 ‡ 32 BMI 7 4.5 (ie values  28.7 kg=m2). According to this cut-off age) in men, and (885 ‡ 32 BMI 7 4.5 age) in point, 238 subjects were non-obese and 84 were women. These formulas yield predictions quite similar obese. Insulin resistance was de®ned as the lowest to those of Harris ± Benedict equations, the only differ- decile of the distribution of total glucose uptake in ence being that in this dataset the slope of the regression non-obese subjects ( < 24 mmol=min= kgLBM); using of REE on age is virtually identical in men and women this de®nition, 254 subjects were insulin sensitive and (Figure 1). 68 were insulin resistant. When REE was expressed per kg of LBM and used Data are given as meanÆ SEM. When insulin in a multivariate model, male gender (P < 0.0001), sensitivity (ie total insulin-mediated glucose disposal increasing age (P < 0.003), and a higher RQ rate) was used as an independent variable in multiple (P ˆ 0.0004) Ð but not insulin sensitivity (P ˆ 0.45) regression analysis, steady-state plasma insulin con- nor BMI (P ˆ 0.82) Ð were independently associated centrations were also entered into the model to with a lower REE. In particular, the multiple regres- account for small between-group differences in sion equation predicted an average decline in REE of attained insulin plateaus (see Results). Plasma insulin 9% between the ages 18 and 80 y (Figure 2). Over the concentrations were transformed into their natural logarithms for use in statistical analyses to account for their skewed distribution. Two-way ANOVA, simple and multiple regression analyses were carried out by standard techniques. Paired group comparisons were carried out with the use of the Wilcoxon signed rank test. In all regression analyses, inter-centre varia- bility was accounted for by using dummy variables ((n 7 1) variables for n centres).

Results

The characteristics of the study groups are shown in Table 1. Obese and non-obese subjects were well Figure 1 Comparison of gender-speci®c age-dependence of matched for gender and age. Fat mass, as a percentage resting energy expenditure (expressed in the most common of body weight, was larger in women than in men and in units) as calculated from the current data (EGIR data) and by obese than non-obese subjects. REE was lower in Harris ± Benedict equations. Note that in the EGIR data resting thermogenesis declines with age in a parallel fashion in men and women than in men, and higher in the obese than the women.

Table 1 Characteristics of the study groupsa

Lean Obese

Men Women Men Women P b

n 165 73 48 36 Age (y) 37Æ 141Æ 238Æ 239Æ 2ns BMI (kg=m2) 24.1Æ 0.2 23.8Æ 0.3 33.3Æ 0.7 33.1Æ 0.7 A Fat mass (%) 27Æ 0.3 32Æ 0.4 38Æ 0.6 42Æ 0.6 A, B WHR (cm=cm) 0.88Æ 0.01 0.86Æ 0.03 0.88Æ 0.02 0.94Æ 0.03 C REE (kJ=min) 4.93Æ 0.06 4.21Æ 0.09 5.91Æ 0.16 5.33Æ 0.18 A, B REE (J=min=kgLBM) 91Æ 13 97Æ 12 95Æ 13 97Æ 12 B RQ 0.82Æ 0.01 0.81Æ 0.1 0.82Æ 0.01 0.81Æ 0.01 ns

aValues are meanÆ 1 SEM; BMI ˆ body mass index; WHR ˆ waist-to-hip ratio; REE ˆ resting energy expenditure; RQ ˆ respiratory quotient. bP values for the effect of obesity and gender by two-way analysis of variance: A ˆP < 0.01 or less for the effect of obesity; B ˆ P < 0.05 or less for the effect of gender; C ˆP < 0.05 or less for the interaction between obesity and gender. Insulin resistance and thermogenesis S Camastra et al 1310 observed range (0.6 ± 1.0) of RQ, the mean decrease in wise, GIT was 283Æ 107 J=min (P < 0.01) in the REE was 16 J=min=kgLBM. insulin sensitive obese individuals, and 7 12Æ 74 Total insulin-mediated glucose disposal (insulin sen- J=min (P ˆ ns) in the insulin resistant obese group. sitivity) was reduced to a similar extent in insulin Two-way ANOVA con®rmed that GIT was signi®- resistant as compared with insulin sensitive subjects, cantly associated with insulin resistance (P ˆ 0.0007), whether they were obese or not (Table 2). When total whether normalised by LBM or not, but not with glucose disposal was partitioned into total oxidation and obesity (P ˆ ns). As a fraction of REE, GIT averaged storage, both these components, as well as insulin 4.1Æ 0.6% (P < 0.0001 vs zero) in insulin sensitive stimulated glucose oxidation, were signi®cantly subjects, and 0.2Æ 1.1% (P ˆ ns vs zero) in insulin reduced in insulin resistant as compared to insulin resistant subjects (Figure 3). REE and GIT were sensitive subjects, regardless of obesity. In contrast, reciprocally related to one another (P < 0.0001 after insulin inhibition of fasting lipid oxidation was a sig- controlling for gender, age and BMI). ni®cant function of both insulin resistance and obesity. In univariate association in the whole dataset, GIT By also controlling for age, sex and steady-state plasma was directly related to insulin sensitivity (P < 0.0001), insulin concentrations in a multiple regression model, but not BMI (P ˆ 0.35), age (P ˆ 0.07) or gender all three insulin effects Ð stimulation of glucose sto- (P ˆ 0.12). By multiple regression analysis, GIT rage, stimulation of glucose oxidation and suppression remained linearly related to insulin sensitivity of lipid oxidation Ð were directly related to insulin (P < 0.0001) after adjusting for age, sex, BMI and sensitivity (with partial rs of 0.86 (P < 0.0001), 0.30 steady-state plasma insulin level, with a predicted (P < 0.0001), and 0.10 (P ˆ 0.08), respectively), change of 1.3 J=min=kgLBM (or 70 J=min on average) whereas only insulin-inhibited lipid oxidation was also for each 10 mmol= min=kgLBM change in insulin independently related to BMI (partial r ˆ 0.15, sensitivity (Figure 4). In individual regressions adjust- P < 0.01). Introducing the fat mass into the latter ing for gender, age and BMI, both of the main model removed the effect of obesity. components of insulin-stimulated glucose utilisa- In non-obese subjects, GIT was 173Æ 28 J=min tion Ð glucose oxidation (partial (r ˆ 0.17, P ˆ (P < 0.0001 vs zero) if insulin sensitive, and 0.003) and glycogen synthesis (partial r ˆ 0.20, 7 35Æ 108 J=min (P ˆ ns) if insulin resistant. Like- P ˆ 0.0006) Ð as well as insulin-inhibited lipid oxida-

Figure 2 True age-dependence of resting energy expenditure reconstructed (by multivariate analysis) from the EGIR dataset. The regression line is calculated at mean population values of Figure 3 Glucose-induced thermogenesis (as a percentage of BMI ( ˆ 26.4 kg=m2), gender ( ˆ 33% women), insulin sensitivity resting thermogenesis) strati®ed by obesity and insulin resis- ( ˆ 46 mmol.min71.kgLBM71) and RQ ( ˆ 0.81). tance.

Table 2 Insulin-mediated glucose and lipid metabolism according to obesity and insulin resistance

Lean Obese

Sensitive Resistant Sensitive Resistant P*

n 214 24 40 44 Steady-state plasma insulin (pmol=l) 465Æ 9 538Æ 38 548Æ 21 644Æ 36 A, B Total glucose uptake (mmol=min=kgLBM) 53.3Æ 1.0 23.1Æ 1.2 46.5Æ 1.7 24.4Æ 1.0 B Total glucose oxidation (mmol=min=kgLBM) 21.8Æ 0.5 13.0Æ 1.4 20.1Æ 1.4 15.8Æ 0.9 B Glucose storage (mmol=min=kgLBM) 31.8Æ 1.0 9.9Æ 1.6 27.2Æ 1.7 8.5Æ 1.3 B Insulin-stimulated glucose oxidation (mmol=min=kgLBM) 8.2Æ 0.3 4.3Æ 0.8 7.1Æ 0.9 4.3Æ 0.6 B Insulin-inhibited lipid oxidation (mmol=min=kgLBM) 7 1.9Æ 0.1 7 0.9Æ 0.3 7 1.2Æ 0.2 7 1.0Æ 0.1 C

P values for the effect of obesity and gender by two-way analysis of variance: A ˆP < 0.001 for the effect of obesity; B ˆ P < 0.0001 for the effect of insulin resistance; C ˆP < 0.04 for the interaction between obesity and insulin resistance. Insulin resistance and thermogenesis S Camastra et al 1311 tion (partial r ˆ 7 0.13, P ˆ 0.02) were related to WHR exerts a more adverse effect in women than in GIT. In a multivariate model simultaneously account- men (Figure 5). ing for all three insulin effects, glucose storage, insulin-stimulated glucose oxidation, and insulin- Discussion inhibited lipid oxidation were all independently cor- related with GIT (P < 0.001 for each). In the subset of subjects (n ˆ 89) in whom WHR The present study group included non-diabetic subjects measurements were available, GIT was inversely of either sex (34% women) with a wide range of ages associated with WHR (P < 0.001 after adjustment by (18 ± 80 y) and BMI (18 ± 50 kg=m2). Though not a gender, age, BMI, insulin sensitivity and steady-state random population sample, the EGIR cohort reproduces plasma insulin concentration). In this model, a sig- the salient characteristics of a general population of non- ni®cant interaction term between WHR and gender diabetic subjects.13,21 The present subset of the EGIR indicated that, all other factors being equal, a high cohort, for which thermogenesis and insulin sensitivity measurements were available, does not differ from the whole cohort in the distribution of any of the measured variables (data not shown), and is the largest collection of such data so far in the literature. Our values for REE, and the gender difference in REE, are in line with previous large-scale measure- ments.22,23 Also in accord with expectation is the strong relation of REE to lean body mass,24 a differ- ence in LBM of 10 kg predicting a change in REE of 1 kJ=min throughout the three-fold range of REE values. Finally, our data demonstrate that REE was lower in older as compared with younger subjects indepen- dently of LBM. It is interesting to extrapolate these cross-sectional age data (Figure 2) to the aging pro- cess. In the whole EGIR cohort (n ˆ 1200), mean body weight was 79.9 and 73.4 kg in the third and ®rst age quartile, respectively (at mean ages of 47 and 24 y). If REE declined linearly with time from the predicted values of 95 J=min=kgLBM in the ®rst age quartile to Figure 4 Glucose-induced thermogenesis as a function of per- centiles (pct) of insulin sensitivity in the whole dataset (adjusted 92.3 J=min=kgLBM in the third quartile, a constant by gender, age, BMI, and steady-state plasma insulin levels). caloric intake over this time period would lead to the Bars are Æ1 SEM. The point estimates are calculated at mean accumulation of 845 MJ of energy, corresponding to population values of BMI ( ˆ 26.4 kg=m2), gender ( ˆ 33% women) and steady-state plasma insulin concentrations 22 kg of fat (1 kg ˆ 37.7 MJ). Actual LBM, however, ( ˆ 499 pmol=l). averaged 51.8 kg in the ®rst age quartile and 53.5 kg in the third quartile. Assuming a time-liner change also for LBM, a constant caloric intake would lead to a cumulative energy balance of 7 117 MJ, ie a loss of 3 kg of fat. As fat mass instead was up 4.8 kg in the third as compared to the ®rst quartile, the true excess of energy intake over expenditure must have amounted to 297 MJ (or 35 kJ=day for 23 y, stored as fat). These calculations serve to emphasise that the decline in thermogenesis inherent in aging per se, though small percentage-wise, is a powerful risk factor for weight gain, and that the increase in LBM occurring between the 3rd and 5th decades of age provides a highly effective protection against weight gain. The clinical implication of these results is that obesity limits itself through its attendant increase in LBM. Physical exer- cise, by building up muscle mass, has the potential to counter the age-related decline in thermogenesis. In the case exempli®ed above, an increase in muscle mass Figure 5 Independent impact of a high waist-to-hip ratio (WHR) on glucose-induced thermogenesis. Data are point estimates of only 0.75 kg would prevent any change in fat mass from 89 cases adjusted by gender, age, BMI, insulin sensitivity, as the individual ages 23 y. and steady-state plasma insulin levels at mean group values of Obese subjects had higher mean REE values than gender ( ˆ 25% women), BMI ( ˆ 27.8 kg=m2), insulin sensitivity ( ˆ 41 mmol.min71.kgLBM71) and steady-state plasma insulin non-obese subjects; this difference was cancelled concentrations ( ˆ 505 pmol=l). when LBM was factored in.25,26 Of more interest is Insulin resistance and thermogenesis S Camastra et al 1312 the ®nding that REE, when adjusted by LBM, was not vation, that the prevalence of insulin resistance in related to insulin sensitivity. Thus, within the same obesity is lower than previously thought.21 unit mass of metabolically active lean tissue, resting GIT was controlled by a different set of factors as thermogenesis and response of glucose uptake to compared to REE, despite the fact that a higher REE insulin are independent phenomena. However, REE was generally associated with a lower GIT. Thus, in was consistently higher in subjects with a lower each unit of LBM and for the same plasma insulin fasting RQ, in men as well as in women irrespective level, GIT was similar in men and women across a of age and BMI. This ®nding is not explained by body wide range of ages and BMI. The only signi®cant composition (as the association was adjusted by modulator was insulin sensitivity, both as a category LBM), but may be related to the quality of lean (Table 2) and as a continuous variable (Figure 4): a tissues. In fact, a relative predominance of slow- robust thermogenic response to infused glucose twitch, oxidative (type I) over fast-twitch, glycolytic depends on the body's ability to dispose of the glucose. (type IIB) ®bres in skeletal muscle (which alone This ®nding con®rms the prediction put forward by represents  40% of body weight) is associated with Ravussin et al,32 who showed Ð in euglycaemic clamp a higher oxidative capacity.27 This, in turn, results in a experiments carried out in nine obese, insulin resistant shift in substrate oxidation towards fat at the cost of a volunteers Ð that GIT could be normalised by forcing higher resting oxygen consumption (whence a higher normal rates of total glucose disposal (using two-fold REE). Muscle ®bre composition is related to the level higher exogenous insulin infusion rates). The result is of aerobic ®tness. An acute bout of physical exercise also coherent with the conclusion of Segal et al,12 who (1 h cycling at 100 W) lowers postprandial RQ,10 and found no major effect of obesity on GIT in groups of regular aerobic training lowers resting RQ.28 The young men matched for degree of insulin sensitivity. clinical implications of the observed association In the latter study, meal-induced thermogenesis was between REE and RQ emerge from prospective stu- still impaired in obese non-diabetic subjects who had a dies of weight changes in non-diabetic humans. In normal theremogenic response to infused glucose. Pima Indians, a lower REE was found to be a Food ingestion elicits a marked and sustained increase signi®cant independent predictor of subsequent in splanchnic blood ¯ow,33 which favours heat leakage weight gain.29 More recently, in a Caucasian popula- across the abdominal wall.34 In their elegant study, tion a higher RQ has been found to predict a greater Brundin and coworkers34 showed that in obese sub- weight accumulation over a period of 3 y.30 jects the postprandial increase in energy expenditure During the clamp, insulin action on total glucose was reduced at the same time as the amount of heat metabolism ( ˆ insulin sensitivity) resulted from the drained into the hepatic vein was increased. Further- simultaneous stimulation of glucose storage and glu- more, when abdominal heat dispersion was prevented cose oxidation. The latter, in turn, was an inverse in lean subjects by arti®cial thermal insulation, post- function of the inhibition of resting lipid oxidation prandial thermogenesis was signi®cantly impaired. (P < 0.0001), as predicated by substrate competi- Thus, in obese individuals who exhibit a normal tion.31 Remarkably, whereas insulin's effects on glu- thermogenic response to infused glucose, post-meal cose storage and oxidation were not in¯uenced by thermogenesis can still be inhibited due to an excess of obesity, the degree to which euglycaemic hyperinsu- abdominal fat insulation. This mechanism also pro- linaemia was able to suppress basal lipid oxidation vides a plausible explanation for the current ®nding was signi®cantly impaired in the obese group inde- that a higher waist-to-hip ratio is associated with an pendently of insulin sensitivity (Table 2). This result impaired GIT at any level of insulin sensitivity, parti- held true after adjusting for LBM and steady-state cularly in women (Figure 5). While the hormonal plasma insulin levels (as well as other confounders). background of central obesity35 may itself play a role The likely explanation is that comparable hyperinsu- in reducing GIT, our results support the view that linaemia (ie steady-state clamp values), although insulin resistance of glucose metabolism and abdom- equally effective in suppressing lipolysis in each inal thermal insulation are intrinsic regulators of ther- unit of fat mass in lean and obese subjects, could mogenesis. In type 2 diabetes and other insulin not normalise total lipid oxidation in the obese resistant states (eg essential hypertension), abdominal because of their expanded total fat mass. In fact, obesity is frequent.36 In these phenotypes, therefore, when fat mass was introduced as a covariate, the the combination of insulin resistance and central fat difference in insulin-induced suppression of lipid accumulation is the likely physiological basis for the oxidation between obese and non-obese subjects was reduced thermogenesis and proneness to weight gain, cancelled. One consequence of this ®nding is that, if and, conversely, for the relative resistance to weight in insulin sensitive obese individuals lipid oxidation loss in response to calorie restriction.37 was fully responsive to insulin inhibition, insulin- A ®nal issue is the relationship of GIT to the stimulated glucose oxidation would be even higher various components of insulin action. Glycogen (due to reduced substrate competition). This implies synthesis was the strongest correlate of GIT in the that some obese individuals would be super-sensitive whole study group. This ®nding has been interpreted to insulin if they lost their excess fat mass. This as evidence that GIT is driven by the energy require- conclusion is in accordance with our previous obser- ment of glycogen synthesis.11 In our data, insulin- Insulin resistance and thermogenesis S Camastra et al 1313 mediated enhancement of glucose oxidation and 17 Lukaski HC. Methods for the assessment of human body inhibition of lipid oxidation also were independently composition: traditional and new. Am J Clin Nutr 1987; 46: 537 ± 556. related to GIT. This is easily explained by considering 18 De Fronzo R. Glucose intolerance and aging: evidence for that the excess energy liberated in the balance tissue insensitivity to insulin. Diabetes 1979; 28: 1095 ± 1101. between augmented glucose oxidation and reduced 19 Robert JJ, Cummins JC, Wolfe RR, Durkot M, Matthews DE, lipid oxidation is what cells use to ®nance the energy Zhao XH, Bier DM, Young VR. Quantitative aspects of cost of glycogen storage. Thus, it is the whole insulin- glucose production and metabolism in healthy elderly subjects. Diabetes 1982; 31: 203 ± 211. induced shift of intracellular substrate ¯uxes, rather 20 Ferrannini E. The theoretical bases of indirect calorimetry: a than a single component pathway, that eventually review. Metabolism 1988; 37: 287 ± 301. determines the extent of GIT. 21 Ferrannini E, Natali A, Bell P, Cavallo-Perin P, Lalic N, In summary, this cross-sectional study in Caucasians Mingrone G and European Group for the Study of Insulin con®rms and integrates information obtained in differ- Resistance (EGIR). Insulin resistance and hypersecretion in obesity. J Clin Invest. 1997; 100: 1166 ± 1173. ent ethnic groups (eg the Pima Indians of Arizona), 22 Harris JA, Benedict FG. A biometric study of basal metabo- suggesting that the basic control mechanisms of thermo- lism in man. Washington, DC: The Carnegie Institute of genesis are independent of race. In addition, the current Washington, 1919 publ No. 279. results are fully predictive of the changes in thermogen- 23 James WPT, Davies HL, Bailes J, Dauncey MJ. Elevated esis that can be expected with aging or weight gain as metabolic rates in obesity. Lancet 1978; 1: 1122 ± 1125. 24 Ravussin E, Lillioja S, Anderson TE, Christin L, Bogardus C. determined by long-term, prospective studies. Determinants of 24-hour energy expenditure in man. Methods and results using a respiratory chamber. J Clin Invest 1986; 78: References 1568 ± 1578. 1 Pittet Ph, Chappuis Ph, Acheson K, De Techtermann F, 25 Ravussin E, Burnand B, Schutz Y, JeÂquier E. Twenty-four- JeÂquier E. Thermic effect of glucose in obese subjects study hour energy expenditure and resting metabolic rate in obese, direct and indirect calorimetry. Br J Nutr 1976; 35: 281 ± 292. moderately obese, and control subjects. Am J Clin Nutr 1982; 2 Golay A, Schutz Y, Beyer HU, ThieÂbaud D, Cuchod B, 35: 566 ± 573. Maeder E, Felber JP, JeÂquier E. Glucose-induced thermogen- 26 Segal KR, Lacayanga I, Dunaif A, Gutin B, Pi-Sunyer FX. esis in non diabetic and diabetic obese subjects. Diabetes Impact of body mass and percent fat on metabolic rate and 1982; 31: 1023 ± 1028. thermogenesis in men. Am J Physiol 1989; 256: E573 ± 3 Bessard T, Schutz Y, JeÂquier E. Energy expenditure and E579. postprandial thermogenesis in obese women before and after 27 Krotkiewski M, Bylund-Fallenius AC, Holm J, Bjorntorp P, weight loss. Am J Clin Nutr 1983; 38: 680 ± 693. Grimby G, Mandroukas K. Relationship between muscle 4 Kaplan ml, Leveille GA. Calorigenic response in obese and morphology and metabolism in obese women: the effects nonobese women. M J Clin Nutr 1976; 23: 1108 ± 1113. of long-term physical training. Eur J Clin Invest 1983; 13: 5 Schwartz RS, Halter JB, Bierman LB. Reduced thermic effect 5 ± 12. of feeding in obesity: role of Norepinefrine. Metabolism 1983; 28 Keim NL, Barbieri TF, Van Loan MD, Anderson BL. Energy 32: 114 ± 117. expenditure and physical performance in overweight women: 6 Schetty PS, Jung RT, James WPT, Barrand MA & Callingham response to training with and without caloric restriction. BA. Postprandial thermogenesis in obesity. Clin Sci 1981; 60: Metabolism 1990; 39: 651 ± 658. 519 ± 525. 29 Ravussin E, Lillioja S, Knowler WC, Christin L, Freymond D, 7 Felig P, Cunningham J, Levitt M, Hendler R, Nadel E. Energy Abbot WGH, Boyce V, Howard BV, Bogardus C. Reduced expenditure in obesity in fasting and postprandial state. Am J rate of energy expenditure as a risk factor for body weight Physiol 1983; 244: E45 ± E51. gain. New Engl J Med 1988; 318: 467 ± 472. 8 Sharief NN, McDonald I. Differences in dietary-induced 30 Marra M, Scal® L, Covino A, Esposito-DelPuente A, Contaldo thermogenesis with various carbohydrates in normal and over- F. Fasting respiratory quotient as a predictor of weight changes weight men. Am J Clin Nutr 1982; 35: 267 ± 272. in non-obese women. Int J Obes 1998; 22: 601 ± 603. 9 Welle SL, Campbell RG. Normal thermic effect of glucose in 31 Randle PJ, Garland PB, Hales CN, Newsholme EA. The glu- obese women. Am J Clin Nutr 1983; 37: 87 ± 92. cose fatty acid cycle. Its role in insulin sensitivity and the 10 Segal KR, Gutin B, Nyman AM, Pi-Sunyer FX. Thermic metabolic disturbances of diabetes mellitus. Lancet 1963; i: effect of food at rest, during exercise, and after exercise in 785 ± 789. lean and obese men of similar body weight. J Clin Invest 1985; 32 Ravussin E, Kevin JA, Vernet O, Danforth E, JeÂquier E. 76: 1107 ± 1112. Evidence that insulin resistance is responsible for the 11 Ravussin E, Bogardus C, Schwartz RS, Robbins DC, Wolfe R, decreased thermic effect of glucose in human obesity. J Clin Horton ES, Danforth E, Sims EAH. Thermic effect of infuse Invest 1985; 76: 1268 ± 1273. glucose and insulin in man. Decreased response with increased 33 Brundin T, Branstromand R, Wahren J. Effects of oral vs i.v. insulin resistance in obesity and noninsulin-dependent diabetes glucose administration on splanchnic and extrasplanchnic mellitus. J Clin Invest 1983; 72: 893 ± 902. O2 upake and blood ¯ow. Am J Physiol 1996; 271: E496 ± 12 Segal KR, Albu J, Chun A, Edano A, Legaspi B, Pi-Sunyer FX. E504. Independent effects of obesity and insulin-resistance on post- 34 Brundin T, Thorne A, Wahren J. Heat leakage across the prandialthermogenesisinmen.JClinInvest1992; 89: 824 ± 833. abdominal wall and meal-induced thermogenesis in normal- 13 Ferrannini E. Vichi S, Beck-Nielsen H, Laakso M, Paolisso G, weight and obese subjects. Metabolism 1992; 41: 49 ± 55. Smith U and European Group for the Study of Insulin Resistasnce 35 Rosmond R, Dallman MF, and Bjontorp P. Stress-related (EGIR). Insulin action and age. Diabetes 1996; 45: 947 ± 953. cortisol secretion in men: relationship with abdominal obesity 14 WHO Expert Committee on Diabetes Mellitus. econd report. and endocrine, metabolic and hemodynamic abnormalities. World Health Organization: Geneva, 1980. J Clin Endocrinol Metab 1998; 83: 1853 ± 1859. 15 De Fronzo RA, Tobin JD, Andres R. Glucose clamp techni- 36 Kissebah AH, Krakower GR. Regional adiposity and morbid- que: a method for quantifying insulin secretion and resistance. ity. Physiol Rev 1994; 74: 761 ± 811. Am J Physiol 1979; 237: E214 ± E223.16 Hume R. Prediction 37 Tappy L, Felber JP, JeÂquier E. Energy and substrate metabo- of lean body mass from height and weight. J Clin Path 1966; lism in obesity and postobese state. Diabetes Care 1991; 14: 19: 389 ± 391. 1180 ± 1188.