An impaired fat oxidation has been implicated to play a role in the etiology of obesity, but it is unclear to what extent impaired fat mobilization from adipose tissue or oxidation of fat is responsible. The present study aimed to examine fat mobilization from adipose tissue and whole body fat oxidation stimulated by exercise in seven formerly obese women (FO) and eight matched controls (C). Lipolysis in the periumbilical subcutaneous adipose tissue, whole body energy expenditure (EE), and substrate oxidation rates were measured before, during, and after a 60-min bicycle exercise bout of moderate intensity. Lipolysis was assessed by glycerol release using microdialysis and blood flow measurement by 133Xe clearance technique. The FO women had lower resting EE than C (3.77 ± 1.01 vs. 4.88 ± 0.74 kJ/min,P < 0.05) but responded similarly to exercise. Adipose tissue glycerol release was twice as high in FO than in C at rest (0.455 ± 0.299 vs. 0.206 ± 0.102 μmol ⋅ 100 g−1 ⋅ min−1,P < 0.05) but increased similarly in FO and C in response to exercise. Despite higher plasma nonesterified fatty acids (NEFA) in FO (P < 0.001), fat oxidation rates during rest and recovery were lower in FO than in C (1.32 ± 0.84 vs. 3.70 ± 0.57 kJ/min,P < 0.02) and fat oxidation for a given plasma NEFA concentration was lower at rest (P < 0.001) and during exercise (P = 0.01) in the formerly obese group. In conclusion, fat mobilization both at rest and during exercise is intact in FO, whereas fat oxidation is subnormal despite higher circulation NEFA levels. The lower resting EE and the failure to use fat as fuel contribute to a positive fat balance and weight gain in FO subjects.
- adipose tissue
- energy expenditure
- respiratory quotient
- substrate oxidations
obesity develops due to an interaction between genetic components and certain environmental factors such as a high-fat diet. A recent prospective study by Heitmann et al. (16) showed that only subjects with a family history of obesity gained weight when consuming high-fat diets. Among the genetically determined factors are a low resting energy expenditure (8, 23) and a low ability to oxidize fat (30), which both have been shown in prospective studies to be risk factors for weight gain and obesity. Compared with never obese subjects, formerly obese (FO) subjects have been reported to have lower fat/carbohydrate oxidation in the fasting state (3), postprandially (22), and on a 24-h basis (3). Furthermore, our group has reported that FO subjects failed to increase fat oxidation appropriately when the dietary fat content is increased from medium to high (4). The impaired fat oxidation seems in particular to be expressed postprandially after intake of high-fat meals (22). However, these studies have not examined whether the lower fat oxidation of FO is caused by a defective fat mobilization or by a reduced tissue capacity to use fat substrates.
The aim of the present study was therefore to investigate whether the lower fat oxidation in FO subjects is related to an impaired fat mobilization or fat oxidation using moderate exercise to stimulate lipolysis and fat utilization.
SUBJECTS AND METHODS
Seven FO and eight control (C) subjects participated in the study. The body weight of all FO subjects had previously exceeded 120% of their ideal body weight (Metropolitan Life Insurance Tables 1983). The excessive weight was lost by a conventional dietary energy restriction program, and body weight was kept stable within 110% of ideal body weight for at least 2 mo before the study. Individual weight loss was 15–20 kg. Subjects of the control group were selected to match the FO with regard to the characteristics given in Table 1. Height was measured to the nearest 0.5 cm and body weight to the nearest 100 g, with subjects wearing light clothes. Measurements of circumferences were taken to the nearest 0.5 cm. Body composition was estimated by the bioimpedance method using an Animeter (HTS-Engineering, Odense, Denmark). Fat-free mass (FFM) and fat mass (FM) were calculated by the equations developed from a study using the four-compartment model on a cohort of the Danish population reported by Heitmann (15). None of the subjects took any medicine, except for one FO who took contraceptives. Two FO and two C subjects were light cigarette smokers. The study protocol was approved by the Municipal Ethical Committee of Copenhagen and Frederiksberg, and all subjects gave informed consent.
All subjects arrived at the laboratory at 8 AM, after an overnight fast. Smoking was not allowed in the morning. Subjects did not participate in any strenuous physical activity for the 3 days before the experimental day. They rested in the supine position, and, during local analgesia, a catheter was introduced percutaneously into a radial artery. In two FO subjects access to an artery could not be achieved, so an antebracheal vein was cannulated. The O2 saturation was satisfactory for only one of these subjects (0.91 ± 0.06), and the blood samples from the other subjects, except for the glycerol levels, were excluded. Adipose tissue microdialysis was performed as previously described in detail by Simonsen et al. (25). After in vivo calibration of the microdialysis probes (seeMicrodialysis), two resting measurements were performed at time t= −15 min and t = 0 min while the subjects sat quietly on a cycle ergometer. Fromt = 0 tot = 60 min the subjects exercised on an electrically braked bicycle ergometer at a workload corresponding to 50% of their maximal oxygen uptake (V˙o 2 max). During the experiment, microdialysis samples were collected in 20-min periods fromt = −20 min tot = 80 min. Arterial blood samples were taken every 15 min between t = −15 to t = 75 min, when the recovery phase started. Heart rate was continuously monitored by a portable heart rate monitor (Sport Tester PE 3000, Polar Electro, Kempele, Finland).
All subjects consumed a standardized diet at home delivered free of charge from the department for the 3 days preceding the experimental day. The energy requirements of the subjects were computed from equations in which FFM and FM are used to predict 24-h energy expenditure (EE) based on previous measurements in our respiration chambers (19) and multiplied by 1.12 to account for a higher free-living activity. The energy content of the standardized diet was calculated by Dankost dietary assessment software (National Food Agency, Denmark), and energy was provided from 30% fat, 55% carbohydrate, and 15% protein. Subjects were allowed to drink coffee and tea freely but not alcoholic beverages.
Separated by at least 1 wk,V˙o 2 max was measured twice by the Douglas bag method, as described by McArdle (21). Each test started with a warm-up period of at least 8 min on an electrically braked bicycle ergometer, followed by a sequential increase in workload by first 50 W and, after 2 min, by 30 W until exhaustion. Verbal encouragement was given to attain maximal performance. Heart rate was monitored, and a Douglas bag was filled at every exercise level. The highest V˙o 2 achieved was taken as V˙o 2 max if four of the five criteria ofV˙o 2 max were fulfilled: 1) leveling off,2) respiratory exchange ratio >1.00, 3) plasma lactate >8 mM,4) maximum heart rate 220 minus age, and 5) ventilatory coefficient (VE/V˙o 2) > 30. To estimate the work load corresponding to 50% of subjects’V˙o 2 max,V˙o 2 for each of the exercise levels during theV˙o 2 max test was plotted against power output.
EE and substrate oxidations were measured by indirect calorimetry. The subjects breathed through a low-resistance, scuba, one-way mouthpiece. After ∼10 min of adaptation, expiratory gas was collected in Douglas bags for 1–2 min. Expiratory gas was continuously analyzed for oxygen and carbon dioxide with a Godart Rapox Oxygenometer (Bilthoven, Holland) and a Beckman LB-1 Medical Gas Analyzer (Fullerton, CA). When end-expiratory carbon dioxide fraction was constant, collection of expiratory gas in Douglas bags was started. If respiratory steady state was not achieved during the collection or if respiratory quotient (RQ) exceeded 1.1, the measurement was not used. Volume and dry gasses were measured, and the results were converted to standard temperature and dry pressure. Oxygen was measured with an electrochemical oxygen sensor (Ametec Oxygen Analyzer, Pittsburg, PA), and carbon dioxide was measured by an infrared carbon dioxide sensor (Ametec Carbon Dioxide Analyzer). Calculations of EE and substrate oxidation rates were performed as previously described (7). Protein oxidation was assumed to be constant and amounting to 15% of EE. The error of calculating EE by omitting the exact correction from urinary nitrogen is negligible and impossible to estimate during such a short period of time. The reliability was assessed by the coefficient of variation on resting EE repeated at a 1-wk interval in eight subjects. The coefficient of variation was found to be 3% (7).
Subcutaneous adipose tissue intercellular glycerol concentration was performed with the microdialysis technique (2), which we previously have validated against fat vein catheterization (25). Briefly, a single 4-cm dialysis fiber was glued at both ends to a nylon tube with an outer diameter of 60 mm. Two sterile dialysis catheters (Gambro 18, 3,000 Da) were inserted parallel to each other, separated by 1 cm, into the periumbilical subcutaneous tissue using 18-gauge cannulas. The tubes were connected to a precision pump (CMA 100, Carnegie Medicin, Solna, Sweden) and perfused with a saline solution with a flow rate of 2.9 μl/min. Because exchange across the microdialysis membrane does not reach 100% equilibrium, in vivo calibration of each microdialysis fiber was performed by the no net flux technique, as described by Lönnroth et al. (20) and Simonsen et al. (25). The relative recovery of glycerol was found to be 43.4 ± 15.0% (means ± SD,n = 15). It may be as low as 5%, but such low values increase the error when the concentrations are corrected to 100%. The recovery of glycerol depends on the volume of interstitial fluid the dialysis membrane is exposed to. The recovery may vary considerably between individuals and fibers, and this is the background for calibration of each fiber in each individual before initiation of the experiment. By this procedure, the actual recovery is determined for each individual fiber in use (for details see Ref. 25). The dialysis fiber was considered fully functional when it delivered a constant flow, whereas a decrease or arrest of the flow was assumed to be due to clotting or hematoma. The relative recovery was measured in the same fibers twice separated by 24 h in 11 subjects. The relative recovery was 17 ± 6% and 19 ± 4% (B. Stallknecht and J. Bülow, unpublished results).
Adipose tissue blood flow.
Adipose tissue blood flow (ATBF) was measured by the133Xe-washout method. Briefly, 200–300 kBq of 133Xe dissolved in sterile saline were injected in the periumbilical subcutaneous adipose tissue between the two microdialysis fibers. A NaI (T1) scintillation detector was placed externally for registration of the 133Xe photopeak at 81 keV. Counts were accumulated during consecutive 50-s intervals and plotted on a semilogarithmic linear diagram as a function of time. The rate constant of the elimination (k) was calculated from the best-fitted straight lines during rest and exercise. ATBF was then calculated as: ATBF =k ⋅ λ ⋅ 100 (ml ⋅ 100 g−1 ⋅ min−1), where λ is the tissue-to-blood partition coefficient for133Xe at equilibrium. λ was estimated by measuring the local skinfold thickness near the umbilicus using a Harpender caliber [λ = 0.22 ⋅ local skinfold (mm) + 2.99] (11). The regional rate of glycerol release was estimated according to the Fick principle: release = (cv − ca) × Q, where cv and ca denote venous and arterial concentration and Q is blood flow. The conversion of interstitial glycerol concentration to venous plasma water concentration was performed as follows: cv = (ci − ca) × (1 −e −PS/Q) + ca, where ci denotes interstitial water concentration, PS the permeability surface area product, which was assumed to be 2 ml/(100 g ⋅ min) for glycerol (25), and Q the plasma water flow. The plasma water flow was calculated as: blood flow ⋅ [1 − hematocrit (Hct) ⋅ 10−2] ⋅ 0.94. Venous plasma concentration was converted to whole blood concentrations using the formula: cb ⋅ 0.93 ⋅ [1 − (0.3 ⋅ Hct)], where cb denotes blood concentration (2).
The blood samples were centrifuged for 10 min at 3,000g and 4°C. Plasma glucose and lactate concentrations were determined using a YSI 2300 Glucose/lactate analyzer (Yellow Springs Instruments, Yellow Springs, OH). Glycerol in plasma and dialysate and triglycerides were determined as described elsewhere (7). Nonesterified free fatty acids (NEFA) were extracted immediately after blood samples had been centrifuged and were determined later (7). Plasma insulin concentration was measured using radioimmunoassay kits purchased from Novo-Nordisk, Copenhagen, Denmark. For determination of catecholamines, blood was collected in tubes containing reduced glutathione and ethylene glycol-bis (β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid. The samples were centrifuged, and the plasma was stored at −80°C until analysis (3).
Results are presented as means ± SD. All group comparisons were made by one- and two-way analysis of variance (ANOVA) for repeated measurements or by paired t-test. Logarithmic transformation was performed if data were not normally distributed. Statistical analyses were performed with Statgraphics software version 4.2 (Graphic Software Systems, Rockville, MD) and Sigmastat program (Jandel Scientific, Erkrath, Germany). Values were considered statistically significant whenP < 0.05.
Physical and anthropometrical characteristics.
The physical and anthropometrical characteristics of the seven FO women and the eight matched C subjects are given in Table 1. There were no significant differences between the two groups for any of the measurements.
Table 2 presents EE in the resting state (REE) and during exercise and recovery for FO and C. REE was significantly lower for FO compared with C (P = 0.047). REE did not correlate significantly with FFM (r = 0.54, P = 0.06), which could be attributed to the low variability in FFM among the subjects (REE = 4%). A positive correlation existed between REE andV˙o 2 max(r = 0.62,P = 0.02). Most of the difference in REE between FO and C could be accounted for by differences inV˙o 2 max, and after adjustment for differences in both FFM andV˙o 2 max the group difference was attenuated (4.07 ± 0.65 vs. 4.64 ± 0.81 kJ, FO vs. C; P = 0.19). The targeted exercise load was 50% of each subject’sV˙o 2 max, and the actual exercise intensity did not differ between FO and C (46.8 ± 5.01 vs. 47.9 ± 3.44%). EE during exercise and recovery was not significantly different between FO and C subjects (Table 2).
RQ and substrate oxidations.
Resting RQ was significantly higher in FO than in C (0.849 ± 0.07 vs. 0.742 ± 0.015; P < 0.005). Thus fat oxidation was significantly lower and carbohydrate oxidation significantly higher in FO than in C (Table 2). Fat oxidation expressed as percentage of EE was also lower at rest in FO than in C (Table 2 and Fig. 1). No correlation was found by linear regression analysis between oxidative fat energy percentage and FM for both groups taken together. However, a positive correlation was present among the C subjects (r = 0.82,P = 0.04). No group difference was seen in RQ or oxidation rates during exercise and recovery. Mean fat oxidation during exercise was 678.5 ± 138.9 and 763.3 ± 204.8 kJ/h for FO and C subjects, respectively. The mean fat oxidation (%) during exercise was 51.1 ± 8.1 vs. 51.6 ± 12.1% for FO and C, respectively. Carbohydrate oxidation rates are shown in Table 2.
No group difference was seen in heart rate during rest (FO vs. C 76.1 ± 15.8 vs. 77.9 ± 9.6 beats/min), exercise (136.9 ± 18.9 vs. 134.9 ± 8.8 beats/min), or recovery (84.0 ± 14.5 vs. 79.6 ± 9.3 beats/min).
Glucose and insulin concentrations.
There were no group differences in plasma glucose during rest, exercise, or recovery (Table 3). Plasma insulin concentrations were lower in FO compared with C both at rest and during exercise (group effect: P = 0.03), and the response to exercise differed significantly (group × time effect: P = 0.011; Table3).
Glycerol, NEFA, and triglyceride.
Plasma glycerol concentration increased similarly in both groups in response to exercise (P < 0.05) and remained significantly elevated above resting values during recovery (Fig. 2). Resting NEFA concentration was higher in FO (P < 0.02) and remained higher in FO during exercise (P < 0.002). The response of NEFA to exercise did not differ between FO and C (ANOVA). NEFA concentration during recovery was similar in the two groups (P > 0.05; Table 2, Fig.3). A significant correlation between NEFA and fat oxidation (%) was found at rest (r = 0.38,P < 0.05) and during recovery (r = 0.46,P = 0.02) but not during exercise. Fat oxidation (%) expressed in relation to plasma NEFA or adjusted for differences in plasma NEFA concentrations was lower in FO at rest by 69% (P < 0.001) and during exercise by 38% (P = 0.01) than in the C group. Plasma triglyceride concentrations were elevated during the exercise experiment and remained elevated above fasting values during recovery in both groups (data not shown).
Fasting epinephrine concentrations did not differ between FO and C (P = 0.08; Table 3), and the exercise-induced increases were similar in the two groups. Norepinephrine increased in both groups during exercise (P < 0.03) and decreased similarly at recovery.
ATBF and glycerol release.
Subcutaneous ATBF was higher at rest in FO (FO 2.46 ± 1.13 vs. C 1.36 ± 0.56 ml ⋅ 100 g−1 ⋅ min−1,P < 0.03; Table4). The subcutaneous ATBF did not increase during exercise but remained significantly higher in FO (2.32 ± 0.99 ml ⋅ 100 g−1 ⋅ min−1) than in C (1.39 ± 0.39 ml ⋅ 100 g−1 ⋅ min−1) (P < 0.03). Interstitial glycerol concentration was similar in the two groups at rest and during and after exercise (Table 4). The interstitial glycerol concentration increased significantly above resting values during exercise and remained elevated in the recovery period for both FO and C. As a result of the higher blood flow, a higher subcutaneous adipose tissue glycerol release was found in the FO than in C (0.46 ± 0.30 vs. 0.21 ± 0.10 μmol ⋅ 100 g−1 ⋅ min−1,P < 0.03; Table 4). Glycerol release increased significantly during exercise (P < 0.05) and remained elevated above resting values during recovery (P < 0.006) but did not differ between the groups during exercise (FO vs. C: 0.66 ± 0.36 vs. 0.40 ± 0.26 μmol ⋅ 100 g−1 ⋅ min−1,P = 0.20) or during recovery (FO vs. C: 0.80 ± 0.53 vs. 0.45 ± 0.42 μmol ⋅ 100 g−1 ⋅ min−1,P = 0.074).
The aim of this study was mainly to establish whether the previous finding of a lower fat oxidation in FO subjects (4, 9, 22, 28) was caused by impaired fat mobilization from adipose tissue or was due to impaired skeletal muscle utilization when moderate exercise was used to increase fat utilization and demand. The major findings were that fat mobilization from the periumbilical subcutaneous adipose tissue as assessed by glycerol release was not decreased in FO subjects during rest, exercise, or recovery compared with C subjects. This finding was supported by the observation that plasma glycerol concentrations increased similarly in the two groups in response to exercise. Despite higher circulating levels of NEFA, the FO had lower fat oxidation during rest and recovery and the difference in fat oxidation was more pronounced when we adjusted for differences in plasma NEFA concentrations. Therefore, the study suggests that the major cause of the lower fat oxidation in FO may be an impaired skeletal muscle uptake and/or oxidation of fatty acids. In addition, the FO had lower resting EE for a given FFM and also had lower insulin levels despite normoglycemia. A low REE, a low fat oxidation, and a high insulin sensitivity have been shown to be risk factors for weight gain and obesity, and in the present study the FO displayed all these traits.
Fat oxidation rates and lipolysis are influenced by FFM, FM, aerobic capacity, composition of the antecedent diet, and energy balance. To control for these confounding factors, the subjects were matched with regard to age, body weight, FFM, FM, and aerobic capacity (12) and they received a weight-maintenance diet with a fixed macronutrient composition for 3 days before the experiment. Although efforts were made to control for these factors influencing fat metabolism, we cannot entirely exclude the fact that differences between FO and C subjects in the macronutrient composition of their habitual diet may have influenced the findings. However, we have recently shown that, when the fat-to-carbohydrate ratio of the diet is markedly changed from one day to another, the oxidative autoregulation fully adjusts within 3 days (B. Buemann, S. Toubro, and A. Astrup, unpublished data). We therefore find it unlikely that the lower fat oxidation of the FO can be explained by differences in the habitual diet of the groups. Protein oxidation was not measured but was assumed to be similar in the two groups and constant throughout the experiment. The error of calculating EE by omitting the exact correction from urinary nitrogen is negligible. The error on substrate oxidations may be more important, but we find it unlikely that protein oxidation differs between the two groups because we had controlled the major determinants of protein oxidation. 1) The two groups were well matched for size of FFM, the major determinant of protein metabolism. Although the reliability of the bioimpedance method for this purpose may be questioned, by the use of dual-energy X-ray absorptiometry scanning we have confirmed in a subgroup of six of the FO and six C subjects that no difference in body composition was found (A. Raben, E. Mygind, B. Saltin, and A. Astrup, unpublished results). 2) The antecedent diet was controlled for 3 days with the energy content from protein fixed at 15%. Moreover, in previous studies of FO there has been no indication of an altered protein oxidation under steady-state conditions (3-5,22, 27).
Low REE, high NEFA, and low insulin concentrations are features of underfeeding and a negative energy balance. These traits were observed among the FO compared with the C subjects, and we have considered whether relative underfeeding of the FO may have taken place. We find this possibility unlikely because relative underfeeding means a negative energy balance, which inevitably decreases RQ. Consequently, in case of relative underfeeding, the RQ of the FO should have been lower than in the never obese group and not higher, such as we found.
Tobacco smoking is known to increase sympathoadrenal activity and influence fat metabolism, and smoking may be an important confounder in studies of energy and fat metabolism. However, in the present study both groups included only two light smokers, and they were not allowed to smoke on the experimental day. Moreover, when smoking was included as a covariate in the statistical analysis, the results essentially did not change.
The present study does not elucidate the cause of the lower fat oxidation among the FO subjects. Fat oxidation is inversely associated with the rate of glucose oxidation (24). Increased glucose oxidation may directly limit long-chain fatty acid oxidation by inhibiting the transport of fatty acids into the mitochondria (14). Therefore, the rate of glucose oxidation is largely determined by the delivery of glucose and by insulin sensitivity of skeletal muscle, whereas the rate of fat oxidation is determined by filling up the gap between oxidation of glucose and the oxidation required to cover total EE. Despite similar plasma glucose concentrations, the FO had lower insulin concentrations throughout the study. This suggests an increased insulin sensitivity of the FO. This is an interesting finding, which could not be explained by differences in FFM and aerobic capacity (Table 1). It is possible that the increased insulin sensitivity may contribute to the higher glucose oxidation and lower fat oxidation rates at rest and recovery among the FO. This is in line with the observation that a high insulin sensitivity is a risk factor for weight gain (26). Although studies in FO subjects have not been equivocal (27), Raben et al. (22) found that another group of FO women had a lower glucose and insulin response on three different diets compared with a control group. The present study therefore supports the fact that a high insulin sensitivity may contribute to a low fat oxidation and to a positive energy balance.
In the present study we investigated both fat utilization using indirect calorimetry and glycerol release from the abdominal subcutaneous adipose tissue by combining the microdialysis technique and the 133Xe clearance method. The assumptions when using these methods for estimating glycerol release are reviewed by Arner and Bülow (2), and the microdialysis technique has been validated against the fat vein catheterization technique by our group and has been shown to produce rather accurate estimates of glycerol and glucose balances across the adipose tissue depot (25). The existance of marked regional variation in ATBF is well established, but there are no indications of a difference between right and left abdominal sites (10). However, regional differences in lipid mobilization during exercise are also reported, and the abdominal subcutaneous fat is much more lipolytic than that of the gluteal-femoral region, but less than the intra-abdominal visceral fat (1, 13). We chose the subcutaneous abdominal fat not for its representativeness, but because it plays an important role for mobilization of fatty acids during exercise. However, it is a limitation of the study that it assessed lipolysis at only one site, and we cannot rule out the possibility that FO may have impaired lipolysis in other adipose tissues. The finding of a very similar exercise-induced increase in plasma glycerol concentrations in FO and C subjects does not support any important difference during exercise. The higher plasma NEFA concentrations during rest and exercise in the FO than in the C subjects may be interpreted as a result of a lower NEFA oxidation. Plasma NEFA concentrations increased immediately after exercise in the control group but not in FO group. A postexercise increase in NEFA has been reported by others (17-18,29) and is likely to reflect that, whereas the utilization of NEFA is rapidly reduced, lipolysis remains increased during recovery (Table 4).
REE was significantly lower in the FO subjects, which is in accordance with studies showing that a low REE is a risk factor for subsequent weight gain (23). The finding is also in agreement with the study of Astrup et al. (5), in which data of 28 FO subjects and 28 matched C subjects were compiled and an 8% lower REE was observed in the FO group. In a recent meta-analysis, REE was adjusted for FFM, and FM was 3% lower in a group of 124 FO subjects compared with a matched C group (6). In addition, a low REE was threefold more prevalent among FO than C subjects (6).
In conclusion, the present study shows that FO women had an intact ability to mobilize fat from the abdominal subcutaneous adipose tissue compared with well-matched C subjects during rest, exercise, or recovery. However, for a given plasma NEFA concentration fat oxidation was lower. In addition, REE for a given body size and composition was lower among the FO women. The study therefore points to an impaired uptake and utilization of fat in muscle as a mechanism that, in susceptible individuals, may contribute to a positive energy balance, weight gain, and obesity.
We thank the staff of the metabolic unit of the Research Department of the Human Nutrition: Tina Cuthbertson, Lis Kristoffersen, Inger-Lise Grøntfelt, Inge Timmermann, and Bente Knap. Furthermore, we thank Birgitte Kjærskov, Department of Clinical Physiology, Bispebjerg Hospital, and Karen Klausen, Institute of Medical Physiology, Panum Insititut, for providing technical assistance.
Address for reprint requests: A. Astrup, Research Dept. of Human Nutrition, RVAU, Rolighedsvej 30, DK-1958 Frederiksberg, Denmark. E-mail:
- Copyright © 1998 the American Physiological Society