In healthy young subjects, training increases insulin sensitivity but decreases the capacity to secrete insulin. We studied whether training changes β-cell function in type 2 diabetic patients. Patients, stratified into “moderate” and “low” secretors according to individual C-peptide responses to an intravenous glucagon test, were randomly assigned to a training program [ergometer cycling 30–40 min/day, including at least 20 min at 75% maximum oxygen consumption (V̇o2 max), 5 days/wk for 3 mo] or a sedentary schedule. Before and after the intervention (16 h after last training bout), a sequential hyperglycemic (90 min at 11, 18, and 25 mM) clamp was performed. An intravenous bolus of 5 g of arginine was given at the end. Training increased V̇o2 max 17 ± 13% and decreased heart rate during submaximal exercise (P < 0.05). During the 3 mo of sedentary lifestyle, insulin and C-peptide responses to the clamp procedures were unchanged in both moderate and low secretors. Likewise, no change in β-cell response was seen after training in the low secretors (n = 5). In contrast, moderate secretors (n = 9) showed significant increases in β-cell responses to 18 and 25 mM hyperglycemia and to arginine stimulation. Glucagon responses to arginine as well as measures of insulin sensitivity and Hb A1c levels were not altered by training. In conclusion, in type 2 diabetic patients, training may enhance β-cell function if the remaining secretory capacity is moderate but not if it is low. The improved β-cell function does not require changes in insulin sensitivity and Hb A1c concentration.
- hyperglycemic clamp
an inverse relationship between the level of daily physical activity and the incidence of type 2 diabetes is now well established by the results from several prospective epidemiological studies (12, 15, 22, 23, 38). Likewise, intervention studies have shown that increased physical activity can prevent or delay type 2 diabetes in people at risk (8, 30, 36). For patients with newly diagnosed type 2 diabetes, exercise and diet are the first choices of treatment. Previous studies on the effect of physical training in the treatment of type 2 diabetes have focused on the effects on insulin action, and most (1, 5, 14, 20, 28, 29, 35, 39) but not all (21, 31, 34) have shown an increase in insulin action with physical training.
In contrast, the effect of physical training on β-cell function in patients with type 2 diabetes has not drawn much attention. In healthy humans, as judged both from dose-response characterization of the relationship between plasma glucose and proinsulin and C-peptide concentrations and from β-cell stimulation with other secretagogues, during training insulin secretory capacity decreases along with the increase in insulin action (6, 16, 18, 19, 25, 27). A similar relationship would appear less expedient in patients with type 2 diabetes. However, it has been proposed that β-cells in type 2 diabetic patients and in healthy subjects, respectively, respond in opposite directions to training. In type 2 diabetes patients, a diminished glucose stress resulting from an increase in insulin sensitivity of target tissues might improve the secretory capacity of overloaded β-cells (11, 32). Some indirect evidence supports this view. After training, enhanced insulin or C-peptide responses to oral or intravenous glucose tolerance testing have been found in some studies of patients with type 2 diabetes (1, 20, 31). Different responses between subgroups within the heterogenous group of type 2 diabetic patients would explain why not all studies using glucose tolerance tests have indicated that training alters β-cell function (33, 35).
Indeed, the heterogeneity of the population of patients with type 2 diabetes has not always been fully acknowledged in studies dealing with lifestyle modifications as a tool to improve glucose homeostasis. As regards effects on insulin action, this fact may not be of major importance, since the insulin resistance of type 2 diabetes correlates almost linearly with the fasting glucose concentration, i.e., with the severity of the disease (3). However, when effects on insulin secretion are studied over time, it is important to realize that insulin secretion (basal and stimulated) correlates with fasting glucose in a bell-shaped manner (3). Hyperinsulinemia develops gradually with time in response to increasing glucose stimulation, but at some point the maximally achievable secretory capacity of the β-cells is reached, and at higher basal glucose levels insulin secretion decreases to normal or subnormal levels. Thus it may be expected that, depending on the residual insulin secretory capacity of the β-cells, the effect of physical training on β-cell function may vary greatly among patients with type 2 diabetes, and the response to training may not necessarily follow the pattern seen in normoglycemic nondiabetic people. It has yet to be established whether the effect of training on the β-cell response to secretagogues in type 2 diabetic patients depends on the preexercise secretory capacity and whether it may be qualitatively different from the changes seen in healthy young people.
Accordingly, to evaluate the hypothesis that in type 2 diabetics the effect of physical training on β-cell function depends on the residual secretory capacity, we have now compared the effect of 3 mo of endurance training on insulin secretion in two groups of type 2 diabetic patients with moderate and low, respectively, insulin secretory capacity. One-half of the recruited diabetics initially remained sedentary, and all patients were studied both before and after the 3-mo intervention period.
MATERIALS AND METHODS
Type 2 diabetic male patients were recruited from outpatient diabetic clinics. All subjects gave their informed consent to participate in the study, which was approved by the Ethics Committee for Copenhagen and Frederiksberg. The patients did not suffer from diseases other than diabetes and had no diabetic complications. Depending on the C-peptide concentration 6 min after 1 mg of intravenous glucagon, the patients were allocated to groups with either moderate (n = 12) or low (n = 5) β-cell secretory capacity (moderate and low secretors, respectively). The groups had a 6-min C-peptide response greater than or less than 1.1 nmol/l, respectively (Table 1). Thereafter, the patients were randomized (by drawing a lot) to either training or nontraining. However, after completion of the nontraining program, four of the moderate secretors and all of the low secretors initially randomized to nontraining entered the training program. None of the patients had ever been treated with insulin, and all were treated with diet. In addition, six moderate secretors in the training group and three in the nontraining group and all of the low secretors were treated with sulfonylureas or biguanides. On experimental days, patients abstained from medication. For the majority, the medicine had been withheld for 24 h. The half-life of the drugs being known, no enhancing effects on insulin secretion should remain at the time of experiments. This is in line with the modest β-cell secretory responses found. Also, procedures and medication were identical before and after the 3-mo intervention period. Clinical characteristics are given in Table 2. Percent body fat was calculated from skin fold measurements (7), and waist and hip circumferences were measured at the level of umbilicus and of trochanter major, respectively.
Test and training protocol.
Before and after the 3-mo intervention periods, all subjects performed a stepwise incremental bicycle test on a bicycle ergometer. Heart rate and oxygen consumption (Jaeger Instruments, Hoechberg, Germany) were measured continuously during the test. Maximal oxygen consumption (V̇o2 max) was determined according to the leveling-off criterion. Individual training programs were made for all training subjects according to pretraining V̇o2 max and submaximal heart rates. Training was carried out at home on an ergometer cycle and included a minimum of 5 sessions/wk for 12 wk. Each training session started with a 5-min warm-up, after which the subjects had to complete altogether at least 20 min of exercise at 75% V̇o2 max. However, if necessary, they were allowed to alternate between periods of at least 5-min duration at this load and 5-min periods corresponding to 40% of V̇o2 max. The total exercise period usually lasted 30–40 min. To ensure that all training subjects had performed the desired amount of training, a hidden revolution counter on the ergometer was checked frequently by the investigators, and the subjects kept a training diary including information for each training bout on exercise duration, revolutions per minute, workload, and heart rate, e.g., recorded by a portable heart rate monitor. All subjects performed at least the desired amount of training and some up to 50% more, the average amount being similar in low and moderate secretors. Furthermore, all subjects had submaximal heart rates measured during bicycle ergometer tests every 4 wk, and the workload was adjusted accordingly to ensure the same relative load throughout the study period.
The hyperglycemic clamps were carried out before and after each intervention period. After training periods, the clamps were done 16 h after the last exercise bout. For 3 days before each examination, all subjects ingested >250 g carbohydrate/day. To exclude a change in diet, the subjects weighed all their food and drink for 3 days (2 weekdays and 1 Saturday or Sunday) before and at the end of the interventions (training or usual sedentary lifestyle). Daily energy intake and distribution on macronutrients (fat, carbohydrate, protein, and alcohol) were subsequently calculated using the National Danish Food Composition Tables (DANKOST, Levnedsmiddelstyrelsen, Denmark).
The subjects arrived in the laboratory at 0800 after an overnight fast of at least 10 h. After being weighed, they were put to bed and had a catheter (Venflon, Viggo, Sweden) inserted into a dorsal hand vein, with the tip of the catheter in the retrograde direction and placed as close as possible to the metacarpophalangeal joints. The hand was heated in a specially designed heating pad, providing arterialized blood samples (oxygen saturation 97 ± 1%). All blood samples were drawn from this catheter. A second catheter (650 × 1.8 mm; Secalon Cathy, Viggo, Sweden) was inserted into the medial cubital vein and introduced 35 cm from the brachiocephalic vein to avoid chemical phlebitis in response to the glucose infusate. Electrocardiogram, heart rate, and blood pressure were monitored throughout the experiment. By infusion of 20% glucose, plasma glucose concentrations were sequentially raised to 11, 18, and 25 mM and maintained for 90 min at each level in accordance with previously used protocols (27). With the use of exponentially decreasing infusion rates during transition, the new steady-state glucose level was established within 7–12 min. At the end of the 25 mM clamp, 5 g l-arginine diluted in 50 ml of sterile water were injected over 30 s, and blood was sampled at intervals during another 50 min, while the 25 mM plasma glucose concentration was maintained. Subjects who had fasting plasma glucose concentrations >11 mM had only the two highest clamp steps performed. Blood for immediate determination of plasma glucose was drawn at least every 5 min throughout the clamp, and the glucose infusion rate was adjusted within 2 min according to the result.
Blood was always sampled into iced tubes. Blood for determination of insulin, C-peptide, and glucagon was stabilized with 500 KIE Trasylol and 1.5 mg EDTA per ml blood and centrifuged immediately at +4°C. All blood samples were kept at −20°C until analysis. Plasma glucose was analyzed by the glucose oxidase method (YSI 23AM; Yellow Springs International, Yellow Springs, OH). Concentrations of arginine were measured spectrophotometrically. Plasma concentrations of insulin, C-peptide, and glucagon were determined with the use of commercially available RIAs (kindly donated by Novo Nordisk, Bagsværd, Denmark). All analyses were carried out in duplicate, and samples from a given subject were always in the same assay run. Heparinized plasma was analyzed for creatine kinase activity with a commercially available kit (Roche Diagnostics, Mannheim, Germany).
Statistical analysis and calculations.
Results are presented as means ± SE. Data obtained during glucose clamping were analyzed by two-way repeated-measures ANOVA. The Student-Newman-Keuls test was used for post hoc analysis to locate differences. Other data were tested by means of nonparametric ranking tests, the Mann-Whitney test for unpaired data and the Wilcoxon test for paired data. P < 0.05 was considered significant in a two-tailed test.
Body composition and training markers.
V̇o2 max increased with training in both moderate secretors (26 ± 1 vs. 30 ± 2 ml·min−1·kg−1, P < 0.05) and low secretors (34 ± 5 vs. 40 ± 5 ml·min−1·kg−1, P < 0.1). V̇o2 max was always higher in the latter compared with the former group (P < 0.05). In the nontraining groups, V̇o2 max did not change (data not shown). Submaximal heart rates at given workloads decreased with training (150 W, 148 ± 5 vs. 136 ± 5 beats/min in moderate secretors and 139 ± 5 vs. 124 ± 5 beats/min in low secretors; 100 W, 121 ± 5 vs. 111 ± 4 beats/min in moderate secretors and 117 ± 3 vs. 109 ± 4 beats/min in low secretors; all P < 0.05) but did not change in the nontraining groups (data not shown). Body weight, body mass index (BMI), percent body fat (25 ± 1 vs. 16 ± 3%, P < 0.05), and waist-to-hip ratio (1.02 ± 0.02 vs. 0.95 ± 0.02, P < 0.05) as well as fasting plasma insulin and C-peptide concentrations were significantly higher and fasting plasma glucose lower in moderate compared with low secretors (Table 2). None of these variables changed significantly (P > 0.05) during training or the corresponding period of sedentary lifestyle, and neither did creatine kinase activities in plasma (Table 2). Also, Hb A1c did not change with training (Table 2). The daily intake of carbohydrate, fat, protein, and alcohol was not changed by the interventions (data not shown).
Hyperglycemic clamp data.
During the hyperglycemic clamps, the coefficient of variation of plasma glucose concentrations in the final hour of each stage was always <5%, and the level of hyperglycemia was not different between patient groups or before and after 3 mo of either training or sedentary lifestyle (data not shown). During arginine stimulation, plasma concentrations of arginine were similar in all groups, whether trained or not, and the average peak value was 3.8 ± 0.3 mmol/l 1 min after the bolus injection.
As expected, in both moderate and low secretor groups, the insulin and C-peptide responses to increasing levels of hyperglycemia and arginine stimulation did not differ before and after 3 mo of sedentary life (Figs. 1 and 2). In contrast, in the type 2 diabetic patients with a moderate secretory capacity, the β-cell response to hyperglycemia and arginine was increased after compared with before training (Figs. 1 and 2). The effect of training attained statistical significance at 25 mM glucose for plasma insulin concentrations and at 18 mM glucose and during arginine stimulation for plasma C-peptide concentrations (Figs. 1 and 2). No effect of training was seen on the β-cell response in type 2 diabetics with a low secretory capacity (Figs. 1 and 2).
The basal plasma glucagon concentration was always similar in low and moderate secretors and similarly reduced during hyperglycemia in the two groups (Fig. 3). The increase in plasma glucagon concentrations upon arginine stimulation also did not differ between patient groups or with state of training (Fig. 3).
As a measure of insulin sensitivity, glucose clearance was calculated from glucose infusion rates and plasma glucose concentrations during glucose clamping and normalized for the corresponding plasma insulin levels. This index did not change with training (P > 0.05) in either moderate [at 18 mM glucose, 0.004 ± 0.001 (before) vs. 0.004 ± 0.001 (after) ml·min−1·kg−1·pmol−1·l−1; at 25 mM glucose, 0.006 ± 0.001 vs. 0.006 ± 0.001 ml·min−1·kg−1·pmol−1·l−1] or low (at 18 mM, 0.01 ± 0.01 vs. 0.02 ± 0.01 ml·min−1·kg−1·pmol−1·l−1; at 25 mM, 0.02 ± 0.01 vs. 0.02 ± 0.01 ml·min−1·kg−1·pmol−1·l−1) secretors.
The present study indicates that, in type 2 diabetic patients, physical training may enhance β-cell function. However, the effect of training depends on the remaining β-cell secretory capacity. Thus, as judged from responses of plasma insulin and C-peptide concentrations to stimulation by glucose during hyperglycemic clamping and of plasma C-peptide to stimulation by arginine, training enhances β-cell secretory capacity in diabetics with a moderate secretory capacity (Figs. 1 and 2). In contrast, in diabetics with a low secretory capacity, training does not affect β-cell function (Figs. 1 and 2). Although overall training enhanced insulin and C-peptide responses to hyperglycemia and arginine in parallel in moderate secretors, the effect of training attained statistical significance at the 0.05 level in different periods for the two variables (Figs. 1 and 2). This probably reflects the small number of patients who could be recruited to these demanding studies and analytical variability as well as difference in half-life between the two peptides. A role of the latter factor is in line with the finding that the training-induced percent increases in insulin responses tended to be higher than the increases in C-peptide responses (Figs. 1 and 2). Lack of steady state is also indicated by the finding that even at the end of hyperglycemic clamping, insulin and C-peptide concentrations in plasma often tended, albeit insignificantly, to increase with time. However, even though measured peripheral plasma concentrations did not completely correspond with steady-state levels, they probably reflected β-cell secretory capacity and C-peptide concentrations more unambiguously than insulin concentrations because C-peptide, in contrast to insulin, is not extracted in the liver.
In comparison with nondiabetic people, in the present study, not only the low secretors but also the moderate secretors had a very modest β-cell response to glucose stimulation. Thus, in healthy, young, untrained, lean subjects clamped at 11, 20, and 35 mM glucose, insulin concentrations have been found to be ≈360, 1,800, and 6,000 pM and C-peptide concentrations ≈2, 6, and 11 pmol/ml, respectively (26). In healthy, middle-aged, untrained, obese (BMI 29 kg/m2, n = 7) or lean (BMI 24 kg/m2, n = 7) subjects with plasma glucose clamped at 11, 18, and 25 mM, we have found insulin and C-peptide concentrations to be ≈500, 2,500, and 5,500 pM and ≈3, 6, and 9 pmol/ml, respectively, in the obese and ≈300, 1,000, and 2,700 pM and ≈2, 4, and 7 pmol/ml, respectively, in the lean subjects (unpublished data). Comparisons with data in Figs. 1 and 2 emphasize that the β-cell function of the patients in the present study was indeed compromised and, even in the moderate secretors, far from normalized by the training program.
Less direct estimates of the effect of training on insulin secretion in type 2 diabetics have previously been provided by glucose tolerance testing. Some of these studies did not find any effect of training (33, 35). However, in accordance with an increase in β-cell secretory capacity, in one study using an oral glucose tolerance test, the plasma insulin concentration in the 30th min was higher after training (31), and in another study, the integrated C-peptide response was increased after training (20). In a study in which training was combined with diet therapy, an increase in the first-phase insulin response to an intravenous glucose tolerance test was seen in patients with type 2 diabetes (1). In light of the present findings, it is likely that the inconsistency between previous studies in part reflects that they did not subdivide the studied diabetics according to their β-cell function before training. Interestingly, in rats, a study of experimental non-insulin-dependent diabetes has been carried out that fully agrees with the present study (9). The rats were made mildly to severely diabetic by partial pancreatectomy and studied by hyperglycemic clamps. Exercise training improved glucose-stimulated insulin secretion in mildly and moderately diabetic rats but did not influence secretion in more severely diabetic rats (9).
It has been shown previously that intensive treatment of type 2 diabetic patients with insulin, sulfonylurea, or diet, which lower plasma glucose concentrations, causes recovery of overloaded β-cells and in turn improvement in insulin secretion (37). We had imagined that exercise training would have a similar effect due to contraction-mediated glucose uptake in muscle and increased sensitivity to insulin of glucose uptake in muscle (4). Reduction of the glucose stress might very well have a greater impact on moderately than on more severely failing β-cells because the relationship between secretory capacity of failing β-cells and prevailing plasma glucose concentration is hyperbolic (3). However, in the studied diabetic patients, both fasting plasma glucose and Hb A1c levels, which reflect the average plasma glucose concentration during 8–12 wks before analysis, did not change during training (Table 2). Neither did training-induced changes in β-cell secretory capacity correlate with increases in measures of insulin sensitivity. Thus, during hyperglycemic clamping, glucose clearance rates normalized for plasma insulin concentrations were similar before and after training in moderate secretors. Correspondingly, during hyperinsulinemic (630 and 17,000 pM) isoglycemic clamping, glucose clearance was also similar before and after training in these patients (2.9 ± 0.5 and 7.2 ± 0.7 ml·min−1·kg−1, respectively, before, and 2.8 ± 0.5 and 7.0 ± 0.7 ml·min−1·kg−1, respectively, after training; P > 0.05, unpublished data). This was so even though the training program was carefully monitored and caused marked adaptations in V̇o2 max and heart rate response to exercise. Lack of significant changes in insulin action and Hb A1c with training in type 2 diabetic patients is not a rare finding (2, 31, 34). In fact, a meta-analysis has been necessary to clearly demonstrate a reduction in Hb A1c (2).
Apparently, the observed increase in β-cell function in the present study cannot without reservations be ascribed to improvement of glucose control with training. It has previously been shown that exercise-induced muscle damage may enhance β-cell function (17). However, in the present study, creatine kinase activity in plasma, which increases upon muscle damage, was not higher during β-cell evaluation after compared with before training (Table 2). The studied patients were insulin resistant [e.g., during hyperinsulinemic (630 pM) isoglycemic clamping, glucose clearance was, before training, lower in moderate (2.9 ± 0.5 vs. 5.8 ± 0.6 ml·min−1·kg−1, P < 0.05) and low (4.9 ± 1.1 vs. 7.5 ± 1.1 ml·min−1·kg−1, P < 0.1) secretors than in healthy controls matched for age, fitness, and BMI; unpublished data]. Recently, evidence has been presented that insulin-resistant muscle produces peptides that may increase β-cell mass and/or secretion (10, 13, 24). So, it may be speculated that in the present study, secretion of such factors was enhanced by training, and that moderately failing β-cells were more susceptible to these factors than more severely failing β-cells. In this context, it is also interesting to note that in the above-mentioned study of partially pancreatectomized rats, it was concluded that mechanisms other than reduced glycemia seem to account for the positive β-cell adaptation with training (9).
From a biological perspective, it is interesting to note that the β-cell adaptation to training seen in diabetics with a moderately reduced capacity for insulin secretion is the reverse of that seen in healthy young and 65-yr-old subjects (19, 27). In healthy subjects, training induces a decrease in the β-cell secretory capacity that is accurately matched to an accompanying increase in insulin action (6, 27). It has been proposed that this pancreatic adaptation implicates sparing of β-cell function and in turn reduced risk of type 2 diabetes (4, 27). On the other hand, in subjects who have developed type 2 diabetes, the opposite adaptation, i.e., an increase in β-cell function, seems much more appropriate and may represent a health benefit of training in this population. However, as judged from the present findings, the clinical impact of the adaptation seems modest. As mentioned above, β-cell function was far from normalized by training, and an improvement could only be detected at glucose levels >11 mM. Furthermore, despite the increase in β-cell function, glucose control, evaluated from fasting plasma glucose and Hb A1c levels, was not improved in the face of constant energy intake. It is possible that an improvement in insulin sensitivity is necessary for improved glucose control with training, and it cannot be excluded that in such a setting a more dramatic recovery of β-cell function would be seen.
The diverging adaptations between healthy subjects and diabetics might reflect the bell-shaped relationship between β-cell secretory capacity and prevailing plasma glucose concentration (3). Subtle reductions in plasma glucose will lower secretory capacity if this is on the ascending part of the curve, i.e., in healthy subjects, but will tend to increase secretory capacity if it is on the descending part of the curve, i.e., in diabetics with failing β-cell function. A possible alternative explanation for the diverging adaptations is that circulating factors inhibiting and enhancing β-cell secretory capacity are produced in healthy and insulin-resistant subjects, respectively (10), and that the secretion of these factors may be increased by training. In contrast to β-cell responses to stimulation, in the present study, glucagon responses to arginine never changed with training in diabetics (Fig. 3). This corresponds to findings in young healthy subjects (6).
In the present study, the diabetics had their β-cell secretory capacity evaluated by a glucagon test at enrollment and before allocation into groups of moderate and low secretors, respectively. The two groups did not differ in regard to age and time elapsed since diagnosis of diabetes (Table 2). Nevertheless, in addition to differences in fasting levels of glucose, insulin and C-peptide, the two groups also showed significant differences in BMI, waist-to-hip ratio, and percent body fat, indicating that they represent distinct subgroups within the overall population of type 2 diabetic patients. The data from the two groups are compatible with the view that, in type 2 diabetics, as is the case in nondiabetics, an increase in body fat per se reduces insulin sensitivity (see above) and stimulates insulin secretion. It also cannot be excluded that the higher body weight in moderate compared with low secretors by some unknown mechanism favored the increase in β-cell function in response to training.
In conclusion, in type 2 diabetic patients, training may enhance β-cell function, whereas α-cell function is not affected. Training has opposite effects on β-cell function in type 2 diabetic patients and healthy subjects, with a potential increase of insulin secretion in the former and a decrease in the latter. In type 2 diabetics, the β-cell adaptation depends on the remaining secretory capacity and may be seen in the absence of significant changes in insulin sensitivity and Hb A1c concentration.
Financial support from the Danish National Research Foundation (grant no. 504-14), the Danish Medical Research Council (grant no. 12-9360), the Novo Foundation, the Nordisk Insulin Foundation Committee, the P. Carl Petersen Foundation (grant nos. 890097 and 900101), the Foundation of 1870, the Jacob and Olga Madsen Foundation, and the Danish Hospital Foundation for Medical Research (grant no. 4589) is gratefully acknowledged.
We thank Birgit Mollerup, Gerda Hau, Vibeke Staffeldt, Lisbeth Kall, Regitze Kraunsø, and Anette Dalgård for dedicated technical assistance. Chief Physician Jens Halkjær-Kristensen, Dept. of Internal Medicine, Rigshospitalet, is thanked for placing laboratory facilities at our disposal. We also thank the patients who volunteered in these comprehensive studies.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 by American Physiological Society