Association Between Cardiorespiratory Fitness and the Determinants of Glycemic Control Across the Entire Glucose Tolerance Continuum

Type 2 diabetes (T2D) is characterized by chronic hyperglycemia that develops when pancreatic β-cell insulin secretion fails to compensate for the deterioration in insulin sensitivity (1). Physical activity aimed at improving cardiorespiratory fitness is prescribed as part of standard-of-care treatment for T2D (2), primarily because randomized controlled clinical trials show that exercise training reduces hyperglycemia in patients with T2D (3–5) and delays the onset of T2D in at-risk individuals (6). Interestingly, in a longitudinal study of 8,633 nondiabetic men, Blair and colleagues showed that high cardiorespiratory fitness (as determined by maximal oxygen consumption [VO_2max], measured during exhaustive incremental workload exercise) confers protection against developing T2D-related hyperglycemia (7). A further longitudinal study by Church et al. (8), examining 2,316 men with T2D, reported that high cardiorespiratory fitness reduced cardiovascular disease mortality. Consequently, poor fitness is considered a key determinant of the pathophysiological progression of glucose intolerance. However, because poor glucose disposition, driven by inadequate β-cell insulin secretory function in the presence of poor insulin sensitivity, is the fundamental cause of hyperglycemia in T2D, it is prudent to determine whether cardiorespiratory fitness is related to these pathophysiological factors. Indeed, we and others have shown that aerobic exercise training that improves cardiorespiratory fitness also increases insulin sensitivity (9–14) and improves β-cell insulin secretory function (10,14,15) in patients with T2D. Nonetheless, whether the predictive value of cardiorespiratory fitness for determining longitudinal glycemic control is explained by an association between fitness and the underlying determinants of glycemic control (insulin sensitivity and/or insulin secretory function) is not clear.

With the a priori knowledge (9–16) that exercise training improves VO_2max and β-cell insulin secretory compensation for changing insulin sensitivity (the glucose disposition index) and the evidence that both variables are reduced in normoglycemic first-degree relatives of T2D patients (17), we hypothesized that low cardiorespiratory fitness would be associated with low disposition index, the underlying pathophysiological determinant of glucose intolerance. Therefore, our aim was to examine this relationship in a large cohort representing the entire glucose tolerance continuum from normal glucose tolerance (NGT) to T2D.

Potential participants underwent medical screening to determine their eligibility for the study. This included a medical history assessment, an electrocardiogram, and blood chemistry screening. Evidence of prior or current chronic pulmonary, hepatic, renal, gastrointestinal, or hematological disease; weight loss (>2 kg in the last 6 months); smoking; pregnancy; and contraindication to an exercise test were used as exclusion criteria. Subjects were recruited by newspaper/radio advertisement from the local municipal areas in Copenhagen, Denmark, and Cleveland, OH. All subjects provided oral and written informed consent prior to participation, and the methods were approved by ethics committees at both locations (Institutional Review Board, Cleveland Clinic, and Scientific Ethics Committee of the Capital Region of Denmark). Data from some subjects participating in this work have previously been published (3,9,10,18,19). Subjects included in the study were stratified by glucose tolerance status (i.e., NGT, impaired glucose tolerance, or T2D) according to their 2-h glucose on an oral glucose tolerance test (OGTT) (2). The final study cohort included 313 subjects.

Tests took place at the Clinical Research Unit at the Cleveland Clinic and at the clinical research laboratory of the Centre of Inflammation and Metabolism, Rigshospitalet. Subjects being treated with antidiabetic drugs (metformin, N = 38; sulfonylureas, N = 14; GLP-1 analogs, N = 12; and dipeptidyl peptidase-4 inhibitors, N = 2) withheld their medications for 5 days prior to metabolic testing, during which time diet and physical activity records were used to ensure there were no pretesting variations in dietary or activity habits. Subjects also abstained from structured exercise for at least 24 h prior to metabolic testing.

Height and weight were measured using standard techniques. Whole-body adiposity was estimated using DEXA (Lunar iDXA; GE Healthcare, Madison, WI). Subjects performed an exercise test to determine their VO_2max during exhaustive incremental workload exercise to volitional exhaustion, which provides a measure of cardiorespiratory fitness. Heart rate was measured using a chest-worn monitor (Polar Electro, Kempele, Finland) and VO₂ was measured online using an automated system (Oxycon Pro, Erich Jaeger, Hoechberg, Germany; CPET, Cosmed, Rome, Italy). VO_2max was determined when at least two of the following criteria were met: respiratory exchange ratios >1.1; no further increase in VO₂ despite increasing workloads, with heart rate greater than age-predicted maximum; or volitional fatigue. The VO_2max test was conducted 48–72 h prior to subsequent metabolic assessments to avoid the acute effects of physical exercise on glycemic control. On a separate day, after an 8- to 10-h overnight fast, subjects came to the laboratory and an antecubital venous cannula was placed. Baseline blood samples were collected and subjects ingested 75 g anhydrous glucose dissolved in 300 mL water (standard OGTT). After glucose ingestion, venous blood samples were collected every 15 min for 2 h. Venous blood was collected into tubes containing EDTA and 500 kiU/mL aprotinin (serine protease inhibitor; Sigma) or into plain tubes containing no anticoagulant. Plain tubes were allowed to clot for 30 min at room temperature, while EDTA tubes were kept on ice. Blood was centrifuged at 2,000g for 15 min at room temperature, and respective serum/plasma was stored at −80°C until analysis.

Glucose concentrations were measured using a bedside analyzer (YSI Stat, Yellow Springs; ABL, Radiometer Medical, Brønshøj, Denmark); insulin and C-peptide concentrations were determined by electrochemiluminescence immunoassay (E-modular; Roche, Basel, Switzerland); total cholesterol and triglycerides were determined by colorimetric assays (P-modular; Roche). A further venous blood sample was collected into an EDTA tube so that hemoglobin A_1c (HbA_1c) levels could be determined by high-performance liquid chromatography (HPLC) (Tosoh G7 analyzer; San Francisco, CA).

Insulin sensitivity during OGTT (Si_OGTT) was calculated using a model previously validated against the gold-standard hyperinsulinemic-euglycemic clamp method (19): Si_OGTT (µmol/kg/min/pmol ⋅ L) = 0.138 − (0.00231 × BMI) − (0.00118 × G₁₂₀) − (0.0000135 × I₃₀) − (0.00000678 × I₉₀), where BMI is measured as weight in kilograms divided by the square of height in meters, G₁₂₀ is glucose at 120 min, and I₃₀ and I₉₀ are insulin at 30 and 90 min during OGTT. GSIS_OGTT was determined from the OGTT plasma C-peptide response rather than the insulin response to overcome the confounding effects of hepatic insulin extraction, which removes up to 80% of insulin at first pass prior to measurement in the peripheral venous blood (20). Early- and late-phase oral glucose–stimulated insulin secretion (GSIS_OGTT) was calculated as the area under the C-peptide curve from 0 to 30 and 30 to 120 min during OGTT, respectively. We previously showed that the log-log relationship between these measures of GSIS_OGTT and Si_OGTT is described by an inverse linear model (19); therefore, the oral disposition index (DI_OGTT), which is a measure of glucose disposal and estimates pancreatic β-cell insulin secretory compensation for changing insulin sensitivity during OGTT, was calculated as the product of GSIS_OGTT and Si_OGTT.

Relationships between VO_2max and measures of glycemic control (HbA_1c, fasting glucose, and 2-h OGTT glucose) and GSIS_OGTT were best fit by log-log curvilinear regression. Relationships between VO_2max and measures of Si_OGTT and DI_OGTT were best fit by linear regression. For testing of the difference between correlation coefficients, r values were converted to z scores using Fisher r-to-z transformation and were then compared using the Cohen test (21). Forced-entry multiple regression was used to determine the relative contributions of VO_2max and other independent variables (age, sex, body weight, BMI, and whole-body fat percentage) to the variance in glycemic control (HbA_1c, fasting glucose, 2-h OGTT glucose), Si_OGTT, GSIS_OGTT, and DI_OGTT. One-way ANOVA and Bonferroni post hoc tests were used to assess differences between glucose tolerance groups. Statistical significance was set at P < 0.05 and determined using Prism v6 (GraphPad, La Jolla, CA) and SPSS v20 (IBM, New York, NY).

Table 1 shows subject characteristics stratified by glucose tolerance status. Glycemic control (HbA_1c, fasting glucose, and 2-h glucose) was progressively reduced across the glucose tolerance continuum. Si_OGTT was markedly lower in subjects with impaired glucose tolerance and T2D compared with those with NGT. Early- and late-phase GSIS_OGTT were greatest in impaired glucose tolerant subjects, and lowest in the T2D group. Early- and late-phase DI_OGTT showed progressive deterioration across the glucose tolerance groups.

Table 2 and Figs. 1 and 2 show the correlation coefficients for the associations between VO_2max and glycemic control, Si_OGTT, GSIS_OGTT, and DI_OGTT. VO_2max was inversely associated with clinical markers of glycemic control: HbA_1c (Fig. 1A), fasting glucose (Fig. 1B), and 2-h glucose during OGTT (Fig. 1C); however, these relationships were strongest in NGT subjects, while significant correlations were not evident in the subgroup of subjects with T2D (Table 2). Furthermore, these relationships were significantly weaker in subjects with T2D compared with those with NGT (NGT vs. T2D z scores for HbA_1c, fasting glucose, and 2-h glucose were −2.51 [P < 0.05], −2.48 [P < 0.05], and −2.67 [P < 0.01], respectively). Fig. 2A demonstrates that VO_2max was directly correlated with Si_OGTT. This significant relationship was evident across the entire glucose tolerance continuum (Table 2), although with decreasing r values across the continuum where the relationship in T2D subjects was significantly weaker than in NGT subjects (NGT vs. T2D z score = 2.52, P < 0.05). Fig. 2B and C indicate that VO_2max was inversely related to early- and late-phase GSIS_OGTT; nevertheless, these relationships were weaker in subjects with T2D, particularly during late phase (NGT vs. T2DM z score = −1.73, P = 0.08), where the relationship was not significant (Table 2). Fig. 2D and E show that VO_2max was directly related to early- and late-phase DI_OGTT. This finding was also consistent across the whole glucose tolerance continuum (Table 2). These significant relationships persisted when analyzing absolute VO_2max (liters per minute) vs. HbA_1c (r = −0.37, P < 0.0001), fasting glucose (r = −0.22, P < 0.001), 2-h OGTT glucose (r = −0.34, P < 0.0001), Si_OGTT (r = 0.44, P < 0.0001), GSIS_OGTT (r = −0.26, P < 0.05; r = −0.21, P < 0.05 [early and late phase, respectively), and DI_OGTT (r = 0.25, P < 0.0001; r = 0.26, P < 0.0001). However, r values were lower compared with the analyses made against relative VO_2max (milliliters per kilogram per minute) indicating that differing body mass accounts for some of the variance in these relationships. Multiple regression analysis (Table 3) where VO_2max was entered simultaneously with the other independent variables (age, sex, weight, BMI, and % body fat) showed that although VO_2max was significantly associated with glycemic control (HbA_1c, fasting glucose, and 2-h OGTT glucose), Si_OGTT, GSIS_OGTT, and DI_OGTT, the other independent variables also explained some of the variance. By comparing significant β-values in Table 3, these analyses also indicate that the majority of the variance in glycemic control was related to VO_2max, while the majority of variance in GSIS_OGTT, Si_OGTT, and DI_OGTT was related to body weight and BMI, respectively.

In the current study, we demonstrate that cardiorespiratory fitness is inversely related with clinical markers of glycemic control, such that high HbA_1c, high fasting glucose, and impaired oral glucose tolerance are associated with low VO_2max. We further found that cardiorespiratory fitness is positively related to insulin sensitivity and inversely related to GSIS_OGTT. However, the major novel finding of this work is that cardiorespiratory fitness is associated with pancreatic β-cell insulin secretory compensation for changing insulin sensitivity (DI_OGTT) in a cohort representing the entire glucose tolerance continuum.

The Italian Diabetes and Exercise Study (IDES [22]) demonstrated that changes in fitness after aerobic training predicted improvements in HbA_1c (and other cardiovascular disease risk factors). In this study, we show that VO_2max is inversely associated with HbA_1c (and other markers of glycemic control such as fasting and 2-h OGTT glucose) in a large cohort representing the entire glucose tolerance continuum, supporting the findings from Blair’s group, who showed that high VO_2max confers protection against the development of T2D over a 6-year period even when adjusted for age and parental diabetes (7). These data provide strength to the notion that maintaining cardiorespiratory fitness will help prevent T2D. That said, it should be noted that when the T2D subgroup was analyzed in isolation, the relationship between VO_2max and glycemic control (HbA_1c and fasting and OGTT glucose) was not evident (Table 2), possibly suggesting that once T2D is present cardiorespiratory fitness is a lesser determinant of the state of glycemia.

Hyperglycemia in individuals with T2D develops when insulin secretion no longer compensates for poor insulin sensitivity, indicating a state of β-cell dysfunction. Indeed, exercise training improves glycemic control (3–6,23), and it is traditionally thought that this benefit is driven entirely by increases in muscle glucose uptake and/or increased insulin sensitivity (rev. in 24,25). Furthermore, VO_2max is associated with insulin sensitivity, as first shown in 1983 by Rosenthal et al. (26), and we now confirm this finding across the entire continuum of glucose tolerance. We also found VO_2max to be inversely associated with GSIS_OGTT; however, as for variables of glycemic control (HbA_1c and fasting and 2-h glucose), such associations were less evident in subjects with T2D. Given that VO_2max is well correlated with insulin sensitivity even in subjects with T2D, the poor relationship between VO_2max and variables of glycemic control in subjects with diabetes is potentially explained by the fact that the solid inverse relationship between insulin sensitivity and insulin secretion is also diminished in subjects with T2D, as evidenced by the deterioration in glucose disposition. Thus, again, in T2D once the disease is established, cardiorespiratory fitness is a lesser determinant of insulin secretion. However, it is important to note that glycemic control is not entirely governed by insulin sensitivity. The ability to restore β-cell function is particularly relevant for the treatment of T2D, a disease in which β-cell insulin secretory function is no longer able to overcome the poor underlying degree of insulin sensitivity. Thus, it is of importance to examine insulin secretory compensation (i.e., the disposition index). Interestingly, aerobic exercise training increases the disposition index in subjects without (16,27) and in subjects with (10) T2D, and Chen et al. (28) have demonstrated that self-reported weekly minutes of physical activity are directly associated with disposition index. In support of such findings, we now demonstrate that high VO_2max is also predictive of high β-cell insulin secretory compensation for poor insulin sensitivity (DI_OGTT) across the entire glucose tolerance continuum. While these relationships do not indicate causality, our findings do highlight an important association between cardiorespiratory fitness and the key determinant of T2D-related hyperglycemia. However, it should also be noted that the genetic component of VO_2max (29) has not been addressed in this study and should be examined in future work.

The underlying mechanisms as to why VO_2max (or aerobic training) may be positively associated with β-cell compensation are not well defined. Physical activity per se, which is a major determinant of VO_2max, decreases blood glucose and lipids and may thereby relieve glucolipotoxicity in β-cells, allowing restoration of appropriate insulin secretory function (30). Further to this, increased physical activity lowers circulating levels of proinflammatory cytokines (tumor necrosis factor-α, leptin) and increases levels of various growth factors, hormones, and cytokines (growth hormone, insulin-like growth factor-1, GLP-1, interleukin-6, and interleukin-1 receptor antagonist) that have direct beneficial effects on β-cells (30). Irrespective of the speculated mechanisms, our findings support the need for future work examining the role of exercise in β-cell health.

There are some limitations to our study. Firstly, it is important to note that pausing antidiabetic drugs for 5 days prior to metabolic testing may induce acute increases in glycemia, potentially inducing glucotoxicity. On the contrary, this time period may not be sufficient to wash out all drugs. With these issues in mind, care should be taken when interpreting, for example, the relationship between VO_2max and HbA_1c and whether it would be altered when pausing the antidiabetic drugs for 5 days. Therefore, future studies should carefully consider the rationale for withholding drugs in the context of the primary outcomes of the study. Particular attention should also be drawn to the strength of the correlations. Despite finding significant relationships in the whole cohort, VO_2max explains just ∼11% of the variance in measures of glycemic control (HbA_1c and fasting and OGTT glucose), which indicates that other factors, such as the other independent variables included in the multiple regression analyses (Table 3), are also involved. Furthermore, for all variables, the strength of the relationships with VO_2max progressively diminishes with advancing glucose intolerance. This is particularly interesting because it highlights the complexity of the pathophysiology of T2D. Indeed, prior work in subjects with T2D has found a relationship between aerobic exercise training–induced changes in fitness and glycemic control (31). Furthermore, we previously found that patients randomized to a training intervention that does not increase VO_2max fail to improve glycemic control compared with patients randomized to training that increases VO_2max (3). However, we (32) and others (33) have also shown that exercise training–induced improvements in VO_2max do not always extrapolate to improved glycemia. Accordingly, while VO_2max explains ∼11% of the variance in measures of glycemic control, it accounted for ∼54% of the variance in insulin sensitivity. Thus, although cardiorespiratory fitness is predictive of insulin sensitivity, this does not extrapolate to the same magnitude of prediction of glycemic control. On the other hand, ∼17% of the variance in DI_OGTT was accounted for by VO_2max. As such, the underlying mechanisms linking cardiorespiratory fitness to improvements in glycemic control require more detailed investigation. Lastly, it is important to note that body weight and BMI explain a large part of the variance of GSIS_OGTT and of Si_OGTT and DI_OGTT, respectively (Table 3). Consequently, it is likely that weight loss combined with exercise training–induced increases in cardiorespiratory fitness is required to mediate the most optimal beneficial adaptation in glycemic control.

In conclusion, our study has advanced the prior knowledge that high cardiorespiratory fitness is associated with good glycemic control by showing that VO_2max is also positively associated with the disposition index in a cohort of adults representing the entire glucose tolerance continuum. While the large contribution of body weight and BMI must not be ignored, these findings are clinically significant because they also highlight the importance of an individual’s cardiorespiratory fitness level in relation to the fundamental pathophysiological cause of hyperglycemia-related chronic disease.

Acknowledgments. The authors thank Julianne Filion, RN (Department of Pathobiology, Cleveland Clinic), for her clinical nursing support and Lisbeth Andreasen, MSc (Department of Clinical Biochemistry, Rigshospitalet), for her assistance with biochemical assays.

Funding. This work was funded by a Paul Langerhans Program grant from the European Foundation for the Study of Diabeteshttp://dx.doi.org/10.13039/501100001648/Amylin Pharmaceuticals, Inc. (to T.P.J.S.). Additional support for this study came from the Danish Center for Strategic Research in Type 2 Diabetes (Danish Council for Strategic Research, grant nos. 09-067009 and 09-075724), and the National Institutes of Healthhttp://dx.doi.org/10.13039/100000002 (NIH) (RO1 AG12834) (to J.P.K.). S.K.M. was supported by an NIH T32 grant (DK007319 [training grant] to J.P.K.). The Centre of Inflammation and Metabolism is supported by the Danish National Research Foundation (grant no. 02-512-55).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. T.P.J.S. designed the study and drafted the manuscript. T.P.J.S., S.K.M., K.K., S.H.K., J.M.H., M.J.L., and J.P.K. implemented the study and edited the manuscript. T.P.J.S., S.K.M., J.M.H., M.J.L., and J.P.K. had access to and analyzed data. T.P.J.S., S.K.M., K.K., J.M.H., M.J.L., and J.P.K. discussed the data. T.P.J.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.