Intensity and Amount of Physical Activity in Relation to Insulin Sensitivity:  The Insulin Resistance Atherosclerosis Study FREE

PHYSICAL ACTIVITY has been related to reduced incidence of non–insulin-dependent diabetes mellitus (NIDDM).^1- 2 Most of the available evidence derives from estimates of total energy expenditure or participation in vigorous exercise. The potential benefit of moderate-intensity physical activity on NIDDM risk is unclear.³ Although mechanisms for the effect of physical activity on glucose tolerance have yet to be fully elucidated, improved insulin sensitivity (S_I) may be a key factor. From controlled studies, exercise training is associated directly with improved S_I,^4- 7 although some doubt remains as to the extent to which this effect is mediated by patterns of short-term fuel availability and use by active muscle or by obesity.^8- 11 Hughes et al¹² showed that exercise training of between 50% and 75% of maximal capacity can improve S_I in individuals with impaired glucose tolerance. From community studies, increased levels of overall habitual physical activity have been positively associated with surrogate measures of S_I among individuals without diabetes^13- 14 and among those with impaired glucose tolerance,¹⁵ independent of obesity.

Thus, questions remain regarding the nature and amount of physical activity required to have a sustained, beneficial impact on glucose and insulin metabolism at the individual and the community levels. The Centers for Disease Control and Prevention (CDC), Atlanta, Ga, and the American College of Sports Medicine (ACSM), Indianapolis, Ind, have recently recommended that every US adult should accumulate at least 30 minutes of moderate-intensity physical activity (3 to 6 metabolic equivalents [METs]) on most, preferably all, days of the week.¹⁶ The same recommendation was put forth by a 1996 National Institute of Health Consensus Statement.¹⁷ While ample evidence is available relative to physical activity, cardiovascular health, and overall mortality, it is not clear whether adherence to this recommendation would be expected to favorably affect S_I. The Insulin Resistance Atherosclerosis Study (IRAS) afforded an opportunity to assess whether self-reported participation in physical activity of moderate as well as vigorous intensity was associated with better S_I in a large, diverse cohort in whom S_I was measured by the frequently sampled intravenous glucose tolerance test (FSIGT).

The recruitment goal for 1600 participants was to obtain nearly equal representation of participants across glucose tolerance status (normal; impaired glucose tolerance [IGT]; non–insulin-taking NIDDM; ethnicity [African American, Hispanic, and non-Hispanic white]; sex; and age [40-49 years, 50-59 years, 60-69 years]). Participants were recruited at 4 clinical centers between October 1992 and April 1994. Ethnicity was determined by self-report. Two of the clinical centers (Los Angeles, Calif, and Oakland, Calif) were assigned to recruit African American and non-Hispanic white participants. In these centers, individuals were sampled from the members of a nonprofit health maintenance organization. In the other 2 clinical centers (San Luis Valley, Colo, and San Antonio, Tex), Hispanic and non-Hispanic white participants were recruited from ongoing population-based epidemiologic studies. A total of 1625 people were included in the final sample: 38% non-Hispanic white, 34% Hispanic, and 29% African American. All participants provided informed consent as approved by their respective field center's institutional review board. Further details of sampling and the self-report of ethnicity have been described elsewhere.¹⁸

Participants were asked to fast for 12 hours prior to each of 2 visits, abstain from heavy exercise and alcohol for 24 hours, and refrain from smoking the morning of the visit. A 2-hour, 75-g oral glucose tolerance test (Orange-dex, Custom Laboratories, Baltimore, Md) was performed during the first visit, and World Health Organization criteria¹⁹ were used to assign glucose tolerance status. Individuals currently taking oral hypoglycemic medications were classified as having NIDDM regardless of oral glucose tolerance test results.

Insulin sensitivity was assessed during the second visit (within 4 weeks) using the FSIGT^20- 21 with minimal model analysis.²² Two modifications of the protocol were used: injection of insulin rather than tolbutamide²⁰ and a reduced number of plasma samples (12 rather than 30).²¹ Glucose, in the form of a 50% solution (0.3 g/kg of body weight), and regular human insulin (0.03 U/kg) were injected at 0 and 20 minutes, respectively. Blood specimens were collected over a 3-hour period (at −5, 2, 4, 8, 19, 22, 30, 40, 50, 70, 100, and 180 min). Insulin sensitivity was calculated by mathematical modeling methods; the time course of plasma glucose was fit using nonlinear least squares methods with the plasma insulin values as a known input to the system (according to the method known as MINMOD, which was developed by Richard N. Bergman, PhD, in 1986). A comparative validity study (n=55, including 11 subjects with normal glucose tolerance, 20 with impaired glucose tolerance, and 24 with NIDDM) was conducted comparing this technique of measuring S_I with estimates derived from the glucose clamp technique (r=0.55, P<.001).²³

Plasma glucose concentrations were measured in duplicate using the glucose oxidase technique on an autoanalyzer (Yellow Springs Equipment Co, Yellow Springs, Ohio). Plasma insulin was determined by radioimmunoassay.²⁴

Physical activity was assessed using 2 approaches. First, usual frequency of vigorous activity was ascertained using 5 predefined responses that ranged from "rarely to never" to "5 or more times per week."²⁵ This scale has been shown previously to be predictive of incident NIDDM.² Second, a 1-year recall of physical activities was administered by centrally trained and certified interviewers. For quality control purposes, audiotapes of the physical activity interviews were monitored centrally on a quarterly basis throughout the data collection period. The structured interview was a modification of a validated instrument²⁶ that incorporated activities common among IRAS study participants, including ranching-related and homemaking activities. These activities were queried in groups according to home, work, or leisure time and according to intensity of activities based on published values in METs (ratio of metabolic rate during the activity to the resting metabolic rate).²⁷ Groupings are given in Table 1. For each activity group, usual frequency and duration of participation was recorded, from which estimated energy expenditure (EEE) was determined. Energy expended per year was estimated by summing across all activity groups, plus the energy expended during reported time spent in sleep (assigned a MET value of 1.0), plus the EEE from light activities (assigned a MET value of 1.5, eg, sitting). This was derived by subtraction assuming that all time not accounted for in moderate (MET assignment for activity grouping, 3.5-5.0) or vigorous (MET assignment for activity grouping ≥6.0) activities or sleep was spent in light activities. Further details regarding the assessment are available from the authors.

Weight was measured to the nearest 0.1 kg, height was recorded to the nearest 0.5 cm, and body mass index (BMI) was calculated as weight in kilograms divided by the square of the height in meters (kg/m²). Girth measurements were estimated as the average of duplicate measures (taken to the nearest 0.5 cm using a steel tape). Minimum waist circumference was measured on bare skin during midrespiration at the natural indentation between the 10th rib and the iliac crest. Hip girth was measured at the maximum circumference of the buttocks. Waist-to-hip ratio (WHR) was calculated as a surrogate measure of visceral adiposity. Nutrient intake was assessed with a 114-item food frequency interview modified from the National Cancer Institute–Health Habits and History Questionnaire^28- 29 to include regional and ethnic food choices across the 4 clinical centers. The nutrient database (HHHQ-DIETSYS Analysis Software, Version 3.0, National Cancer Institute, Bethesda, Md, 1993) was expanded to accommodate the new foods based on values obtained from the Minnesota Nutrition Data System, Program Version 2.3.³⁰ Additional standardized interviews were used to ascertain previous physician diagnosis of diabetes, medication use, smoking status, and alcohol intake.

After excluding individuals with missing S_I (n=143) or missing physical activity data (n=16) and 1 statistical outlier, 1467 individuals were included in the present analyses. For statistical models that included additional variables, sample sizes varied slightly because of occasional missing values. Analyses were conducted in the full sample, and, to focus specifically on the potential effect of light-to-moderate intensity activities on S_I, analyses were repeated among the subset of individuals who reported essentially no time spent in vigorous activities (<1 hour per month for any vigorous activity group [≥6 METs] on the 1-year recall, n=446).

For descriptive purposes, quintiles of physical activity were considered in relation to S_I. Because a threshold effect of activity and S_I was not evident, activity variables (either total EEE or its components, vigorous EEE, and nonvigorous EEE) were included in linear regression analyses in their original, continuous form so that study hypotheses could be evaluated with maximal statistical power. Regression analysis assumes that the distribution of the residual values from the fitted model are normally distributed. Approximately 15% of the sample had an S_I value of 0. This most likely reflects an inability of the FSIGT to compute finite values for individuals who are extremely insulin resistant. In addition, the distribution of S_I was skewed to the right. Therefore, we calculated the natural log of S_I, adding a constant 1 to all values since the log of 0 cannot be taken. With this transformation, the distributions of the resulting residual values approached normality. To confirm internal consistency of the results, analyses were repeated excluding individuals with an original S_I value of 0; results were essentially unchanged. For comparison with existing epidemiologic studies that used insulin levels as a surrogate for insulin resistance, analyses were repeated using the natural log of fasting insulin as the dependent variable.

Covariates included in regression models as potential confounders were age, sex, ethnicity, clinical center, smoking status, alcohol intake, percentage of calories from dietary fat, and use of antihypertensive medications. To evaluate whether associations between physical activity and S_I were statistically independent of obesity and fat distribution, BMI and WHR were then added to the models. Finally, we evaluated whether associations between physical activity and S_I were comparable across various subgroups of the study sample by inclusion of the appropriate interaction term (1 at a time) for diabetes status, ethnicity, and sex. All analyses were conducted using the SAS statistical computing software.³¹

Participant characteristics are given in Table 2 for the full sample (N=1467) and for the subset of those who did not participate in vigorous activities (n=446). In the full sample, 2% of total time was spent in vigorous activities, and 7% was spent in moderate activities. In the subset, 6% of time was spent in moderate activities. For descriptive purposes, Figure 1 shows unadjusted average values of S_I according to level of physical activity for all participants. Average S_I (untransformed, min⁻¹·µU⁻¹·mL⁻¹·10⁻⁴) for individuals who reported rarely or never participating in vigorous activities was 1.14 (SD, 1.35), and, for those who reported participation in vigorous activities 5 or more times per week, the average S_I was 2.40 (SD, 2.55). This pattern of higher S_I among participants with higher levels of physical activity was consistent for the 1-year EEE in total, vigorous, and nonvigorous activities.

As shown in Table 3, after adjustment by regression analysis for potential confounders (age, sex, ethnicity,clinical center, percentage of caloric intake as dietary fat, alcohol intake, smoking status, and presence of hypertension), frequency of participation in vigorous activities was positively associated with S_I (S_I of 0.90 min⁻¹·µU⁻¹·mL⁻¹·10⁻⁴ for "rarely or never" with a 95% confidence interval [CI] of 0.83-0.97, compared with an S_I of 1.59 for "5+ times per week" with a 95% CI of 1.39-1.79; overall P<.001). In addition, an inverse association was seen for participation in vigorous activity in relation to fasting insulin (Table 3, overall P<.001).

Pearson correlation coefficients for physical activity variables from the 1-year activity recall in relation to S_I and fasting insulin are given in Table 4. These data demonstrate positive associations between physical activity and S_I (for all participants, r=0.14 for total EEE and S_I; P<.001). However, in the full sample, only total EEE and vigorous EEE were statistically significantly associated with fasting insulin; nonvigorous EEE was not associated with fasting insulin. In the nonvigorous subset, total EEE (incorporating only nonvigorous activity by definition) was significantly associated with S_I (r=0.13, P<.01) but not with fasting insulin.

Regression model results are presented in terms of predicted change in S_I or fasting insulin for an increase in physical activity of 836.8 kJ/d (200 kcal/d), based on current CDC and ACSM recommendations, for a hypothetical 70-kg individual, adjusted for demographic and behavioral variables (Table 5). Prior to adjustment for BMI and WHR, the magnitude of association was similar for total EEE (model 1) and for the independent effects of vigorous and nonvigorous EEE (model 2). Inclusion of BMI and WHR in the models attenuated the association in each case but did not entirely account for the statistically significant predicted increase in S_I with higher EEE. In model 1, the increase of 836.8 kJ/d (200 kcal/d) in total EEE was associated with a 1.87% increase in S_I (95% CI, 0.88-2.87), after accounting for BMI and WHR. As a point of reference, from the same model, a 1-unit decrement in BMI was associated with a 3.2% higher S_I. Findings were similar for vigorous and nonvigorous EEE.

Similar results were obtained in terms of presumed improvement in S_I (ie, lower fasting insulin level) with increased physical activity (Table 5). However, the association of nonvigorous activity with fasting insulin (model 2) failed to reach statistical significance even prior to inclusion of BMI and WHR.

Table 5 shows results restricted to the subset of participants who reported no vigorous activity (model 3); hence, the variation in EEE for this subset was restricted to energy expended across the restricted range of intensity from sleep to moderate activity (METs for all reported activities <6). Findings in this subset confirm results from the full sample in that higher levels of physical activity in the low- to moderate-intensity range were associated with higher S_I as measured directly by the FSIGT (P<.01), although inclusion of BMI and WHR attenuated the association (P=.14). As in the full sample, nonvigorous activity was associated with fasting insulin in the hypothesized direction but failed to reach statistical significance even prior to inclusion of BMI and WHR.

Stratified analyses were conducted for subgroups of diabetes status, ethnicity, and sex (Table 6). Within the subgroups, results were generally similar to those for the full sample, with a positive association observed between activity and S_I. Of the 9 interaction terms used to test whether estimates of association between activity and S_I were different across the subgroups, none were statistically significant.

Increased participation in nonvigorous as well as overall and vigorous physical activity was associated with higher S_I in a large, culturally and ethnically diverse sample of men and women, including individuals with normal glucose tolerance, IGT, and mild NIDDM. Overall obesity and fat distribution appeared to mediate some, but not all, of the observed association.

From Table 5, the estimated magnitude of effect of isocaloric EEE on S_I was remarkably similar for total, vigorous, and nonvigorous activities. Inclusion of EEE in vigorous and nonvigorous activities in the same model and consideration of their independent effects on S_I were justified by the relative lack of colinearity between these 2 variables (r=−0.18). Results from the subgroup of individuals who reported no participation in vigorous activity provided further confirmation of the relation of greater participation in nonvigorous activities with higher S_I.

Energy expenditure (both vigorous and nonvigorous) represents the cumulation of complex behaviors. It is assumed that error in the measurement of EEE is random with respect to the outcome variable, S_I. Such error would be expected to result in underestimation of the true association between EEE and S_I.

Recently, it was shown that walking was an effective adjunct to diet therapy in reducing weight and improving S_I among obese patients with NIDDM.³² In contrast, a study of 12 obese men¹⁰ (6 with NIDDM and 6 without diabetes) demonstrated that a 7-day exercise program at 70% maximum oxygen consumption (O₂max) resulted in improved S_I, but an isocaloric, 7-day program at 50% O₂max was not effective. The authors suggested that vigorous activity may confer a greater improvement in S_I than nonvigorous activity because there is greater use of muscle glycogen as an energy substrate during vigorous activity than during mild- or moderate-intensity activity. In addition, exercise training has recently been shown to increase insulin-stimulated glycogen synthesis in muscle.⁸ On the other hand, there is markedly enhanced fat oxidation (whether from circulating free fatty acids or muscle triglyceride) during activity of any intensity,³³ as well as increased translocation of muscle glucose transporter (GLUT-4), which is stimulated by contracting muscle³⁴ and has been shown to increase with exercise of between 50% and 75% of maximal capacity.¹² These phenomena, along with increased delivery of insulin to active muscle related to increased blood flow, suggest at least the potential for activity of any intensity to affect favorably S_I. It is possible that the observation of essentially equivalent effects of vigorous and nonvigorous activity on S_I in the present study relates to the reduced dependence of nonvigorous EEE on muscle glycogen, thereby allowing for longer duration of the activity without hypoglycemia or muscle discomfort.⁹

Some studies have shown that the effect of physical training on S_I may be transient.⁴ In the present data, habitual physical activity levels (vigorous and nonvigorous) were shown to be related to S_I. These findings are consistent in the sense that the level of ongoing physical activity, not just isolated bouts of activity, may be a key determinant of S_I in a free-living cohort.

For each type of physical activity, inclusion of BMI and WHR in the model (both variables associated with S_I at P<.001) attenuated but did not entirely account for the observed association of activity with S_I. Initially, expenditure of 836.8 kJ/d (200 kcal/d) was associated with a 2.68% higher S_I, after adjustment for potential confounders (Table 5). The addition of BMI to the model (not shown) reduced this effect size estimate to 1.97%. Finally, inclusion of both BMI and WHR (Table 5) yielded a further attenuation to 1.87%. This is consistent with the potential for multiple mechanisms. Although causal pathways cannot be determined by cross-sectional data analyses, results suggest the possibility of reduced overall obesity and reduced central deposition of adipose tissue as mediators of the beneficial effect of physical activity on S_I. In addition, S_I may also be improved with activity because of beneficial alterations in isocaloric fuel processing or other pathways.

Particularly for the association of nonvigorous activity with S_I, observed associations were generally stronger for the variable S_I derived from the FSIGT compared with the variable fasting insulin. This is not unexpected, given that S_I is a direct measure of insulin sensitivity, whereas fasting insulin is a surrogate measure that is known to be determined not only by S_I but also by insulin secretion and hepatic clearance of insulin.³⁵ Boyko et al³⁶ have demonstrated substantial confounding of S_I by insulin secretion when fasting insulin is used as a surrogate of S_I. In addition, studies have shown that the validity of fasting insulin as a surrogate for S_I worsens with increasing glucose intolerance³⁷; therefore, the validity of fasting insulin was presumably worse among the "nonvigorous" subset, since this group included a higher proportion of individuals with IGT or mild NIDDM (Table 2).

The potential impact of increased EEE (either vigorous or nonvigorous) on future incidence of NIDDM or coronary heart disease (via improvement in S_I) cannot be estimated directly from these cross-sectional data. However, Manson et al² demonstrated prospectively that the relative risk over 5 years for NIDDM incidence was 0.71 (P=.006) for participants in the Physician's Health Study who exercised vigorously at least once per week compared with those who exercised less frequently, after adjustment for BMI. Because the same question used in the Physician's Health Study was used in the present study and was very strongly associated with S_I (Table 3) and because the magnitude of the association between EEE and S_I was comparable for vigorous and nonvigorous activities (Table 5), it is not unreasonable to speculate that regular participation in either vigorous or nonvigorous activity would result in a clinically meaningful improvement in S_I with a consequent reduction in disease risk.

Sedentary living is extremely common among US adults. Major national surveys yield estimates between 22% and 30% of adults who report no participation in leisure-time physical activity (including walking).^38- 39 However, within various groups of individuals at high risk for diabetes, including those who are overweight, older, African American, or Hispanic, walking was reported as the most prevalent activity, with 28% to 45% reporting walking at least occasionally.³⁸ Similarly, walking was a preferred activity among individuals with diabetes.⁴⁰ Therefore, if nonvigorous activity is confirmed to have beneficial effects on S_I, the potential for prevention of related chronic diseases, including NIDDM and cardiovascular disease, may be considerable because interventions could be designed that incorporate an already common and presumably acceptable behavior that is inexpensive and accessible to large segments of the population. Initially, it will be necessary to firmly quantitate the potential benefit of nonvigorous activity on S_I. This will require prospective data, both from observational, community studies and from clinical trials. In the meantime, the findings from the present cross-sectional study lend further support for the current recommendations of the CDC and ACSM encouraging all US adults to participate in at least 30 minutes of moderate-intensity physical activity on most days of the week.