Make Your Diabetic Patients Walk: Long-term impact of different amounts of physical activity on type 2 diabetes

Eligibility criteria included diagnosis of type 2 diabetes of at least 2 years and age >40 years before recruitment. Patients who had illnesses that could seriously reduce life expectancy or cause cardiac, liver, or renal failure were excluded. All eligible patients consecutively attending our outpatient diabetes clinic over a period of 3 months (October 1999 to January 2000) were recruited. After obtaining written informed consent, 182 patients (age 62 ± 0.7 years, 88 men and women, diabetes duration 7.6 ± 0.3 years) were randomized to the study, which closed on December 2002. The details of the intervention have been reported elsewhere (9). Briefly, it consisted of a first counseling session of at least 30 min conducted by a physician and designed to advise on physical activity, followed 1 month later by a telephone call made by the same doctor and then by 15-min sessions every 3 months in the outpatient clinic for a total of seven maintenance visits; the duration of the study was 2 years. Patients were treated for diabetes and other diseases according to accepted international guidelines. The diet suggested to all patients contained 55% of calories from complex carbohydrates, 30% from fat, and 15% from protein. Overweight and obese patients were given a diet with a negative balance of ∼300 kcal/day, including in the calculation the energy expenditure due to voluntary physical activity.

Levels of voluntary physical activity were assessed every 3 months with the Modifiable Activity Questionnaire (14) and calculated as the product of the duration (hours times weeks) of the various activities weighted by an estimate of metabolic equivalent (MET) of each activity (5). The outcome of the intervention was the consistent achievement of an energy expenditure >10 METs · h⁻¹ · week⁻¹ through moderate intensity (3–6 METs) aerobic leisure time physical activity. Subjects were instructed to exercise at moderate intensity (40–60% of heart rate reserve) between 3 and 6 METs (15). On a daily basis, all patients were asked to record type, intensity, and duration of performed physical activities. Their written reports were examined and discussed with the physician every 3 months. At the end of the study, we collected the detailed reports of 104 weeks from each patient. Most subjects (94%) varied their weekly amounts of energy expenditure <10 METs · h⁻¹ · week⁻¹. If larger fluctuations in energy expenditure were reported, subjects were transferred into the next group with higher energy expenditure (see statistical analysis).

At each session, all subjects underwent measurements of body weight, BMI, waist circumference, resting heart rate, and blood pressure according to NHANES procedures (16). Plasma glucose and HbA_1c were measured as previ-ously described (9), serum total cholesterol and triglycerides were measured using a colorimetric assay (Menarini Diagnostics, Florence, Italy), serum HDL cholesterol was measured using a direct enzymatic method (Menarini Diagnostics, Florence, Italy), and LDL cholesterol was calculated using the Friedewald formula (17). Coronary heart disease (CHD) risk was estimated using Framingham risk tables (18).

Costs were generally calculated using standard unit costs of the resources used, averaged on the year 2000 costs of resources in Italy (tables produced by Italian National Institute of Statistics and by Umbria region). Cost analysis included direct medical costs and direct and indirect social costs and is reported in U.S. dollars per capita per year, adjusted to year 2000. Direct medical costs include expenses for medications and other costs usually paid by the National Health Service (NHS), e.g., counseling intervention (hourly rate of physicians $49), laboratory testing, hospitalization, and outpatient care. Direct social costs include the value of the time that participants spent in practicing physical activity, cost of related items (shoes, fitness equipment, etc.), transportation to exercise places, and admission to health clubs. Indirect social costs include time that participants reported as lost from work or usual activities as a result of counseling visits, illness, or injury; each day lost to morbidity was valued at $100. Sensitive analysis was performed to estimate the economic impact of a less effective physical activity counseling. For this purpose we assumed that instead of 125 of 179 (69%), only 90 of 179 (50%) or 60 of 179 (33%) diabetic subjects achieved the target of the intervention (>10 METs · h⁻¹ · week⁻¹). The 35 and the 65 subjects not reaching the target in the 50 and 33% simulation were equally distributed between groups 0 and 1–10 and equally subtracted from the other four groups.

Data obtained over the last year before study entry (basal) and the last year of intervention were analyzed using NCSS 2004 (Kaysville, UT). The primary end point of the study was the estimated 10-year CHD risk. Secondary end points were body weight, BMI, waist circumference, fasting plasma glucose (FPG), HbA_1c, systolic and diastolic blood pressure, heart rate, total cholesterol, LDL and HDL cholesterol, and triglycerides. Additional end points were medical and social costs. Results are presented as the means ± SE and 95% CI. Analysis of change from baseline values for each continuous variable were analyzed using paired samples t tests. The average rate of change for each variable was calculated using linear regressions according to the least-squares method. To assess the impact of different amounts of energy expenditure, a post hoc analysis was performed on the 179 subjects, dividing them into six groups. Ten METs · h⁻¹ · week⁻¹ of energy expenditure was chosen as criterion to form the groups because this amount of energy expenditure was the target of the intervention (9) and roughly corresponds to the recommendations of several scientific societies (1–4,12). The 179 subjects were divided into the six following groups: group 0 (no activity, n = 28), group 1–10 (6.8 ± 0.3 METs · h⁻¹ · week⁻¹, n = 27), group 11–20 (17.1 ± 0.4, n = 31), group 21–30 (27.0 ± 0.5, n = 27), group 31–40 (37.5 ± 0.5, n = 32), and group >40 (58.3 ± 1.8, n = 34), including all the subjects who achieved change in energy expenditure >40 METs · h⁻¹ · week⁻¹. Analysis was conducted using ANOVA with post hoc analysis (Tukey’s test), and proportions were analyzed using the McNemar test. A P value <0.05 indicated significant differences. To maintain an overall type I error rate of 5% (α = 0.05), P values were corrected according to the modified Bonferroni procedure proposed by Finner (19), which is more powerful than the traditional Bonferroni approach for multiple test adjustments.

Originally, 182 patients were randomized to the intervention. Three patients did not complete the study. One dropped out of the follow-up, and two died of causes unrelated to physical activity. Data analysis refers to 179 patients who were divided into six groups based on their change in increments in energy expenditure (METs per hour per week): group 0 (no activity, n = 28), group 1–10 (6.8 ± 0.3, n = 27), group 11–20 (17.1 ± 0.4, n = 31), group 21–30 (27.0 ± 0.5, n = 27), group 31–40 (37.5 ± 0.5, n = 32), and group >40 (58.3 ± 1.8, n = 34). Group 0 and group 1–10 (31% of the entire population) did not reach the study target (constant energy expenditure >10 METs · h⁻¹ · week⁻¹), which was achieved by the subjects in the other four groups (69% of the entire population).

Age, diabetes duration, male-to-female ratio, and baseline levels of energy expenditure through voluntary physical activity did not differ in the six groups (Table 1). The intervention increased energy expenditure through voluntary physical activity by 26 ± 2 METs · h⁻¹ · week⁻¹ (P < 0.0001) in the whole group (Table 2). As referred during the regular (every 3 months) ambulatory visits, the different amounts of physical activity practiced by the subjects in the six groups was not conditioned by associated diseases but was the result of only the subjects’ free choice to use exercise to improve their state of health.

There was no significant change in any of the parameters, except for a significant increase (P < 0.01) in medical costs (drugs) and indirect social costs for a total of $828 · capita⁻¹ · year⁻¹ after 2 years’ follow-up (Tables 1 and 3).

There was no significant change in any of the health parameters examined and no significant change in medical or indirect social costs after 2 years (Tables 1 and 3).

There were significant reductions in HbA_1c (P < 0.0001), systolic (P = 0.0286) and diastolic (P < 0.003) blood pressure, total cholesterol (P < 0.0001), triglycerides (P < 0.0001), and 10-year CHD risk (P = 0.0003). These results were associated with significant drops in medical (P = 0.0159) and social costs (P = 0.011), for a total saving (P < 0.0001) of $386 · capita⁻¹ · year⁻¹ after 2 years’ follow-up (Tables 1 and 3).

There were significant (P < 0.0001) reductions in body weight, BMI, waist circumference, FPG, HbA_1c, systolic (P = 0.048) and diastolic (0.0156) blood pressure, heart rate, total and LDL (P = 0.0229) cholesterol, and triglycerides; an ∼4% reduction in 10-year CHD risk; and a significant (P < 0.0001) increase in HDL cholesterol. These results were associated with significant (P < 0.0001) decreases in medical and social costs for a total saving of ∼$1,452 · capita⁻¹ · year⁻¹ after 2 years’ follow-up (Tables 1 and 3).

There were significant (P < 0.0001) reductions in body weight, BMI, waist circumference, FPG, HbA_1c, systolic and diastolic blood pressure, heart rate, total and LDL (P = 0.012) cholesterol, and triglycerides; a 4–5% reduction in 10-year CHD risk; and a significant (P < 0.0001) increase in HDL cholesterol. These results were associated with significant (P < 0.0001) drops in medical and social costs, for a total saving of ∼$2,000 · capita⁻¹ · year⁻¹ after 2 years’ follow-up (Tables 1 and 3).

There were significant (P < 0.0001) reductions in body weight, BMI, waist circumference, FPG, HbA_1c, systolic and diastolic blood pressure, heart rate, total and LDL cholesterol, and triglycerides; an ∼3% reduction in 10-year CHD risk; and a significant (P < 0.0001) increase in HDL cholesterol. These results were associated with significant (P < 0.0001) drops in medical and social costs, for a total saving of $855 · capita⁻¹ · year⁻¹ after 2 years’ follow-up.

Amounts of energy expenditure (METs per hour per week) were inversely related (P < 0.0001) with changes in body weight (r = 0.62), BMI (r = 0.62), waist circumference (r = 0.77), FPG (r = 0.52), HbA_1c (r = 0.70), systolic (r = 0.33) and diastolic (r = 0.23, P = 0.023) blood pressure, resting heart rate (r = 0.76), triglycerides (r = 0.56), and percent 10-year CHD risk (r = 0.39) and positively related with changes in HDL cholesterol (r = 0.39, P < 0.0001). Accordingly, on average, for every increase of 1 MET · hour⁻¹ · week⁻¹, body weight decreased by 0.08 kg, BMI by 0.03 kg/m², waist circumference by 0.16 cm, FPG by 0.03 mmol/l, HbA_1c by 0.02%, systolic blood pressure by 0.14 mmHg, diastolic blood pressure by 0.08 mmHg, blood pressure resting heart rate by 0.16 bpm, serum triglycerides by 0.014 mmol/l, and percent 10-year CHD risk by 0.08%; HDL increased by 0.004 mmol/l.

Costs were significantly (P < 0.0001) related to energy expenditure. METs per hour per week were inversely related with medical prescription costs (r = 0.51, −$18), other medical costs (r = 0.33, −$23), indirect social costs (r = 0.40, −$36), and total costs (r = 0.60, −$66) and positively related with direct social costs (r = 0.44, $13). In 2 years, a 3-mile daily walk is estimated to reduce drug costs by $550, other medical costs by $700, indirect social costs by $1,100, and total costs by $2,000 and to increase direct social costs by $400. Reduction in costs for insulin had a major impact on medication costs. In fact, after 2 years, the number of subjects on insulin therapy fell by 25% (before 59 of 179, after 44 of 179, P = 0.0003). There was a significant (P < 0.0001) inverse correlation between METs per hour per week and daily units of insulin (r = 0.38, −0.35 IU/day), suggesting that over 2 years, a 3-mile daily walk reduces insulin dosage by ∼11 units/day.

The 50% compliance model (90 of 179 achieving >10 METs · h⁻¹ · week⁻¹) predicted significant reductions in medical and social costs with expected total savings of ∼$500 · capita⁻¹ · year⁻¹ after 2 years. The 33% compliance model (60 of 179 achieving the target) predicted no significant change in medical and social costs after 2 years (Table 3).

This study demonstrates that physical activity is an effective cost-saving tool in the care of type 2 diabetes. Two years’ individual exercise counseling of a randomized population of type 2 diabetic subjects provided valuable health benefits and financial advantages for the NHS and the individual patient. These results confirm that the advice of several scientific societies (1–4,12) recommending ≥30 min moderate-intensity physical activity (>10 METs · h⁻¹ · week⁻¹) on most days, if not every day, is also valid for type 2 diabetic subjects and demonstrate a significant dose-response relationship.

Post hoc analysis showed that energy expenditure must be >10 METs · h⁻¹ · week⁻¹ for significant beneficial effects. In fact, energy expenditure ranging between 11 and 20 METs · h⁻¹ · week⁻¹ significantly reduced HbA_1c, total cholesterol, triglycerides, and blood pressure, with a 2.6 ± 0.6% reduction of 10-year CHD risk. These benefits occurred in the absence of any significant weight loss, suggesting that regular aerobic physical activity improves glucose control, lipid profile, and blood pressure independently of weight reduction. This conclusion concurs with results of a meta-analysis examining the effects of exercise on glucose control in type 2 diabetes (8) and on blood pressure in general population (20). However, from our data, it is not possible to rule out that changes in body composition, like a selective loss of visceral fat, contributed to the observed effects. Regarding the effects of exercise on body weight, our results, based on anthropometric measurements (body weight, waist circumference), confirm that visceral fat is a very sensitive target of physical activity (waist circumference vs. energy expenditure: r = −0.77) and that to induce long-term losses of body weight, increments in energy expenditure >20 METs · h⁻¹ · week⁻¹ are required (1,12,13).

Constant energy expenditure >20 METs · h⁻¹ · week⁻¹ is needed to decrease body weight, BMI, waist circumference, heart rate, and LDL cholesterol and to augment HDL cholesterol. This amount of energy expenditure induces greater reductions in HbA_1c, total cholesterol, triglycerides, and blood pressure, leading to an ∼4–5% decrease in the 10-year CHD risk. The results in our type 2 diabetic subjects confirm reports of prospective studies in the general population that show an inverse linear dose response between amount of physical activity and all-cause mortality, total cardiovascular disease, and CHD incidence and mortality (1–4,21). This dose-response relationship demonstrates that full beneficial outcomes of exercise are obtained by the most physically active diabetic subjects. Differences in caloric intake among groups might have contributed to the final results. To control for differences in energy balance, all overweight and obese patients were given a diet with a negative balance of 300 kcal/day, including in the calculation the energy expenditure due to voluntary physical activity (9). Thus, the fact that despite greater caloric intakes, subjects of groups with increased energy expenditure had greater beneficial effects emphasizes the effectiveness of regular exercise in the management of type 2 diabetes.

As has been shown in the general population (2), the health benefits of exercise tend to become less evident in patients when the increase in energy expenditure is beyond a certain level, as was indicated by benefit analysis in groups with >20 METs · h⁻¹ · week⁻¹ (Table 1). Since the threshold for full benefits was observed in group 21–30 (average energy expenditure 27 METs · h⁻¹ · week⁻¹), we recommend a target of 27 METs · h⁻¹ · week⁻¹ as a reasonable target of energy expenditure for previously sedentary type 2 diabetic subjects. This goal corresponds to a 3-mile (∼5-km) daily walk (1 h/day at a pace of 3 mph or 45 min/day at a pace of 4 mph), which, according to our results, might be expected over 2 years to reduce body weight by 2.4 kg, BMI by 0.9 kg/m², waist circumference by 4.8 cm, FPG by 0.9 mmol/l, HbA_1c by 1.5%, systolic and diastolic blood pressure by 10 and 7 mmHg, respectively, resting heart rate by 5 bpm, triglycerides by 0.4 mmol/l, and 10-year CHD risk by 2.4% and to augment HDL cholesterol by 0.12 mmol/l. However, it must be emphasized that greater amounts of energy expenditure resulted in added improvements in anthropometric and biochemical markers of the metabolic syndrome, as is clearly shown by the group comparisons reported in Table 1. Subjects in groups with energy expenditure of 31–40 and/or >40 METs · h⁻¹ · week⁻¹ had significantly greater beneficial effects than the subjects in groups with energy expenditure <20 METs · h⁻¹ · week⁻¹. Brisk walking was the most common form of leisure time physical activity practiced by our type 2 diabetic subjects; however, it cannot always be recommended. To subjects affected by peripheral neuropathy or lower-limb complications, we suggested exercises with reduced physical impact like stationary cycling, water gym, or swimming.

Cost analysis indicates remarkable financial savings (Table 3). Over 2 years, prescription costs, which were usually medication for diabetes, hypertension, and dyslipidemia, were reduced by $259 and other health care costs by $298 · capita⁻¹ · year⁻¹ in the entire group, even allowing for the costs of counseling, which took a physician 75 min over the first year and 60 min over the second. Our physical activity counseling performed in type 2 diabetic subjects appears more convenient than the approach used for lifestyle intervention in the Diabetes Prevention Program in subjects with impaired glucose tolerance (5). However, comparison of medical and social costs must be done with caution considering that in the U.S., the cost of medical care is generally higher than in Italy.

Our physical activity counseling model is simple and reproducible (9,10) but requires physician training in the use of the social cognitive approach, which adds costs that were not included in the present calculations. In our study, counseling was conducted by physically active physicians who were highly motivated and obtained a 69% compliance rate after 2 years (9). Lower rates of long-term compliance might have been achieved in other settings. For this reason, Table 3 reports a sensitive analysis simulating compliance rates of 50 and 33%. Interestingly, physical activity counseling remains financially advantageous, even when only one of two or three diabetic subjects reaches the target of >10 METs · h⁻¹ · week⁻¹. In fact, although the 33% compliance rate model does not significantly reduce costs versus baseline, it remains cost saving compared with sedentary diabetic subjects whose health costs increased over 2 years by $516 · capita⁻¹ · year⁻¹. Again, compared with this latter group, even small amounts of energy expenditure (1–10 METs · h⁻¹ · week⁻¹) save money as shown by the lack of increase over 2 years in costs for medical prescriptions in group 1–10 (Table 3). Furthermore, savings with our cost analysis, limited as they are to a 2-year time frame, are largely underestimated because long-term benefits related to the significant reduction in CHD risk (22) are not taken into account.

Post hoc subgroup cost analysis shows that as with the health benefits, the greatest financial benefits are achieved by groups with energy expenditure >20 METs · h⁻¹ · week⁻¹ (average energy expenditure: 27 METs · h⁻¹ · week⁻¹). Money saving tends to become less evident when the increase in energy expenditure is beyond a certain level. In subjects with highest energy expenditure (group >40), savings due to lower medical costs and indirect social costs are partially counterbalanced by the increased costs of physical activity, i.e., exercise apparel, footwear, and gym costs. Thus, also from the financial point of view, recommending energy expenditure amounts of 27 METs · h⁻¹ · week⁻¹ appears to be an appropriate target for previously sedentary type 2 diabetic subjects.

In conclusion, the present study confirms the long-term efficacy of a program of physical activity counseling (9–11), confirms the beneficial effects of exercise on cardiovascular risk factors (1–4,11), and demonstrates that health benefits are associated with remarkable savings in the NHS care of type 2 diabetes. The present data strongly support the need for, and convenience of, instituting physical activity programs as an essential part of therapy for type 2 diabetes. Our diabetic subjects practiced only aerobic activities at moderate intensity, and most of them chose fitness walking. It remains to be established whether exercise at higher intensities (>6 METs) and mixed aerobic/anaerobic exercise might have provided additional benefits. In addition, future studies are required to establish if type 2 diabetic subjects have individual differences in metabolic responses to regular exercise and the extent to which genetic heterogeneity might affect the variability in training response (23,24).