CI indicates confidence interval. Changes in hemoglobin A1c (HbA1c ) (absolute values) of individual studies included in the meta-analysis of structured exercise training (aerobic exercise, resistance training, and combined aerobic/resistance exercise) vs no intervention in patients with type 2 diabetes. Studies that included more than 1 modality or different training protocols within a same type of structured exercise training were evaluated as separate observations. Weights are from random-effects analysis.aExercise and control subgroups.bExercise and control subgroups with rosiglitazone treatment as cointervention.cSubgroup with exercise frequency of 3 sessions per week.dSubgroup with exercise frequency of 5 sessions per week.
CI indicates confidence interval. Changes in hemoglobin A1c(HbA1c ) for individual studies included in the meta-analysis of physical activity advice vs no intervention in patients with type 2 diabetes according to the association or not of dietary intervention. Two studies provided more than 1 observation and were analyzed as distinct interventions to deliver physical activity. Weights are from random-effects analysis.aSubgroup received advice in printed material.bSubgroup received advice through a Web system.cSubgroup received advice from an individual.dSubgroup received advice in written form.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Umpierre D, Ribeiro PAB, Kramer CK, et al. Physical Activity Advice Only or Structured Exercise Training and Association With HbA1c Levels in Type 2 Diabetes: A Systematic Review and Meta-analysis. JAMA. 2011;305(17):1790–1799. doi:10.1001/jama.2011.576
Author Affiliations: Exercise Pathophysiology Research Laboratory (Mr Umpierre, Ms Paula A. B. Ribeiro, and Dr Jorge P. Ribeiro), Endocrinology Division (Drs Kramer, Leitão, Azevedo, Gross, and Schaan and Ms Zucatti), and Cardiology Division (Dr Jorge P. Ribeiro), Hospital de Clínicas de Porto Alegre, Porto Alegre, Brazil; and Department of Internal Medicine, Faculty of Medicine, Federal University of Rio Grande do Sul, Porto Alegre, Brazil (Drs Azevedo, Gross, Jorge P. Ribeiro, and Schaan).
Context Regular exercise improves glucose control in diabetes, but the association of different exercise training interventions on glucose control is unclear.
Objective To conduct a systematic review and meta-analysis of randomized controlled clinical trials (RCTs) assessing associations of structured exercise training regimens (aerobic, resistance, or both) and physical activity advice with or without dietary cointervention on change in hemoglobin A1c (HbA1c) in type 2 diabetes patients.
Data Sources MEDLINE, Cochrane-CENTRAL, EMBASE, ClinicalTrials.gov, LILACS, and SPORTDiscus databases were searched from January 1980 through February 2011.
Study Selection RCTs of at least 12 weeks' duration that evaluated the ability of structured exercise training or physical activity advice to lower HbA1c levels as compared with a control group in patients with type 2 diabetes.
Data Extraction Two independent reviewers extracted data and assessed quality of the included studies.
Data Synthesis Of 4191 articles retrieved, 47 RCTs (8538 patients) were included. Pooled mean differences in HbA1c levels between intervention and control groups were calculated using a random-effects model. Overall, structured exercise training (23 studies) was associated with a decline in HbA1c level (−0.67%; 95% confidence interval [CI], −0.84% to −0.49%; I2, 91.3%) compared with control participants. In addition, structured aerobic exercise (−0.73%; 95% CI, −1.06% to −0.40%; I2, 92.8%), structured resistance training (−0.57%; 95% CI, −1.14% to −0.01%; I2, 92.5%), and both combined (−0.51%; 95% CI, −0.79% to −0.23%; I2, 67.5%) were each associated with declines in HbA1C levels compared with control participants. Structured exercise durations of more than 150 minutes per week were associated with HbA1c reductions of 0.89%, while structured exercise durations of 150 minutes or less per week were associated with HbA1C reductions of 0.36%. Overall, interventions of physical activity advice (24 studies) were associated with lower HbA1c levels (−0.43%; 95% CI, −0.59% to −0.28%; I2, 62.9%) compared with control participants. Combined physical activity advice and dietary advice was associated with decreased HbA1c (−0.58%; 95% CI, −0.74% to −0.43%; I2, 57.5%) as compared with control participants. Physical activity advice alone was not associated with HbA1c changes.
Conclusions Structured exercise training that consists of aerobic exercise, resistance training, or both combined is associated with HbA1c reduction in patients with type 2 diabetes. Structured exercise training of more than 150 minutes per week is associated with greater HbA1c declines than that of 150 minutes or less per week. Physical activity advice is associated with lower HbA1c, but only when combined with dietary advice.
Exercise is a cornerstone of diabetes management, along with dietary and pharmacological interventions.1,2 Current guidelines recommend that patients with type 2 diabetes should perform at least 150 minutes per week of moderate-intensity aerobic exercise and should perform resistance exercise 3 times per week.1,2 Previous meta-analyses3-6 demonstrated that structured exercise training including aerobic and resistance exercises reduces hemoglobin A1c (HbA1c) levels by approximately 0.6%. However, only 1 previous review separately analyzed associations of aerobic exercise, resistance training, and the combination of aerobic exercise and resistance training on change in HbA1c levels.5 Since publication of this meta-analysis, 2 large randomized trials7,8 were published that reported contradictory findings regarding the types of structured exercise associated with declines in HbA1c levels. Sigal et al7 found that aerobic or resistance exercise training alone improved glycemic control but the effects were more pronounced with both combined. In contrast, Church et al8 observed that only the combination, but not aerobic and resistance training alone, reduced HbA1c levels.
In contrast to structured exercise training, physical activity is defined as any bodily movement produced by skeletal muscle contractions resulting in increased energy expenditure.9 Although structured exercise training may be available to a subset of patients with type 2 diabetes, physical activity advice is more feasible and should be offered to most patients with type 2 diabetes. However, meta-analyses have not been performed to determine whether physical activity advice is associated with similar declines in HbA1c as compared with those associated with structured exercise. This study consists of a systematic review with meta-analysis of randomized controlled clinical trials (RCTs) on the associations of structured exercise training and physical activity advice, respectively, on changes in HbA1c levels in patients with type 2 diabetes. Structured exercise training is categorized according to whether it consists of aerobic exercise, resistance training, or a combination of both.
We searched the following electronic databases covering the period from January 1980 through February 2011: MEDLINE (accessed by PubMed), Cochrane Central Register of Controlled Trials, EMBASE, ClinicalTrials.gov, SPORTdiscus, and LILACS. In addition, we searched the references of published studies manually. The initial search comprised the terms exercise, diabetes mellitus, physical activity, and related entry terms associated with a high-sensitivity strategy for the search of RCTs,10 and was not limited by language. The complete search strategy used for the PubMed database is
shown in eBox 1. Only eligible full texts in English, Portuguese, or Spanish were considered for review. This systematic review and meta-analysis is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.11
We included RCTs that compared any category of structured exercise training (aerobic, resistance, or a combination of both) or physical activity advice with a control group of patients with type 2 diabetes older than 18 years, that evaluated HbA1c as an outcome, and reported means or differences between means and respective dispersion values of HbA1c at baseline and after the intervention. Structured exercise training was defined as an intervention in which patients were engaged in planned, individualized, and supervised exercise programs. Physical activity advice was defined as an intervention in which patients were partially or not engaged in supervised exercise training, but received formal instructions to exercise regularly with or without an individualized exercise prescription. Eligible studies included only individuals able to exercise, with no clinical manifestations limiting physical activity. Exclusion criteria are as follows: (1) studies of patients with type 1 diabetes or gestational diabetes; (2) RCTs that did not provide information regarding the associations of the intervention with HbA1c in the experimental group, the control group, or both; (3) duplicate publications or substudies of included trials; and (4) studies with less than 12 weeks of follow-up.
Titles and abstracts of retrieved articles were independently evaluated by 2 investigators (D.U. and P.A.B.R.). Reviewers were not blinded to authors, institutions, or manuscript journals. Abstracts that did not provide enough information regarding the inclusion and exclusion criteria were retrieved for full-text evaluation. Reviewers independently evaluated full-text articles and determined study eligibility. Disagreements were solved by consensus and if disagreement persisted, by a third reviewer (B.D.S.). To avoid possible double counting of patients included in more than 1 report by the same authors or working groups, patient recruitment periods were evaluated and if necessary, authors were contacted for clarification. The corresponding author was contacted as needed to obtain data not included in the published report.
Two reviewers (D.U. and P.A.B.R.) independently conducted data extraction. Disagreements were solved by consensus or by a third reviewer (B.D.S.). Adherence to protocols, dropout rates, and adverse events were also extracted.
Risk of bias was evaluated according to the PRISMA recommendation.12 Study quality assessment included adequate sequence generation, allocation concealment, blinding of outcomes assessors, use of intention-to-treat analysis, and description of losses and exclusions. Studies without clear descriptions of an adequate sequence generation or how the allocation list was concealed were considered not to have fulfilled these criteria. Quality assessment was independently performed by 2 unblinded reviewers (D.U. and P.A.B.R) and disagreements were solved by consensus or by a third reviewer (B.D.S). The κ agreement rate between reviewers was κ = 0.96 for quality assessment.
Absolute changes in HbA1c were reported as differences between arithmetic means before and after interventions. Data from intention-to-treat analyses were entered whenever available in included RCTs.
Pooled-effect estimates were obtained by comparing the least squares mean percentage change from baseline to the end of the study for each group, and were expressed as the weighted mean difference between groups. Calculations were performed using a random-effects model. Four comparisons were made with each group being compared with a no-intervention (control) group: structured aerobic exercise training, structured resistance exercise training, structured combined aerobic/resistance exercise training, and physical activity advice. An α value = .05 was considered statistically significant.
Statistical heterogeneity of the treatment effect among studies was assessed using Cochran Q test, a threshold P value of .1 was considered statistically significant, and the inconsistency I2 test in which values greater than 50% were considered indicative of high heterogeneity.13 We explored heterogeneity between studies using 3 strategies. First, we reran the meta-analyses removing each study at a time to check if a particular study was explaining heterogeneity. Second, stepwise meta-regression analyses were carried out. Using univariate meta-regression models, we assessed clinical and methodological variables that influenced the association of exercise with HbA1c levels. Likewise, similar procedures were undertaken to analyze particular variables that could explain heterogeneity in the physical activity advice meta-analysis. Thereafter, based on univariate meta-regression analyses, we constructed 4 multivariate models including baseline HbA1c plus exercise frequency (defined as the number of exercise sessions per week [model 1]); baseline HbA1c plus total exercise time spent in the program (defined as the cumulative product of exercise frequency, session duration, and number of weeks of training [model 2]); baseline HbA1c plus a variable indicating total exercise time of 150 minutes or less per week or more than 150 minutes per week [model 3]); and baseline HbA1c plus exercise intensity plus total exercise time spent in the program (model 4). Model 4 included covariates that were not significant in univariate regression, but were included based on clinical judgment of their importance. We evaluated the goodness of fit of each model using the adjusted R2, which denotes the proportion of between-study variation explained by the covariates.14,15 Third, we performed sensitivity analyses to evaluate subgroups of studies most likely to yield valid estimates of the intervention based on prespecified relevant clinical information and on meta-regression analyses. For the structured exercise training meta-analysis results, we used a cutoff of 150 minutes per week to stratify studies according to their weekly amounts of exercise. RCTs evaluating physical activity advice were grouped according to the presence vs absence of a simultaneous dietary recommendation.
Because some studies compared multiple exercise interventions with a single control group, we split this shared group into 2 or more groups with smaller sample sizes weighted in relation to different exercise interventions. This approach was applied in order to have reasonably independent comparisons and overcome a unit-of-analysis error for studies that could contribute to multiple and correlated comparisons, as suggested by the Cochrane Handbook for Systematic Reviews of Interventions.13 Imputation and/or transformation methods were used for few studies that showed results as confidence intervals (CIs) or interquartile ranges.16
Publication bias was assessed using a contour-enhanced funnel plot of each trial's effect size against the standard error.17 Funnel plot asymmetry was evaluated by Begg and Egger tests, and a significant publication bias was considered if the P value was less than .10. The trim-and-fill computation was used to estimate the effect of publication bias on the interpretation of results.18,19 All analyses were conducted using Stata software version 11.0 (Stata Inc, College Station, Texas).
From 4191 potentially relevant citations retrieved from electronic databases and searches of reference lists, 47 RCTs (including 23 RCTs of structured exercise training and 24 RCTs of physical activity advice) met the inclusion criteria. A flow diagram of search and selection is
shown in eFigure 1. Included studies had a total of 8538 patients. Of these, 848 patients were included in studies of structured aerobic exercise training, 261 in structured resistance exercise studies, 404 in structured combined aerobic/resistance exercise training studies, and 7025 in physical activity advice studies. Characteristics of these studies are summarized in Table 1 and Table 2.
Fifteen studies of structured exercise reported data on adherence. Of these, 14 trials reported adherence rates of more than 75%. Dropout rates were less than 20% in all but 2 of the 21 studies that reported this measure (Table 1). Adherence rates were not reported for the physical activity studies because of lack of accuracy (ie, self-reported data and reliance on patient recall). Dropout rates were less than 20% for 19 of the 24 physical activity intervention studies (Table 2).
No major adverse effects were reported (eTable 1). Minor adverse events for the structured exercise interventions and physical activity interventions most commonly included cardiovascular disease events that were not related to the intervention and musculoskeletal injury or
discomfort (eTable 1). One study of a physical activity intervention included a high rate of hypoglycemia. Of 47 RCTs, 30 studies did not report data on adverse
events (eTable 1).
Among the included studies, 36% presented adequate sequence generation (17 of 47), 17% reported allocation concealment (8 of 47), 17% had blinded assessment of outcomes (8 of 47), 96% described losses to follow-up and exclusions (45 of 47), and 13% used the intention-to-treat principle for statistical analyses (6 of 47) (eTable 2
and eTable 3).
Contour-enhanced funnel plots and the Egger regression test suggested an asymmetry in the analysis of structured exercise training (P = .02). However, the trim-and-fill computation revealed that publication bias did not interfere with the interpretation of results (eFigure 2, panel A). Regarding physical activity advice studies, neither the Egger regression test nor the trim-and-fill computation showed any publication
bias (P = .10) (eFigure 2, panel B).
Structured Exercise Training: Aerobic, Resistance, or Both. The overall association of any structured exercise vs control with absolute HbA1c reduction (23 studies; 1533 patients) was −0.67% (95% CI; −0.84% to −0.49%; I2, 91.3%; P for heterogeneity, <.001) (Figure 1). Eighteen studies (848 patients) demonstrated that structured aerobic exercise training was associated with an absolute HbA1c reduction of 0.73% (95% CI, −1.06% to −0.40%; I2, 92.8%; P for heterogeneity <.001) as compared with control.
Four articles (261 patients) demonstrated that structured resistance exercise training was associated with a decline in absolute HbA1c of 0.57% (95% CI, −1.14% to −0.01%; I2, 92.5%; P for heterogeneity <.001) as compared with control.
Seven articles (404 patients) demonstrated that the combination of aerobic and resistance exercise were associated with an HbA1c reduction of 0.51% (95% CI, −0.79% to −0.23%; I2, 67.5%; P for heterogeneity <.001) as compared with control participants.
In univariate meta-regression, baseline HbA1c level, exercise frequency, total time spent in exercise during the study, and weekly exercise duration of more than 150 minutes per week or of 150 minutes or less per week partially explained heterogeneity between structured exercise training studies (eTable 4). These covariates also were significant in the multivariate meta-regression models (eTable 4). Structured exercise of more than 150 minutes per week (18 observations, 826 patients) was associated with an absolute HbA1c reduction of 0.89% (95% CI, −1.26% to −0.51%; I2, 91.4%; P for heterogeneity <.001). Structured exercise of 150 minutes or less per week (12 observations, 687 patients) was associated with an absolute reduction of 0.36% of HbA1c (95% CI, −0.50% to −0.23%; I2, 78.6%; P for
heterogeneity <.001) (eFigure 3). When studies were omitted individually from the meta-analysis to assess possible individual influences on results, heterogeneity and weighted mean differences were unchanged.
Physical Activity Advice. Twenty-four articles comparing physical activity advice (3529 patients) vs control (3496 patients) demonstrated that physical activity advice was associated with a decline in HbA1c of 0.43% (95% CI, −0.59% to −0.28%; I2, 62.9%; P for heterogeneity <.001) (Figure 2).
Covariates used in univariate analysis did not explain
heterogeneity (eTable 4). Similarly, a multivariate meta-regression using baseline HbA1c and dietary recommendation (model 1) as covariates did not explain the between-studies variance (overall, P = .17).
In sensitivity analyses, physical activity associated with dietary advice (12 studies, 6313 patients) was associated with a 0.58% absolute HbA1c reduction (95% CI, −0.74% to −0.43%; I2, 57.5%; P for heterogeneity = .007) as compared with control. Physical activity advice alone (14 studies, 712 patients) was not associated with HbA1c changes as compared with control (Figure 2). When studies were individually omitted from the meta-analysis, heterogeneity and weighted mean differences were unchanged.
Quiz Ref IDOur results demonstrate that in patients with type 2 diabetes, structured aerobic, resistance, or combined exercise training is associated with a HbA1c decline of −0.67%.Quiz Ref IDOur analyses also demonstrate that structured exercise duration of more than 150 minutes per week was associated with greater benefit (0.89% reduction in HbA1c ) than structured exercise duration of 150 minutes or less per week (0.36% reduction in HbA1c). Structured exercise training was associated with a more pronounced HbA1c reduction compared with physical activity advice. Quiz Ref IDA recommendation to increase physical activity was beneficial (0.43% HbA1c reduction), but only if combined with dietary recommendations.
This systematic review and meta-analysis of RCTs demonstrates important findings regarding the prescription of structured exercise training. First, aerobic, resistance, and combined training are each associated with HbA1c decreases, and the magnitude of this reduction is similar across the 3 exercise modalities. Quiz Ref IDInterestingly, the weighted mean difference of −0.67% in HbA1c levels favorably compares with the decline in HbA1c associated with the addition of noninsulin antidiabetic drugs to maximal metformin therapy.65 Second, our findings demonstrate that structured exercise of more than 150 minutes per week is associated with greater declines in HbA1c than structured exercise of 150 minutes or less per week in patients with type 2 diabetes. This finding is important because the current guideline-recommended exercise duration is at least 150 minutes per week.1,2Quiz Ref IDAlthough high-intensity exercise has been previously shown to have an association with HbA1c reduction,4 our findings did not demonstrate that more intensive exercise was associated with greater declines in HbA1c. It is important to mention that, due to a great variability in exercise intensity descriptions, we used an intensity rating as previously reported.5 Baseline HbA1c was one of the variables explaining the heterogeneity between studies, which underscores the greater magnitude of intervention effects in HbA1c among individuals with baseline HbA1c levels of greater than 7%, when compared with those with baseline HbA1c levels of less than 7%.7,8,66,67
To our knowledge, this is the first systematic review to assess the association between physical activity advice interventions and glycemic control. Our results showed that physical activity advice was associated with lesser declines in HbA1c than the studies evaluating structured exercise training. These results are consistent with a recent RCT demonstrating that supervised aerobic and resistance exercise training were more efficacious than physical activity advice alone in achieving declines in HbA1c.68
This review demonstrates that physical activity advice is only associated with HbA1c reduction when accompanied by a dietary cointervention. This highlights the need for a combined recommendation of these lifestyle interventions. Despite the fact that diet alone could improve glycemic control, most RCTs in our meta-analysis that evaluated physical activity plus a dietary intervention included a control group of a dietary intervention. Because HbA1c reduction in type 2 diabetes is associated with improved insulin resistance, and both exercise training/physical activity and body weight reduction induced by low-calorie diets1 have distinct mechanisms to elicit these effects, it is expected that these interventions applied together would result in greater metabolic effects.2,69 Therefore, patients with type 2 diabetes should receive dietary recommendations in combination with advice to increase physical activity. Taken together, these results provide important information for clinical practice.1,2
This study has limitations. Data extraction was unblinded, which is a potential source of bias. Additionally, high heterogeneity was identified in the meta-analyses, especially in the structured exercise training meta-analysis. To address this, we have performed analyses to identify clinical (eg, baseline HbA1c) and methodological differences (eg, amounts of exercise) between studies. Finally, the general quality of the studies was low, reflecting increased risk of bias in some studies. This may have contributed to the heterogeneity of our analyses.
Structured exercise, consisting of aerobic training, resistance training, or a combination of aerobic and resistance exercise training for at least 12 weeks, is associated with improved glycemic control in type 2 diabetic patients. Structured weekly exercise of more than 150 minutes per week was associated with greater declines in HbA1c. Structured exercise training reduced HbA1c to a larger degree than physical activity advice. Physical activity advice is beneficial only if associated with dietary recommendations.
Corresponding Author: Beatriz D. Schaan, MD, ScD, Servico de Endocrinologia—Hospital de Clínicas de Porto Alegre, Rua Ramior Barcelos 2350, prédio 12, 4° andar, 90035-003 Porto Alegre, RS, Brazil (email@example.com).
Author Contributions: Dr Schaan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Leitao, Gross, Ribeiro, Schaan.
Acquisition of data: Umpierre, Ribeiro, Zucatti, Schaan.
Analysis and interpretation of data: Umpierre, Ribeiro, Kramer, Leitao, Azevedo, Gross, Ribeiro, Schaan.
Drafting of the manuscript: Umpierre, Ribeiro, Kramer, Leitao, Zucatti, Ribeiro, Schaan.
Critical revision of the manuscript for important intellectual content: Leitao, Azevedo, Gross, Ribeiro, Schaan.
Statistical analysis: Umpierre, Ribeiro, Kramer.
Obtained funding: Leitao, Azevedo, Gross, Schaan.
Administrative, technical, or material support: Zucatti, Azevedo.
Study supervision: Azevedo, Ribeiro, Schaan.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Gross reports having attended advisory board meetings, received research support, and undertaking clinical trials sponsored by Bristol-Myers Squibb; and delivered continuing medical educational programs sponsored by Bristol-Myers Squibb, GlaxoSmithKline, and Merck Sharp & Dohme. Dr Ribeiro reports having received consulting fees from Merck Sharp & Dohme and Servier; lecturing fees from Abbott, Aventis, Bioassist, Merck Sharp & Dohme, and Servier; and grant support from Boehringer Ingelheim, Bristol-Myers Squibb, Merck Sharp & Dohme, and Servier. The other authors reported no disclosures.
Funding/Support: This study was partially supported by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) grant 576627/2008-9 and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) PNPD 03021/09-2.
Role of the Sponsor: The sponsors of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.