Article Text

Download PDFPDF

Amount and frequency of exercise affect glycaemic control more than exercise mode or intensity
  1. Alison R Harmer1,
  2. Mark R Elkins2,3
  1. 1Discipline of Physiotherapy, Clinical and Rehabilitation Sciences Research Group University of Sydney, Sydney, Australia
  2. 2Sydney Medical School, University of Sydney, Sydney, Australia
  3. 3Centre for Evidence-Based Physiotherapy, The George Institute for Global Health, Sydney, Australia
  1. Correspondence to Dr Mark R Elkins, Department of Respiratory Medicine, Royal Prince Alfred Hospital, Sydney, Australia; mark.elkins{at}

Statistics from

This section features a recent systematic review that is indexed on PEDro, the Physiotherapy Evidence Database ( PEDro is a free, web-based database of evidence relevant to physiotherapy.


The prevention and treatment of diabetes is vital.1 ,2 People with prediabetes and obesity are able to markedly reduce their risk of developing type 2 diabetes by lifestyle modification (increasing physical activity and losing weight via improved diet; number needed to treat (NNT)=7 over 3 years) or by medication (NNT=14).3 Among people who already have type 2 diabetes, exercise improves insulin action and reduces cardiovascular risk.4 Regular exercise per se (ie, without additional effects of a dietary intervention) significantly reduces glycosylated haemoglobin (HbA1c; indicative of improved glycaemic control) and visceral adiposity among people with type 2 diabetes.5 Improved glycaemic control is strongly associated with reduced microvascular complications6 and, among those with newly diagnosed type 2 diabetes, with a sustained reduction in cardiovascular events and mortality.7 Questions remain about whether the characteristics of the exercise undertaken influence the benefits obtained in glycaemic control.3 This systematic review sought to determine which aspects of exercise dosage were associated with greater benefit in glycaemic control among patients with type 2 diabetes.


The aim of this review was to synthesise the best available evidence about whether the frequency, intensity and duration of exercise influence HbA1c among people with type 2 diabetes. A subsequent aim was to assess these relationships independently for aerobic exercise, resistance exercise or combined aerobic and resistance exercise.

Searches and inclusion criteria

Five databases were searched for trials published between 1980 and 2012, using search terms for exercise, diabetes mellitus and physical activity. Reference lists of review articles and other relevant publications were also scanned. Only English, Spanish and Portuguese language publications were included. To be eligible, trials were required to have studied adults with type 2 diabetes randomised to exercise or control, with follow-up of at least 12 weeks. Assessment of eligibility and data extraction were performed by two reviewers working independently, with original authors being contacted regarding missing data.


The intervention group in the included studies was required to have undertaken supervised exercise training so that the frequency, intensity and duration of exercise was more likely to have been achieved as prescribed. The exercise training modes were aerobic, resistance or combined aerobic and resistance. The control group was required not to have received an exercise intervention.

Main outcome measures

The only outcome measure was HbA1c, which reflects the average plasma glucose concentration (glycaemic exposure of tissues) over the preceding 2–3 months.

The characteristics of the exercise interventions that were extracted were frequency (sessions per week), total time (minutes spent exercising throughout the study) and intensity (% of maximum heart rate for aerobic exercise and % of 1-repetition maximum for resistance exercise). Certain other characteristics were extracted for one type of exercise only, such as session duration and weekly exercise volume for aerobic exercise.

Statistical methods

Absolute changes in HbA1c were converted into weighted mean differences for each study. Correlations between these weighted mean differences and the characteristics of the exercise interventions (ie, frequency, total time, intensity, etc.) were then assessed. Those characteristics that were significant in this univariate analysis were entered into multivariate regression analyses. The weighted mean differences were then plotted against each of the characteristics, with ‘bubble plots’ displaying the weighting of each trial.

Quality was assessed using items selected from the PRISMA statement including adequate sequence generation, allocation concealment, blinding of assessors, intention-to-treat analysis and description of losses and exclusions.


Data from 26 eligible studies were included, which provided a pool of data from 2253 participants: 935 for aerobic exercise, 249 for resistance exercise and 1069 for combined exercise. Intervention durations ranged from 12 to 52 weeks. Although all trials described losses to follow-up and exclusions, the other quality items were achieved by a minority of the studies.

When data from all trials were pooled, greater benefit in HbA1c was observed with increases in exercise frequency (sessions per week) and weekly exercise duration (minutes per week). Although statistically significant, these associations were not strong.

When aerobic exercise trials were analysed separately, higher exercise frequency (sessions per week) clearly improved the benefit in HbA1c (figure 1A), whereas the intensity of the aerobic exercise (% of maximum heart rate) clearly had no association with the amount of benefit obtained (figure 1B). The analyses of resistance exercise did not identify any characteristics associated with the amount of HbA1c improvement. Among the trials of combined aerobic and resistance exercise, a significant association was noted: for each additional resistance exercise set per week, there was a corresponding decrease of 0.02% in HbA1c with combined exercise training.

Figure 1

Association of improvement in HbA1c with (A) frequency and (B) intensity, in studies of supervised aerobic training. The size of the circles represents the inverse variance of each study in the pooled analysis. HbA1c, glycosylated haemoglobin; WMD, weighted mean difference; HRmax, maximum heart rate. Reproduced with modification from Umpierre et al.


This study joins a growing number of systematic reviews in the exercise field to use meta-regression.8–10 Meta-regression assesses whether a particular characteristic of the intervention is related to the magnitude of the intervention's effect, thus indirectly estimating the influence of that characteristic on the effect size. Meta-regression should not be used when direct randomised comparisons of the intervention with and without the characteristic are available. If meta-regression is to be used, the strong potential for over-interpretation of findings can be reduced by including a substantial number of trials and minimising the number of characteristics to be analysed—although absolute thresholds do not accompany these recommendations.11 The meta-regression by Umpierre et al appears to follow these recommendations appropriately. The only precaution these authors have not taken is to temper the statistical significance of the meta-regression with a permutation test, which has been recommended when heterogeneity is high.11

Acknowledging the caveats regarding meta-regression outlined above, the main finding of this review was a strong association between the number of sessions of supervised aerobic exercise per week and the improvement in glycaemic control. This is physiologically plausible because among patients with type 2 diabetes, a single bout of moderate-intensity exercise increases GLUT 4 gene expression for at least 3 h12 and only 7 days of moderate-intensity exercise training enhances glucose disposal and skeletal muscle GLUT 4 transporter content.13 Thus, repeated bouts of exercise and/or exercise training may reduce exposure of haemoglobin to glycosylation; hence, increasing number of exercise sessions is consistent with improved HbA1c. Most of the other characteristics of the exercise interventions did not appear to modify the HbA1c treatment effect. However, the effect of some characteristics could not be assessed. For example, all of the resistance exercise trials had a frequency of three sessions per week, preventing analysis of this characteristic.

Four of the included trials explicitly described a dietary component (advice, monitoring or classes—intended to either reduce or maintain body weight) for intervention and control groups. Although exercise is preserved as the contrast between the groups in these trials, diet alone can be potent in reducing HbA1c; hence, this may increase the absolute changes in HbA1c observed in these trials. One other included trial employed a dietary component in the intervention group but not in the control group, which would confound results. Fortunately, this trial involved only 75 participants, limiting its influence on the overall result.

Trials of aerobic exercise included in the systematic review mostly employed continuous exercise of low-to-moderate intensity. Although maximal oxygen uptake was not an outcome of the systematic review, it is nonetheless an important outcome for clinicians to consider when prescribing exercise intensity because maximal oxygen uptake is strongly inversely related to morbidity and mortality among cohorts with cardiovascular disease.14 High-intensity aerobic interval training improves maximal oxygen uptake more than moderate-intensity continuous exercise among patient cohorts, for example, postinfarction heart failure15, but has not been tested among patients with type 2 diabetes in a randomised trial.

Among people with diabetes living in the community, common reasons for poor adherence to recommended exercise include lack of time, poor motivation and uncertainty about how to exercise.16 Exercise practitioners could incorporate motivational techniques, such as goal setting and problem solving, using a communication style that facilitates the involvement of the patient, as such strategies have been shown to improve collaboration and satisfaction.17–19 For other individuals, such as those with complications secondary to their diabetes or with unrelated comorbidities, the usual recommended exercise may not be possible.20 The clinician's role here is to tailor a feasible exercise programme, perhaps supplemented with other interventions to control glycaemia such as passive stretching21 and referral to a dietician if dietary interventions have not been trialled.

Clinical implications

Patients seeking to maximise glycaemic control using aerobic exercise should be advised that exercising more times per week may be more beneficial than increasing the session duration or exercise intensity. This may be of particular importance to those with multimorbidity and among those with higher levels of HbA1c. Those undertaking combined aerobic and resistance exercise may also benefit from adding extra sets of resistance exercise, although this effect appears to be substantially weaker.


View Abstract


  • ▸ Umpierre D, Ribeiro PAB, Schaan BD, et al. The systematic review being appraised in this update. Diabetologia 2013;56:242–51.

  • Contributors ARH and MRE selected the systematic review. MRE wrote the first draft of the manuscript. ARH and MRE contributed to interpretation of the data and revision of the final manuscript, and are guarantors.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.