Sage Journals: Discover world-class research

Abstract

Objective: Although exercise can improve insulin sensitivity, no adequate synthesis exists of exercise intervention studies with regard to their effect on insulin sensitivity. This comprehensive meta-analysis synthesized the insulin sensitivity outcomes of supervised exercise interventions. Method: Extensive literature searching located published and unpublished intervention studies that measured insulin sensitivity outcomes. Eligible studies tested supervised exercise interventions among healthy adults. Primary study characteristics and results were coded. Random-effects meta-analyses of standardized mean differences included moderator analyses. Results: Data were synthesized across 2509 subjects (115 samples, 78 reports). The overall mean effect size for 2-group postintervention comparisons was 0.38 (95% confidence interval [CI] = 0.25-0.51, I² = 0%) and for 2-group pre–post comparisons was 0.43 (95% CI = 0.30-0.56, I² = 52%; higher mean insulin sensitivity for treatment than control subjects). The postintervention mean of 0.38 is consistent with treatment subjects ending studies with a mean fasting insulin of 6.8 mU/L if control participants’ mean fasting insulin were 7.9 mU/L. Exploratory moderator analyses did not document different insulin sensitivity effect sizes across intervention characteristics or sample attributes. Conclusion: This study documented that exercise is a valuable primary care and community health strategy for healthy adults to improve insulin sensitivity and lower the risk for diabetes conferred by insulin resistance.

Keywords

exercise insulin resistance insulin sensitivity meta-analysis systematic review

Introduction

Type 2 diabetes mellitus, which affects 346 million people worldwide, is a major personal, social, and economic burden.^1,2 Diabetes is an important comorbidity for other chronic illnesses commonly encountered in primary care.³ Considerable research addresses community health diet and exercise strategies to prevent diabetes. Exercise is an especially attractive intervention because it is cost-free and offers potential health benefits beyond diabetes prevention.⁴ Exercise is believed to help prevent diabetes by increasing insulin sensitivity (ie, the responsiveness of insulin-sensitive tissues to the metabolic actions of insulin).⁵ Reduced insulin sensitivity, or insulin resistance, precedes type 2 diabetes on the continuum from normoglycemia to clinically diagnosed diabetes. Insulin resistance is marked by decreased insulin-stimulated glucose uptake in skeletal muscle and adipose tissue.⁶ Exercise training is thought to affect insulin sensitivity via mechanisms related to carbohydrate and fat oxidation in skeletal muscle, vascular endothelial function, and synthesis of proteins involved in carbohydrate and fat metabolism.^7,8

The possible positive health consequences of exercise in preventing diabetes have led to numerous primary studies testing exercise interventions. The proliferation of such studies has prompted some researchers to summarize existing knowledge, primarily through narrative summaries which cannot calculate the magnitude of effect or conduct moderator analyses to determine if effect sizes are linked with characteristics of the sample or intervention.^7-10 Previous related meta-analyses have synthesized exercise intervention trials among adults with diabetes or to prevent gestational diabetes.^11-20

Few meta-analyses have addressed diabetes prevention in nonpregnant adults. Four meta-analyses synthesized 2 to 9 primary studies of combined exercise and diet interventions on fasting glucose and incidence of diabetes outcomes among adults with elevated fasting glucose or impaired glucose tolerance test.^21-24 One meta-analysis of only 2 primary studies focused on adults at high risk for diabetes and found that exercise did not significantly reduce diabetes incidence.⁵ No previous meta-analysis has addressed exercise interventions conducted with healthy adults; nor have previous studies examined insulin sensitivity and resistance as outcomes. No previous synthesis has examined supervised exercise interventions with verified exercise dose. This project moves past previous syntheses to address the following questions: (a) What is the overall effect of supervised exercise interventions on insulin sensitivity in healthy adults when exercise subjects are compared to control subjects? (b) In treatment versus control comparisons, do intervention effects vary depending on sample characteristics of age, gender, preintervention body mass index, or preintervention insulin sensitivity? (c) In treatment versus control comparisons, do intervention effects vary depending on intervention characteristics of behavioral target (exercise vs exercise plus diet), exercise intensity, exercise duration per session, exercise session frequency, exercise total dose, and exercise mode (endurance vs endurance plus resistance)?

Methods

Standard systematic review and meta-analysis procedures were used to identify and retrieve potential studies, determine eligibility, code data, analyze primary study results, and interpret findings. PRISMA guidelines were used to conduct the study and report this project.²⁵ This study is part of a larger comprehensive meta-analysis examining physical activity interventions in healthy adults. The study protocol was not registered.

Sample

Inclusion Criteria

Eligible studies were English-language reports of insulin sensitivity and resistance outcomes following exercise interventions conducted with healthy adults. We included studies of exercise sessions supervised by research staff to verify exercise behavior and studies that measured insulin sensitivity, including fasting insulin as a surrogate index for insulin sensitivity, within 7 days after completing supervised exercise sessions.^6,26 The hyperinsulinemic euglycemic glucose clamp is the gold standard for measuring insulin sensitivity/resistance.^26,27 Unfortunately, the hyperinsulinemic euglycemic glucose clamp is rarely used in exercise studies because it is costly, invasive, and time-intensive.^26,28 Surrogate markers of insulin sensitivity are commonly used and recommended for use in some forms of research and clinical practice.^26,27,29 Fasting insulin is a commonly used surrogate marker of insulin resistance.²⁶ We excluded were studies that exclusively reported glucose levels, because these values can be poor indicators of insulin sensitivity.⁶ We excluded projects that measured insulin sensitivity during or immediately after an exercise bout, because insulin sensitivity increases acutely with exercise.^28,30

To reduce bias, we included both published and unpublished studies because the most consistent difference between published and unpublished research is the statistical significance of findings.^31,32 Meta-analyses that include only published studies may overestimate effect sizes. Publication status is an inadequate proxy for study quality. We did not use primary study quality assessments as inclusion criteria, because extant study quality scales lack evidence of validity, mix report quality with study quality, and contain items not relevant to this area of science.³³ Primary study quality attributes were coded and reported. This project focused on treatment-versus-control effect sizes because these provide the most valid information. Given the limitations of single-group designs, we analyzed single-group pre–post comparisons as ancillary information that should be interpreted cautiously. Our findings emphasize the more valid 2-group comparisons.

Small-sample studies were included because it is important to synthesize across the range of research. Small studies might lack statistical power to detect differences, but because meta-analyses do not rely on primary study hypothesis tests these studies can contribute to synthesis findings. Studies were weighted such that small-sample studies had proportionally less influence on results.

Search Strategies

To enlarge the scope of studies beyond previous reviews and thus avoid biases introduced by narrow searches, we employed multiple search strategies.³⁴ Information sources included computerized databases, research registers, ancestry searches, author searches, and hand searches. An expert reference librarian conducted searches in 11 computerized databases (MEDLINE, Dissertation Abstracts, Sport Discus, EMBASE, PsychINFO, Healthstar, Cochrane Controlled Trials Register, Combined Health Information Database, Nursing and Allied Health Database, Database of Abstracts of Reviews of Effectiveness, Educational Resources Information Center) using broad search terms (sample MEDLINE intervention terms: clinical trial, intervention, treatment outcome, program, program development, program evaluation, outcome & process assessment, adherence, behavior therapy, compliance, counseling, evaluation, evaluation study, evidence-based medicine, health care evaluation, health behavior, health education, health promotion, patient education, self care, validation study; physical activity terms: exercise, physical activity, physical fitness, exertion, exercise therapy, physical education and training, walking). Search terms such as diabetes, insulin sensitivity, and insulin resistance were not used because this project was part of a larger comprehensive meta-analysis and limiting use of specific terms such as insulin helped ensure potential studies were not missed due to insufficient keyword terms applied to some potential primary studies. Initial searches were conducted in 2008 as part of the parent project, with automated monthly updates from computerized databases used to identify more recent studies. Research registers searched included the National Institutes of Health Research Portfolio Online Reporting Tool and mRCT, which has 14 active registers and 16 archived registers.³⁵ Ancestry searches of eligible primary studies and review articles were conducted. Computerized author searches were completed for corresponding authors of eligible studies. Hand searches of 114 journals were conducted.³⁵ Although time consuming, these diverse comprehensive search procedures are important because no single approach locates most eligible studies, and the pattern of bias varies among search mechanisms.^34,36 These comprehensive search strategies yielded 54,642 research reports that we considered for inclusion.

Study Selection

A staged eligibility determination process was used to identify eligible studies for the parent project and for this meta-analysis. The staged process was used to ensure all eligible studies for any part of the parent project reached the coding phase. First, title and abstracts were reviewed for visual heralds suggesting a potentially eligible study. Second, full reports were examined to determine whether the study included an exercise intervention in healthy adults. Third, potential primary studies were examined for any eligible outcome for the parent study. Fourth, studies were sorted to identify those with insulin sensitivity outcomes. Finally, potential primary studies were evaluated for adequate data to calculate an insulin sensitivity effect size. A flowchart of potential studies movement through the project is available in supplementary material.

Data Collection Instrument and Process

A coding frame was developed, pilot tested with 50 primary studies in the parent project, and refined to record primary study results and characteristics of participants, research methods, and interventions as well as source attributes.^37,38 Source characteristics included dissemination vehicle (eg, published article, dissertation, presentation at conference), presence of funding, and year of distribution. Sample mean age, gender, ethnic distribution, and anthropometric values such as preintervention body mass index were recorded. From primary reports we extracted intervention characteristics such as behavioral target (exercise only vs exercise plus other behaviors), aerobic only exercise versus aerobic plus resistance exercise, and exercise dose (minutes/session, number of sessions/week, total minutes of exercise). Research methods were coded, such as sample size, group allocation, percentage attrition, and insulin sensitivity measure.

We also extracted insulin sensitivity data. Pre- and postintervention means and standard deviations (SDs) for treatment and control groups were coded. If these data were not available, we coded other statistics that could be converted to the d index of effect size (eg, t statistic). If multiple research reports described the same study, all available reports were used to secure the required data. To prevent redundant data, we cross-checked all author names for all eligible studies to detect potentially overlapping samples.

Two extensively trained coders extracted data independently then compared coded values to achieve 100% agreement. A third doctorally prepared coder confirmed effect size data. Remaining discrepancies were resolved by the principal investigator.

Risk of Bias

Intensive and extensive search strategies were employed to avoid the bias introduced by including studies with larger effect sizes, which are often easier to locate. Search strategies included attempts to locate unpublished studies to reduce bias. Potential bias from selective reporting within studies was managed by coding a priori determined insulin sensitivity measures. Publication bias was assessed using multiple methods described below. Design bias was partially addressed by reporting effect sizes separately for 2-group postintervention comparisons, 2-group pre–post comparisons, and treatment single-group comparisons. Control single-group comparisons were used to explore potential bias from participation in research studies. Potentially biased measures were managed by excluding studies that reported fasting glucose values as measures of insulin sensitivity.

Statistical Analyses

The ability to combine outcomes across studies with different measures of the same construct is an important strength of meta-analyses. This is possible because the effect size is scale-free because of standardization.^39,40 The standardized mean difference (d index) effect size used in this study reflects the difference in means between treatment and control subjects in terms of their shared SD.⁴⁰ Meta-analyses commonly combine primary research that did not use the same measurement operations for the outcome variable.⁴⁰ Effect size numerical values can be interpreted in a consistent manner across studies because of standardization.⁴⁰ We calculated a standardized mean difference effect size for each comparison.^41,42 For 2-group postintervention comparisons, the effect size represents the difference between treatment and control groups’ means divided by the pooled SD, all from postintervention scores. For studies that provided pre- and postintervention data for both treatment and control subjects, a 2-group pre–post effect size was calculated as the treatment pre–post effect size minus the control pre–post effect size. Each of the treatment and control pre–post effect sizes is a single-group effect size calculated as the group’s difference between postintervention and preintervention means divided by its preintervention SD. A positive effect size represents a better mean for treatment subjects than for control subjects (for 2-group comparisons) or at postintervention than preintervention (for pre–post comparisons). Effect sizes were adjusted for bias.⁴¹ We calculated single-group effect sizes for both treatment and control groups. Control group data were analyzed because possible changes could be useful for interpreting treatment pre-post comparisons. A given group’s preintervention and post-intervention insulin sensitivity scores probably are correlated. Primary study authors were contacted to request information about this correlation, but few provided it. We never combined single-group and 2-group effect sizes, which are reported separately in this article. Single-group effect sizes provide information to supplement the more valid 2-group findings.

To give more influence to larger sample studies, we weighted each effect size by the inverse of its sampling variance. Random-effects analyses were used to acknowledge that effect sizes vary due to subject-level sampling error and also from other sources of study-level variations, such as sample or intervention characteristics.⁴³ Ninety-five percent confidence intervals were constructed. Forest plots were developed to graphically depict findings. To facilitate interpretation, we converted effect sizes to the original metric of fasting insulin, the most common insulin sensitivity measure in the primary studies.⁴⁰

We detected outliers statistically by omitting each effect size one at a time and checking for substantially reduced measures of heterogeneity or large externally standardized residuals. Multiple approaches, including estimates of the number of omitted studies, tests of funnel-plot asymmetry, and selection function procedures were used to explore potential publication bias.^44-48

Homogeneity was assessed with a conventional homogeneity statistic (Q).⁴⁹ We used I², an index of between-studies heterogeneity relative to within-study sampling error, to assess the impact of (in)consistency among trials. Statistical and clinical heterogeneity is common in studies that are not virtual replications.⁵⁰ Heterogeneity was expected and handled in 4 ways.^11,51 Heterogeneity was (a) taken into account using random-effects analyses, (b) quantified as estimates of between-studies variance and I², (c) explored using moderator analyses, and (d) viewed as part of the context in which we interpreted results.

We conducted prespecified exploratory moderator analyses to determine if effect sizes were associated with primary study attributes. We examined behavioral target, exercise mode, and type of insulin sensitivity categorical moderators using a meta-analytic analogue of analysis of variance. A meta-analytic analogue of meta-regression was used for the continuous variables of baseline fasting insulin mean and SD scores, preintervention body mass index mean and SD scores, mean age, proportion female, exercise session duration, exercise session frequency, and the total minutes of supervise exercise. Each of these yields an estimate and test of the moderator effect (eg, difference between mean effect sizes, unstandardized regression slope).

Results

Comprehensive searching located 115 comparisons from 78 reports that included 2509 participants.^52-129 Most studies were disseminated as published articles (s = 77; s denotes the number of reports, k indicates the number of comparisons); one dissertation was included. Recent studies were common, with 39 reports disseminated in 2000 or more recently. Most reports (s = 71) indicated some funding for the research.

The 2-group postintervention and 2-group pre–post intervention comparisons included 1100 participants (733 treatment, 367 control) subjects in 34 comparisons. Treatment pre–post comparisons included 2142 participants. Control pre–post comparisons included 367 participants.

Primary Study Characteristics and Quality Attributes

Primary study characteristics are presented in Table 1. Median sample size was 22 participants (range = 5-154). Participant attributes were inconsistently reported (Table 1). The median of mean age was 43 years. Only 14 of the 78 reports provided minority participation rates. The median proportion of females was 48%.

Table 1.

Characteristics of Primary Studies Included in Insulin Sensitivity Meta-Analyses.^a

Characteristic	k	Min	Q ₁	Median	Q ₃	Max
Mean age (years)	71	20	36	43	58	80
Total postintervention sample size per study	78	5	12	22	36	154
Treatment group postintervention sample size per study	78	5	12	20	30	114
Comparison group postintervention sample size per study	20	6	9	14	25	44
Percentage attrition from total sample^a	15	0	4	18	29	64
Percentage attrition from treatment sample	46	0	0	12	30	65
Percentage attrition from control sample	15	0	0	8	23	58
Percentage female	76	0	0	48	100	100
Percentage ethnic minority	14	0	0	0	24	64
Mean preintervention fasting insulin for total sample^b	4	7	7	8	8	9
Mean preintervention fasting insulin for treatment sample	9	7	7	7	9	15
Mean preintervention fasting insulin for comparison sample	4	7	7	8	8	9
Minutes of supervised exercise per session	68	30	40	51	60	480
Mean frequency per week of exercise sessions	74	1	3	3	5	14
Total number of supervised exercise sessions	74	6	30	51	87	349

Abbreviations: k, number of studies providing data on characteristic; Q₁, first quartile, Q₃, third quartile.

Includes all studies that contributed to primary analyses at least one effect size for any type of comparison. Samples within studies were aggregated by summing sample sizes and using weighted mean of other characteristics (weighted by sample size).

Includes only comparisons with both treatment sample and comparison sample.

The median attrition was 18%. Thirty of the 34 two-group comparisons individually randomized participants to treatment and control groups. We excluded some potentially eligible comparisons because they used fasting glucose to measure insulin resistance (k = 42). The median of mean preintervention fasting insulin was 8 mU/L.

Not all studies provided adequate details about the intervention (Table 1). For example, only 68 reported the duration of exercise sessions. The median of mean minutes per supervised exercise session was 51 minutes. The median of mean session frequency was 3 sessions per week. The median of mean total number of supervised exercise sessions was 51.

Overall Effects of Interventions

Table 2 shows the effects of interventions on effect sizes. The mean effect size was 0.38 for 2-group postintervention comparisons and 0.43 for 2-group pre–post comparisons. Forest plots of 2-group postintervention comparisons and of 2-group pre–post comparisons are available in the electronic supplementary material. A mean effect size of 0.28 was documented for treatment pre–post comparisons. These effect sizes indicate that, on average, interventions did increase insulin sensitivity. In contrast, control subjects experienced a small decrease in insulin sensitivity, as evidenced by a mean effect size of −0.10.

Table 2.

Intervention Effect by Type of Comparison: Random-Effects Estimates and Tests.^a

Comparison Type	k	Mean ES	95% CI	SE	${\hat{σ}}_{δ}$	Q	I ²
Two-group postintervention comparison	34	0.38***	(0.25, 0.51)	0.067	0	27.4	0%
Two-group pre–post comparison	34	0.43***	(0.30, 0.56)	0.068	0.27	69.4***	52%
Treatment pre–post comparison	115	0.28***	(0.22, 0.34)	0.030	0.23	275.5***	59%
Control pre–post comparison	21	−0.10^†	(−0.21, 0.00)	0.054	0.15	32.4*	38%

Abbreviations: k, number of comparisons; mean ES, estimated mean of true effect sizes; SE, standard error.

Under homogeneity (H₀: δ_i = δ), Q is distributed approximately as χ² with df = k − 1, where k is number of (possibly dependent) observed effect sizes; this also tests between-studies variance component, ${\hat{σ}}_{δ}$ (H₀: ${\hat{σ}}_{δ}$ = 0). Weighted method of moments used to estimate ${\hat{σ}}_{δ}$ . Potential outliers excluded based on standardized random-effects residuals (0, 0, 4 [3.4% of comparisons], and 0 for 2-group postintervention, 2-group pre–post, treatment pre–post, and control pre–post, respectively). Duplicate samples omitted from analyses of control pre–post comparisons.

†

P < .10, *P < .05, **P < .01, ***P < .001 (for ES and Q).

To facilitate interpretation, we converted the summary effect size to a value of fasting insulin, which was the most commonly reported original metric: The 2-group postintervention mean effect size of 0.38 is consistent with treatment participants’ ending studies with a mean fasting insulin of 6.8 mU/L if controls end with 7.9 mU/L.

Findings from heterogeneity analyses (Q, I², and estimate of between-studies variance) suggest variation in true effect sizes among studies included in 2-group pre–post, treatment pre–post, and control pre–post comparisons (Table 2). In particular, 2-group and treatment pre–post effect sizes exhibited substantial heterogeneity. Two-group postintervention effect sizes were not significantly heterogeneous and exhibited no between-studies variance. Evidence for or against publication bias was weak because of few studies for comparisons.

Exploratory Moderator Analyses

Exploratory moderator analyses addressed sample attributes and intervention characteristics. Tables reporting the moderator analyses are available from the corresponding author and in supplementary material. Some analyses contained few studies (ie, behavioral target, exercise mode, body mass index); findings for these moderators should be interpreted cautiously, as should findings for aggregate participant variables used as moderators.¹³⁰ The difference in mean effect size between interventions, including both aerobic and resistance exercise (effect size = 0.26, k = 4) and exclusively aerobic exercise (effect size = 0.44; k = 30) was not statistically significant. The 5 studies combining exercise and diet interventions (effect size = 0.51) did not have a statistically significantly larger mean effect size than the 29 studies exclusively targeting exercise (effect size = 0.39). Neither mean preintervention fasting insulin level ( ${\hat{β}}_{1} = - 0.59$ , k = 10, natural log of fasting insulin scores) nor mean preintervention body mass index ( ${\hat{β}}_{1} = - 0.02$ , k = 7) was related to insulin sensitivity outcomes. Gender distribution ( ${\hat{β}}_{1} = - 0.01$ , k = 34) and sample mean age ( ${\hat{β}}_{1} = - 0.00$ , k = 34) were not associated with differences in insulin sensitivity effect sizes.

Exercise dose variables (minutes/session [ ${\hat{β}}_{1} = - 0.92$ , k = 23, log₁₀ duration in minutes], sessions/week [ ${\hat{β}}_{1} = - 0.07$ , k = 31, log₁₀ numbers sessions per week], total minutes of supervised exercise [ ${\hat{β}}_{1} = - 0.26$ , k = 23, log₁₀ total minutes of exercise]) were unrelated to insulin sensitivity effect sizes. Analyses of supervised exercise intensity, Q(2) = 2.0, P = .369, demonstrated no significant variation in mean effect size among low, moderate, and high intensity.

An additional post hoc sensitivity analyses was conducted to explore if effect sizes varied between studies with fasting insulin and those with other insulin resistance measures. For a given type of comparison we conducted 2 separate random-effects analyses to obtain 2 estimates of the mean effect size: one for insulin resistance measures and another for fasting insulin measures. That permitted the between-studies variance to differ between measures but required estimating one more parameter than a conventional analysis of variance–type moderator analysis with a residual between-studies variance. We applied standard-normal procedures to those results to estimate and test the difference between the 2 measures’ mean effect sizes. Unfortunately, most studies measured fasting insulin (k = 27) so we had scant data for direct/indirect measures of insulin sensitivity (k = 7). Comparisons of effect sizes between the measures yielded inconsistent findings, which suggest a possible lower effect size for studies with fasting insulin measures in some analyses. Given the small sample size and these analyses were unplanned post hoc comparisons, the findings should be viewed as exploratory to prompt further research.

Discussion

Diabetes is a growing problem among primary care patients. Given the type 2 diabetes epidemic and the link between insulin resistance and diabetes, interventions that increase insulin sensitivity are important.^131,132 This comprehensive meta-analysis documented moderate improvements (mean effect size 0.38 and 0.43) in insulin sensitivity among healthy adults after supervised exercise interventions. No previous meta-analyses have analyzed insulin sensitivity outcomes, so these findings cannot be directly compared with those from similar quantitative syntheses. Meta-analyses of supervised exercise interventions among adults with type 2 diabetes reported standardized mean difference effect sizes of 0.28 to 0.43 for HbA1c outcomes.^11-13 The present findings add important new information about the magnitude of improvements in insulin sensitivity among healthy adults and go beyond previous meta-analyses that documented improved glycemic control among type 2 diabetes samples after exercise interventions.

The moderate improvement in insulin sensitivity among adults without insulin resistance is clinically significant for primary care. Low insulin levels, as have been documented in the people of Kitava in the Trobriand Islands of Papua New Guinea, might partly explain their relative absence of overweight, hypertension, and cardiovascular disease. Their fasting insulin levels, ranging from 3.4 to 3.9 IU/mL, are approximately half that of age-matched Swedes who already had overweight and borderline hypertension.¹³³ The benefit of decreasing insulin resistance as measured by fasting insulin is evident in its role of moving the patient back toward normal along the continuum ranging from normoglycemia to type 2 diabetes. There may be other benefit as well because increases of as little as 15 pmol/L (similar to 2.16 mU/L) have been associated with a 2-fold increase in the risk of coronary heart disease, which also gives context for the magnitude of the effect of exercise.¹³⁴ Moreover, the possibility of a modest decline in insulin sensitivity among healthy control subjects, even over these time-limited interventions, could be important.¹³¹ The larger mean effect size for the 2-group pre–post (vs 2-group postintervention) comparison may partly reflect deterioration among control subjects over the study interval. Improved insulin sensitivity may partially account for reduced diabetes incidence following lifestyle changes.^5,21,22,24 Exercise may affect multiple mechanisms to reduce insulin resistance.^8,135 This study documented that exercise is a valuable primary care and community health strategy for healthy adults to improve insulin sensitivity and lower the risk for diabetes conferred by insulin resistance.

The results of the moderator analyses are exploratory. Given the homogeneity of the effect sizes, there was little heterogeneity for which to account. Also, in some analyses, low power due to few studies made detecting moderators’ effects difficult (eg, exercise mode, behavior target of exercise vs. exercise plus diet). Our findings that do not confirm the body of research that has demonstrated greater improvement in insulin sensitivity with high-intensity compared with moderate-intensity exercise, with endurance plus resistance training, and with exercise training that includes energy restriction were unexpected.^9,135,136 Dose variation was unrelated to effect sizes, but most studies had similar minutes per session, sessions per week, and total number of sessions. Primary studies with greater variations in dose and studies which systematically vary components of dose would be useful. In the current analysis, the median frequency of 3 sessions per week may be insufficient to cause larger improvement in insulin sensitivity outcomes because the effect of exercise on insulin signaling in skeletal muscle is relatively short-lived, lasting up to 48 hours.^135,137 Determining dose response will require intervention research that reaches or exceeds current U. S. physical activity guidelines (ie, at least 30 minutes of moderate-intensity aerobic activity on most, and preferably all, days of the week plus 2 or more days a week of muscle strengthening exercises).¹³⁸ Primary research examining outcomes over longer periods of regular exercise behavior would also help.

Identifying how participant characteristics relate to outcomes is a valuable component of meta-analyses. Our findings suggest that exercise interventions may be similarly effective across body mass index and preintervention fasting insulin variations, though O’Gorman and Krook⁷ reported differing exercise training effects on resting fat oxidation among obese, type 2 diabetic, and lean populations. In our study, effects did not differ by mean age or gender proportion. As mentioned above, low power in some analyses as well as limited heterogeneity in exercise dose across interventions may have concealed true effect size differences. In addition, including only studies of healthy adults could account for this finding. Fasting insulin levels could be a significant modifier of effects by including studies of subjects with significant insulin resistance. Other characteristics, such as genetic factors and distribution of adiposity, likely influenced responsiveness of individual changes in insulin sensitivity to exercise.^7-9 Future primary research should consistently report participant characteristics or report individual participant data so future moderator analyses can examine these sample attributes.

Limitations of primary studies can diminish confidence in findings and create challenges for meta-analyses. A persistent problem in synthesizing studies has been their inadequate reporting of intervention details, such as dose variables in this study.¹³⁹ Variations in measures used to assess insulin sensitivity are another limitation. Measurement of insulin varies among laboratories.¹⁴⁰ This meta-analysis used both fasting insulin and direct/indirect insulin sensitivity measures. The post hoc analysis, which suggested that effect sizes may be smaller in studies using fasting insulin levels, is preliminary and should not be considered confirmatory. The findings should encourage more careful investigation of this type of difference in well-designed primary studies. The cause of any difference in measures is unknown. It is possible the properties of the two measures make them react differently to exercise interventions. A feasible alternative is that any differences are partly an artifactual difference due to something like differential reliability (eg, if fasting insulin scores contain more measurement error “noise” than insulin resistance, that could increase the former’s SD, thereby decreasing its effect size). Measurement of insulin sensitivity instead of fasting insulin would be preferable in future research. Further primary studies which report both fasting insulin and direct/indirect measures of insulin sensitivity would be valuable.

This is the first meta-analysis to quantitatively synthesize findings of studies testing exercise interventions to improve insulin resistance. The focus on exercise interventions with verified exercise dose is valuable. Our searches were comprehensive. The exploratory moderator analyses provided the first attempt to identify systematic variations among studies. To better understand the meaning of the effects, we converted effect sizes to an original metric.

Overall, with regard to insulin sensitivity outcomes, the data we synthesized support the efficacy of exercise interventions. These findings document the importance of primary care and community health intervention to increase physical activity to decrease diabetes risk. Future research should examine greater variations in exercise form and dose to clarify the necessary form and dose to achieve important public health goals. Long-term follow-up studies are needed to trace the linkages among exercise behavior, insulin sensitivity, and development of diabetes.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Financial support provided by a grant from the National Institutes of Health (R01NR009656) to Vicki Conn, principal investigator. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author Biographies

Vicki S. Conn is Potter-Brinton Distinguished Professor at the University of Missouri Sinclair School of Nursing. Her NIH funded research uses meta-analysis to examine physical activity behavior change and medication adherence interventions.

Richelle J. Koopman is Associate Professor in the Department of Family and Community Medicine at University of Missouri. Her research is focused on patient use of online health resources, type 2 diabetes and pre-diabetes, and health information technology.

Todd M. Ruppar is Robert Wood Johnson Foundation Nurse Faculty Scholar and Assistant Professor at the University of Missouri Sinclair School of Nursing. His work focuses on chronic disease self-management.

Lorraine J. Phillips is Assocation Professor at the University of Missouri Sinclair School of Nursing. Dr. Phillip's research focuses on maintaining physical function among aging adults.

David R. Mehr is the William C. Allen Professor in Family and Community Medicine. His research focuses on preventive services and chronic disease care.

Adam R. Hafdahl is an independent statistical consultant. His work is focused on meta-analysis methods.

References

. Genes, environment, and interactions in prevention of type 2 diabetes: a focus on physical activity and lifestyle changes. Curr Mol Med. 2008;8:519-532.

World Health Organization. World diabetes. 2012. http://www.who.int/mediacentre/events/annual/world_diabetes_day/en/index.html.

Fisher

. Multiple chronic conditions: diabetes associated with comorbidity and shared risk factors using CDC WEAT and SAS analytic tools [published online September 10, 2013]. J Prim Care Community Health. doi:10.1177/2150131913503347.

. Prevention of diabetes and cardiovascular disease among prediabetic individuals: lifestyle versus drug interventions. Eur J Cardiovasc Prev Rehabil. 2011;18:810-812.

Orozco

Buchleitner

Gimenez-Perez

Roqué

Figuls

Richter

Mauricio

. Exercise or exercise and diet for preventing type 2 diabetes mellitus. Cochrane Database Syst Rev. 2008;(3):CD003054.

Muniyappa

Lee

Chen

Quon

. Current approaches for assessing insulin sensitivity and resistance in vivo: advantages, limitations, and appropriate usage. Am J Physiol Endocrinol Metab. 2008;294:E15-E26.

O’Gorman

Krook

. Exercise and the treatment of diabetes and obesity. Endocrinol Metab Clin North Am. 2008;37:887-903.

Ostergard

Jessen

Schmitz

Mandarino

. The effect of exercise, training, and inactivity on insulin sensitivity in diabetics and their relatives: what is new? Appl Physiol Nutr Metab. 2007;32:541-548.

Gill

. Physical activity, cardiorespiratory fitness and insulin resistance: a short update. Curr Opin Lipidol. 2007;18:47-52.

10.

Hussain

Claussen

Ramachandran

Williams

. Prevention of type 2 diabetes: a review. Diabetes Res Clin Pract. 2007;76:317-326.

11.

Conn

Hafdahl

Mehr

LeMaster

Brown

Nielsen

. Metabolic effects of interventions to increase exercise in adults with type 2 diabetes. Diabetologia. 2007;50:913-921.

12.

Umpierre

Ribeiro

Kramer

. 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;305:1790-1799.

13.

Boule

Haddad

Kenny

Wells

Sigal

. Effects of exercise on glycemic control and body mass in type 2 diabetes mellitus: a meta-analysis of controlled clinical trials. JAMA. 2001;286:1218-1227.

14.

Han

Middleton

Crowther

. Exercise for pregnant women for preventing gestational diabetes mellitus. Cochrane Database Syst Rev. 2012;7:CD009021.

15.

Oostdam

van Poppel

Wouters

van Mechelen

. Interventions for preventing gestational diabetes mellitus: a systematic review and meta-analysis. J Womens Health (Larchmt). 2011;20:1551-1563.

16.

Tobias

Zhang

van Dam

Bowers

. Physical activity before and during pregnancy and risk of gestational diabetes mellitus: a meta-analysis. Diabetes Care. 2011;34:223-229.

17.

Avery

Flynn

van Wersch

Sniehotta

Trenell

. Changing Physical Activity Behavior in Type 2 Diabetes: A Systematic Review and Meta-Analysis of Behavioral Interventions. Newcastle upon Tyne, UK: Institute of Cellular Medicine, Newcastle University; 2012.

18.

Chudyk

Petrella

. Effects of exercise on cardiovascular risk factors in type 2 diabetes: a meta-analysis. Diabetes Care. 2011;34:1228-1237.

19.

Kirk

Mutrie

MacIntyre

Fisher

. Effects of a 12-month physical activity counselling intervention on glycaemic control and on the status of cardiovascular risk factors in people with type 2 diabetes. Diabetologia. 2004;47:821-832.

20.

Thomas

Elliott

Naughton

. Exercise for type 2 diabetes mellitus. Cochrane Database Syst Rev. 2006;(3):CD002968.

21.

Yamaoka

Tango

. Efficacy of lifestyle education to prevent type 2 diabetes: a meta-analysis of randomized controlled trials. Diabetes Care. 2005;28:2780-2786.

22.

Gillies

Abrams

Lambert

. Pharmacological and lifestyle interventions to prevent or delay type 2 diabetes in people with impaired glucose tolerance: systematic review and meta-analysis. BMJ. 2007;334:299.

23.

Hopper

Billah

Skiba

Krum

. Prevention of diabetes and reduction in major cardiovascular events in studies of subjects with prediabetes: meta-analysis of randomised controlled clinical trials. Eur J Cardiovasc Prev Rehabil. 2011;18:813-823.

24.

Cardona-Morrell

Rychetnik

Morrell

Espinel

Bauman

. Reduction of diabetes risk in routine clinical practice: are physical activity and nutrition interventions feasible and are the outcomes from reference trials replicable? A systematic review and meta-analysis. BMC Public Health. 2010;10:653.

25.

Liberati

Altman

Tetzlaff

. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions: explanation and elaboration. BMJ. 2009;339:b2700.

26.

Singh

Saxena

. Surrogate markers of insulin sensitivity: a review. World J Diabetes. 2010;1:36-47.

27.

Antuna-Puente

Disse

Rabasa-Lhoret

Laville

Capeau

Bastard

. How can we measure insulin sensitivity/resistance? Diabetes Metab. 2011;37:179-188.

28.

Lorenzo

Haffner

Stancakova

Laakso

. Relation of direct and surrogate measures of insulin resistance to cardiovascular risk factors in nondiabetic finnish offspring of type 2 diabetic individuals. J Clin Endocrinol Metab. 2010;95:5082-5090.

29.

Buchanan

Watanabe

Xiang

. Limitations in surrogate measures of insulin resistance. J Clin Endocrinol Metab. 2010;95:4874-4876.

30.

Wojtaszewski

Hansen

Gade

. Insulin signaling and insulin sensitivity after exercise in human skeletal muscle. Diabetes. 2000;49:325-331.

31.

Rothstein

Hopewell

. Grey literature. In: Cooper

Hedges

Valentine

, eds. The Handbook of Research Synthesis and Meta-Analysis. 2nd ed. New York, NY: Russell Sage Foundation; 2009:103-125.

32.

Cook

Guyatt

Ryan

. Should unpublished data be included in meta-analyses? Current convictions and controversies. JAMA. 1993;269:2749-2753.

33.

Valentine

. Judging the quality of primary research. In: Cooper

Hedges

Valentine

, eds. The Handbook of Research Synthesis and Meta-Analysis. 2nd ed. New York, NY: Russell Sage Foundation; 2009:129-146.

34.

White

. Scientific communication and literature retrieval. In: Cooper

Hedges

Valentine

, eds. The Handbook of Research Synthesis and Meta-Analysis. 2nd ed. New York, NY: Russell Sage Foundation; 2009:51-71.

35.

Reed

Baxter

. Using reference databases. In: Cooper

Hedges

Valentine

, eds. The Handbook of Research Synthesis and Meta-Analysis. 2nd ed. New York, NY: Russell Sage Foundation; 2009:73-101.

36.

Conn

Isaramalai

Rath

Jantarakupt

Wadhawan

Dash

. Beyond MEDLINE for literature searches. J Nurs Scholarsh. 2003;35:177-182.

37.

Wilson

. Systematic coding. In: Cooper

Hedges

Valentine

, eds. The Handbook of Research Synthesis and Meta-Analysis. 2nd ed. New York, NY: Russell Sage Foundation; 2009:159-176.

38.

Lipsey

. Identifying interesting variables and analysis opportunities. In: Cooper

Hedges

Valentine

, eds. The Handbook of Research Synthesis and Meta-Analysis. 2nd ed. New York, NY: Russell Sage Foundation; 2009:147-158.

39.

Cooper

. Research Synthesis and Meta-Analysis: A Step-by-Step Approach. 4th ed. Thousand Oaks, CA: Sage; 2009.

40.

Lipsey

Wilson

. Practical Meta-Analysis. Thousand Oaks, CA: Sage; 2001.

41.

Hedges

Olkin

. Statistical Methods for Meta-Analysis. Orlando, FL: Academic Press; 1985.

42.

Morris

DeShon

. Combining effect size estimates in meta-analysis with repeated measures and independent-groups designs. Psychol Methods. 2002;7:105-125.

43.

Raudenbush

. Analyzing effect sizes: random effects models. In: Cooper

Hedges

Valentine

, eds. The Handbook of Research Synthesis and Meta-Analysis. 2nd ed. New York, NY: Russell Sage Foundation; 2009:279-294.

44.

Rosenthal

. The file drawer problem and tolerance for null results. Psychol Bull. 1979;86:638-641.

45.

Gleser

Olkin

. Models for estimating the number of unpublished studies. Stat Med. 1996;15:2493-2507.

46.

Sterne

Egger

. Funnel plots for detecting bias in meta-analysis: guidelines on choice of axis. J Clin Epidemiol. 2001;54:1046-1055.

47.

Sutton

. Publicaton bias. In: Cooper

Hedges

Valentine

, eds. The Handbook of Research Synthesis and Meta-Analysis. 2nd ed. New York, NY: Russell Sage Foundation; 2009:435-452.

48.

Vevea

Hedges

. A general linear model for estimating effect size in the presence of publication bias. Psychometrika. 1995;60:419-435.

49.

Higgins

Thompson

Deeks

Altman

. Measuring inconsistency in meta-analyses. BMJ. 2003;327:557-560.

50.

Lipsey

Wilson

. The way in which intervention studies have “personality” and why it is important to meta-analysis. Eval Health Prof. 2001;24:236-254.

51.

Berlin

. Invited commentary: benefits of heterogeneity in meta-analysis of data from epidemiologic studies. Am J Epidemiol. 1995;142:383-387.

52.

Aldred

Hardman

Taylor

. Influence of 12 weeks of training by brisk walking on postprandial lipemia and insulinemia in sedentary middle-aged women. Metab Clin Exp. 1995;44:390-397.

53.

Andersson

Rebuffe-Scrive

Terning

Krotkiewski

Bjorntorp

. The effects of exercise training on body composition and metabolism in men and women. Int J Obes. 1991;15:75-81.

54.

Asikainen

Miilunpalo

Kukkonen-Harjula

. Walking trials in postmenopausal women: effect of low doses of exercise and exercise fractionization on coronary risk factors. Scand J Med Sci Sports. 2003;13:284-292.

55.

Asikainen

Miilunpalo

Oja

. Randomised, controlled walking trials in postmenopausal women: the minimum dose to improve aerobic fitness? Br J Sports Med. 2002;36:189-194.

56.

Bigard

Lavier

Ullmann

Legrand

Douce

Guezennec

. Branched-chain amino acid supplementation during repeated prolonged skiing exercises at altitude. Int J Sport Nutr. 1996;6:295-306.

57.

Blumenthal

Sherwood

Gullette

. Exercise and weight loss reduce blood pressure in men and women with mild hypertension: Effects on cardiovascular, metabolic, and hemodynamic functioning. Arch Intern Med. 2000;160:1947-1958.

58.

Bouchard

Tremblay

Despres

. The response to exercise with constant energy intake in identical twins. Obes Res. 1994;2:400-410.

59.

Bouchard

Tremblay

Nadeau

. Long-term exercise training with constant energy intake. 1: effect on body composition and selected metabolic variables. Int J Obes. 1990;14:57-73. Erratum: Int J Obes. 1990;14:987.

60.

Bruun

Helge

Richelsen

Stallknecht

. Diet and exercise reduce low-grade inflammation and macrophage infiltration in adipose tissue but not in skeletal muscle in severely obese subjects. Am J Physiol Endocrinol Metab. 2006;290:E961-967.

61.

Christ-Roberts

Pratipanawatr

. Exercise training increases glycogen synthase activity and GLUT4 expression but not insulin signaling in overweight nondiabetic and type 2 diabetic subjects. Metab Clin Exp. 2004;53:1233-1242.

62.

Cox

Cortright

Dohm

Houmard

. Effect of aging on response to exercise training in humans: skeletal muscle GLUT-4 and insulin sensitivity. J Appl Physiol. 1999;86:2019-2025.

63.

Crampes

Marion-Latard

Zakaroff-Girard

. Effects of a longitudinal training program on responses to exercise in overweight men. Obes Res. 2003;11:247-256.

64.

De Glisezinski

Crampes

GHarrant

. Endurance training changes in lipolytic responsiveness of obese adipose tissue. Am J Physiol. 1998;275:E951-E956.

65.

Delecluse

Colman

Roelants

. Exercise programs for older men: mode and intensity to induce the highest possible health-related benefits. Prev Med. 2004;39:823-833.

66.

Dengel

Galecki

Hagberg

Pratley

. The independent and combined effects of weight loss and aerobic exercise on blood pressure and oral glucose tolerance in older men. Am J Hypertens. 1998;11:1405-1412.

67.

Dengel

Pratley

Hagberg

Rogus

Goldberg

. Distinct effects of aerobic exercise training and weight loss on glucose homeostasis in obese sedentary men. J Appl Physiol. 1996;81:318-325.

68.

Despres

Pouliot

M-C

Moorjani

. Loss of abdominal fat and metabolic response to exercise training in obese women. Am J Physiol. 1991;261:E159-E167.

69.

DiPietro

Dziura

Yeckel

Neufer

. Exercise and improved insulin sensitivity in older women: evidence of the enduring benefits of higher intensity training. J Appl Physiol. 2006;100:142-149.

70.

Donnelly

Hill

Jacobsen

. Effects of a 16-month randomized controlled exercise trial on body weight and composition in young, overweight men and women: The Midwest Exercise Trial. Arch Intern Med. 2003;163:1343-1350.

71.

Donnelly

Jacobsen

Heelan

Seip

Smith

. The effects of 18 months of intermittent vs continuous exercise on aerobic capacity, body weight and composition, and metabolic fitness in previously sedentary, moderately obese females. Int J Obes. 2000;24:566-572.

72.

Eriksson

Tuominen

Valle

. Aerobic endurance exercise or circuit-type resistance training for individuals with impaired glucose tolerance? Horm Metab Res. 1998;30:37-41.

73.

Evans

Racette

Peterson

Villareal

Greiwe

Holloszy

. Aerobic power and insulin action improve in response to endurance exercise training in healthy 77-87 yr olds. J Appl Physiol. 2005;98:40-45.

74.

Ferrara

Goldberg

Ortmeyer

Ryan

. Effects of aerobic and resistive exercise training on glucose disposal and skeletal muscle metabolism in older men. J Gerontol A Biol Sci Med Sci. 2006;61:480-487.

75.

Friedlander

Jacobs

Fattor

. Contributions of working muscle to whole body lipid metabolism are altered by exercise intensity and training. Am J Physiol Endocrinol Metab. 2007;292:E107-E116.

76.

Goulet

Melancon

Aubertin-Leheudre

Dionne

. Aerobic training improves insulin sensitivity 72-120 h after the last exercise session in younger but not in older women. Eur J Appl Physiol. 2005;95:146-152.

77.

Goulet

Melancon

Dionne

Aubertin-Leheudre

. No sustained effect of aerobic or resistance training on insulin sensitivity in nonobese, healthy older women. J Aging Phys Act. 2005;13:314-326. Erratum: J Aging Phys Act. 2006;14:359.

78.

Hickey

Houmard

Considine

. Gender-dependent effects of exercise training on serum leptin levels in humans. Am J Physiol. 1997;35:E562-E566.

79.

Horowitz

Leone

Feng

Kelly

Klein

. Effect of endurance training on lipid metabolism in women: a potential role for PPARα in the metabolic response to training. Am J. Physiol Endocrinol Metab. 2000;279:E348-E355.

80.

Houmard

Bruno

Bruner

McCammon

Israel

Barakat

. Effects of exercise training on the chemical composition of plasma LDL. Arterioscler Thromb. 1994;14:325-330.

81.

Houmard

Cox

MacLean

Barakat

. Effect of short-term exercise training on leptin and insulin action. Metab Clin Exp. 2000;49:858-861.

82.

Houmard

Shaw

Hickey

Tanner

. Effect of short-term exercise training on insulin-stimulated PI 3-kinase activity in human skeletal muscle. Am J Physiol. 1999;277:E1055-E1060.

83.

Houmard

Tanner

Slentz

Duscha

McCartney

Kraus

. Effect of the volume and intensity of exercise training on insulin sensitivity. J Appl Physiol. 2004;96:101-106.

84.

Hughes

Fiatarone

Fielding

. Exercise increases muscle GLUT-4 levels and insulin action in subjects with impaired glucose tolerance. Am J Physiol. 1993;264:E855-E862.

85.

Katzel

Bleecker

Rogus

Goldberg

. Sequential effects of aerobic exercise training and weight loss on risk factors for coronary disease in healthy, obese middle-aged and older men. Metab Clin Exp. 1997;46:1441-1447.

86.

Kiens

Jorgensen

Lewis

. Increased plasma HDL-cholesterol and apo A-1 in sedentary middle-aged men after physical conditioning. Eur J Clin Investig. 1980;10:203-209.

87.

Kirwan

Kohrt

Wojta

Bourey

Holloszy

. Endurance exercise training reduces glucose-stimulated insulin levels in 60- to 70-year-old men and women. J Gerontol. 1993;48:M84-M90.

88.

Kohrt

Landt

Birge

Jr . Serum leptin levels are reduced in response to exercise training, but not hormone replacement therapy, in older women. J Clin Endocrinol Metab. 1996;81:3980-3985.

89.

Kohrt

Obert

Holloszy

. Exercise training improves fat distribution patterns in 60- to 70-year-old men and women. J Gerontol. 1992;47:M99-M105.

90.

Krotkiewski

Lonnroth

Mandroukas

. The effects of physical training on insulin secretion and effectiveness and on glucose metabolism in obesity and type 2 (non-insulin-dependent) diabetes mellitus. Diabetologia. 1985;28:881-890.

91.

Krotkiewski

Mandroukas

Sjostrom

Sullivan

Wetterqvist

Bjorntorp

. Effects of long-term physical training on body fat, metabolism, and blood pressure in obesity. Metab Clin Exp. 1979;28:650-658.

92.

Lamarche

Despres

Pouliot

. Is body fat loss a determinant factor in the improvement of carbohydrate and lipid metabolism following aerobic exercise training in obese women? Metab Clin Exp. 1992;41:1249-1256.

93.

Lampman

Schteingart

Santinga

. The influence of physical training on glucose tolerance, insulin sensitivity, and lipid and lipoprotein concentrations in middle-aged hypertriglyceridaemic, carbohydrate intolerant men. Diabetologia. 1987;30:380-385.

94.

Layman

Evans

Baum

Seyler

Erickson

Boileau

. Dietary protein and exercise have additive effects on body composition during weight loss in adult women. J Nutr. 2005;135:1903-1910.

95.

Leick

Lindegaard

Stensvold

Plomgaard

Saltin

Pilegaard

. Adipose tissue interleukin-18 mRNA and plasma interleukin-18: effect of obesity and exercise. Obesity (Silver Spring). 2007;15:356-363.

96.

Leon

Conrad

Hunninghake

Serfass

. Effects of a vigorous walking program on body composition, and carbohydrate and lipid metabolism of obese young men. Am J Clin Nutr. 1979;32:1776-1787.

97.

Lithell

Schele

Vessby

Jacobs

. Lipoproteins, lipoprotein lipase, and glycogen after prolonged physical activity. J Appl Physiol. 1984;57:698-702.

98.

MacRae

Noakes

Dennis

. Effects of endurance training on lactate removal by oxidation and gluconeogenesis during exercise. Pflugers Arch. 1995;430:964-970.

99.

Mensink

Blaak

Corpeleijn

Saris

de Bruin

Feskens

. Lifestyle intervention according to general recommendations improves glucose tolerance. Obes Res. 2003;11:1588-1596.

100.

Monzillo

Hamdy

Horton

. Effect of lifestyle modification on adipokine levels in obese subjects with insulin resistance. Obes Res. 2003;11:1048-1054.

101.

Nicklas

Dennis

Berman

Sorkin

Ryan

Goldberg

. Lifestyle intervention of hypocaloric dieting and walking reduces abdominal obesity and improves coronary heart disease risk factors in obese, postmenopausal, African-American and Caucasian women. J Gerontol A Biol Sci Med Sci. 2003;58:181-189.

102.

Nicklas

Rogus

Berman

Dennis

Goldberg

. Responses of adipose tissue lipoprotein lipase to weight loss affect lipid levels and weight regain in women. Am J Physiol Endocrinol Metab. 2000;279:E1012-E1019.

103.

O’Donovan

Kearney

Nevill

Woolf-May

Bird

. The effects of 24 weeks of moderate- or high-intensity exercise on insulin resistance. Eur J Appl Physiol. 2005;95:522-528.

104.

Okazaki

Himeno

Nanri

Ogata

Ikeda

. Effects of mild aerobic exercise and a mild hypocaloric diet on plasma leptin in sedentary women. Clin Exp Pharmacol Physiol. 1999;26:415-420.

105.

Okura

Nakata

Ohkawara

. Effect of weight reduction on concentration of plasma total homocysteine in obese Japanese men. Obes Res Clin Pract. 2007;1:213-221.

106.

Pasman

Westerterp-Plantenga

Saris

WHM

. The effect of exercise training on leptin levels in obese males. Am J Physiol. 1998;274:E280-E286.

107.

Peltonen

Marniemi

Hietanen

Vuori

Ehnholm

. Changes in serum lipids, lipoproteins, and heparin releasable lipolytic enzymes during moderate physical training in man: a longitudinal study. Metab Clin Exp. 1981;30:518-526.

108.

Perseghin

Price

Petersen

. Increased glucose transport-phosphorylation and muscle glycogen synthesis after exercise training in insulin-resistant subjects. N Engl J Med. 1996;335:1357-1362.

109.

Perusse

Collier

Gagnon

. Acute and chronic effects of exercise on leptin levels in humans. J Appl Physiol. 1997;83:5-10.

110.

Poehlman

Danforth

Jr . Endurance training increases metabolic rate and norepinephrine appearance rate in older individuals. Am J Physiol. 1991;261:E233-E239.

111.

Poehlman

Dvorak

DeNino

Brochu

Ades

. Effects of resistance training and endurance training on insulin sensitivity in nonobese, young women: a controlled randomized trial. J Clin Endocrinol Metab. 2000;85:2463-2468.

112.

Raz

Israeli

Rosenblit

Bar-On

. Influence of moderate exercise on glucose homeostasis and serum testosterone in young men with low HDL-cholesterol level. Diabetes Res. 1988;9:31-35.

113.

Ross

Dagnone

Jones

. Reduction in obesity and related comorbid conditions after diet-induced weight loss or exercise-induced weight loss in men. A randomized, controlled trial. Ann Intern Med. 2000;133:92-103.

114.

Ross

Janssen

Dawson

. Exercise-induced reduction in obesity and insulin resistance in women: a randomized controlled trial. Obes Res. 2004;12:789-798.

115.

Ryan

Nicklas

Berman

. Aerobic exercise is necessary to improve glucose utilization with moderate weight loss in women. Obesity (Silver Spring). 2006;14:1064-1072.

116.

Schenk

. Effects of Exercise on Lipid-Induced Insulin Resistance [dissertation]. Ann Arbor: University of Michigan; 2006.

117.

Segal

Edano

Abalos

. Effect of exercise training on insulin sensitivity and glucose metabolism in lean, obese, and diabetic men. J Appl Physiol. 1991;71:2402-2411.

118.

Seip

Angelopoulos

Semenkovich

. Exercise induces human lipoprotein lipase gene expression in skeletal muscle but not adipose tissue. Am J Physiol. 1995;268:E229-E236.

119.

Short

Vittone

Bigelow

. Impact of aerobic exercise training on age-related changes in insulin sensitivity and muscle oxidative capacity. Diabetes. 2003;52:1888-1896.

120.

Sial

Coggan

Hickner

Klein

. Training-induced alterations in fat and carbohydrate metabolism during exercise in elderly subjects. Am J Physiol. 1998;274:E785-E790.

121.

Tanner

Koves

Cortright

. Effect of short-term exercise training on insulin-stimulated PI 3-kinase activity in middle-aged men. Am J Physiol Endocrinol Metab. 2002;282:E147-E153.

122.

Thamer

Machann

Stefan

. High visceral fat mass and high liver fat are associated with resistance to lifestyle intervention. Obesity (Silver Spring). 2007;15:531-538.

123.

Tymchuk

Tessler

Barnard

. Changes in sex hormone-binding globulin, insulin, and serum lipids in postmenopausal women on a low-fat, high-fiber diet combined with exercise. Nutr Cancer. 2000;38:158-162.

124.

van Aggel-Leijssen

Saris

Hul

van Baak

. Long-term effects of low-intensity exercise training on fat metabolism in weight-reduced obese men. Metab Clin Exp. 2002;51:1003-1010.

125.

van Aggel-Leijssen

Saris

Homan

van Baak

. The effect of exercise training on β-adrenergic stimulation on fat metabolism in obese men. Int J Obes. 2001;25:16-23.

126.

van Aggel-Leijssen

Saris

Wagenmakers

. The effect of low intensity exercise training on fat oxidation of obese women. Obes Res. 2001;9:86-96.

127.

Vermeulen

. Plasma lipid and lipoprotein levels in obese post-menopausal women: effects of a short-term low-protein diet and exercise. Maturitas. 1990;12:121-126.

128.

White

Ismail

Bradley

. Serum insulin and glucose response to graded exercise in adults—part II. The effect of exercise conditioning. Br J Sports Med. 1978;12:137-141.

129.

Wirth

Diehm

Hanel

Welte

Vogel

. Training-induced changes in serum lipids, fat tolerance, and adipose tissue metabolism in patients with hypertriglyceridemia. Atherosclerosis. 1985;54:263-271.

130.

Schmid

Stark

Berlin

Landais

Lau

. Meta-regression detected associations between heterogeneous treatment effects and study-level, but not patient-level, factors. J Clin Epidemiol. 2004;57:683-697.

131.

Kraus

Slentz

. Exercise training, lipid regulation, and insulin action: a tangled web of cause and effect. Obesity (Silver Spring). 2009;17(suppl 3):S21-S26.

132.

Lebovitz

. Insulin resistance—a common link between type 2 diabetes and cardiovascular disease. Diabetes Obes Metab. 2006;8:237-249.

133.

Lindeberg

Eliasson

Lindahl

Ahren

. Low serum insulin in traditional Pacific Islanders—the Kitava Study. Metab Clin Exp. 1999;48:1216-1219.

134.

Folsom

Szklo

Stevens

Liao

Smith

Eckfeldt

. A prospective study of coronary heart disease in relation to fasting insulin, glucose, and diabetes. The Atherosclerosis Risk in Communities (ARIC) study. Diabetes Care. 1997;20:935-942.

135.

Lakka

Laaksonen

. Physical activity in prevention and treatment of the metabolic syndrome. Appl Physiol Nutr Metab. 2007;32:76-88.

136.

Menshikova

Ritov

Toledo

Ferrell

Goodpaster

Kelley

. Effects of weight loss and physical activity on skeletal muscle mitochondrial function in obesity. Am J Physiol Endocrinol Metab. 2005;288:E818-E825.

137.

Praet

van Loon

. Optimizing the therapeutic benefits of exercise in type 2 diabetes. J Appl Physiol. 2007;103:1113-1120.

138.

Centers for Disease Control and Prevention. How much physical activity do adults need? 2012. http://www.cdc.gov/physicalactivity/everyone/guidelines/adults.html. Accessed December 26, 2013.

139.

Conn

Cooper

Ruppar

Russell

. Searching for the intervention in intervention research reports. J Nurs Scholarsh. 2008;40:52-59.

140.

Antuna-Puente

Disse

Rabasa-Lhoret

Laville

Capeau

Bastard

. How can we measure insulin sensitivity/resistance? Diabetes Metab. 2011;37:179-188.

Insulin Sensitivity Following Exercise Interventions

Abstract

Keywords

Introduction

Methods

Sample

Inclusion Criteria

Search Strategies

Study Selection

Data Collection Instrument and Process

Risk of Bias

Statistical Analyses

Results

Primary Study Characteristics and Quality Attributes

Overall Effects of Interventions

Exploratory Moderator Analyses

Discussion

Footnotes

Declaration of Conflicting Interests

Funding

Author Biographies

References