Cardiorespiratory fitness, muscular strength and risk of type 2 diabetes: a systematic review and meta-analysis

Aims/hypothesis The study aimed to quantitatively summarise the dose–response relationships between cardiorespiratory fitness and muscular strength on the one hand and risk of type 2 diabetes on the other and estimate the hypothetical benefits associated with population-wide changes in the distribution of fitness. Methods We performed a systematic review with meta-analysis. The PubMed and EMBASE electronic databases were searched from inception dates to 12 December 2018 for cohort studies examining the association of cardiorespiratory fitness or muscular strength with risk of incident type 2 diabetes in adults. The quality of included studies was evaluated using the Newcastle–Ottawa Scale. Results Twenty-two studies of cardiorespiratory fitness and 13 studies of muscular strength were included in the systematic review with both exposures having ten estimates available for the primary adiposity- or body size-controlled meta-analysis. In random-effects meta-analysis including 40,286 incident cases of type 2 diabetes in 1,601,490 participants, each 1 metabolic equivalent (MET) higher cardiorespiratory fitness was associated with an 8% (95% CI 6%, 10%) lower RR of type 2 diabetes. The association was linear throughout the examined spectrum of cardiorespiratory fitness. In 39,233 cases and 1,713,468 participants each 1 SD higher muscular strength was associated with a 13% (95% CI 6%, 19%) lower RR of type 2 diabetes. We estimated that 4% to 21% of new annual cases of type 2 diabetes among 45–64-year-olds could be prevented by feasible and plausible population cardiorespiratory fitness changes. Conclusions/interpretation Relatively small increments in cardiorespiratory fitness and muscle strength were associated with clinically meaningful reductions in type 2 diabetes risk with indication of a linear dose–response relationship for cardiorespiratory fitness. Registration: PROSPERO (CRD42017064526). Electronic supplementary material The online version of this article (10.1007/s00125-019-4867-4) contains peer-reviewed but unedited supplementary material, which is available to authorised users.

ESM Figure 3. Study-specific relative risks per standard deviation increase in muscular strength in models not controlling for adiposity ESM Figure 4. Risk of small-study bias visualized by funnel-plot of cardiorespiratory fitness estimates including control for adiposity. ESM Figure 5. Risk of small-study bias visualized by funnel-plot of cardiorespiratory fitness estimates excluding control for adiposity. ESM Figure 6. Risk of small-study bias visualized by funnel-plot of muscular strength estimates including control for adiposity ESM Figure 7. Risk of small-study bias visualized by funnel-plot of muscular strength estimates excluding control for adiposity

ESM Table 3. Inclusion and exclusion criteria Component Inclusion criteria Exclusion criteria
Population Studies that include human subjects free of type 2 diabetes at baseline. Cohorts will be included if they consist of participants with conditions that are associated with type 2 diabetes (e.g. obesity, metabolic syndrome, cardiovascular diseases) Studies not excluding subjects with type 2 diabetes at baseline, studies with a population that consists exclusively of individuals with a chronic disease (e.g. cancer).
Exposure Cardiorespiratory fitness* assessed by a maximal or sub-maximal stress test of any form Muscular strength** measured as peak score or mean score. Composite scores including >1 unique test will be included. Both isotonic, isometric and isokinetic tests will be included. There are no criteria regarding muscle groups tested. Tests should allow few (<3) repetitions of a task before reaching momentary muscular fatigue Muscular power*** or endurance**** Outcome type 2 diabetes Not type 2 diabetes (also excluding pre-diabetes).

Study design Cohort studies
Experimental studies, case-control studies, cross-sectional studies, metaanalyses, reviews, reports Other Published in English or Scandinavian language.
Any publication year Other languages For metaanalysis Results should be provided with relative risk, hazard ratio or odds ratio, and corresponding confidence intervals or information to calculate variance associated with estimates. Estimates should be convertible to the unit size found most appropriate for harmonization Insufficient information/not possible to convert estimates to chosen unit for harmonization *Cardiorespiratory fitness is the ability to perform large muscle, dynamic, moderate-vigorous intensity activity for prolonged periods [1]. **Muscular strength is the ability of a muscle to exert maximal force [1]. ***Muscular power is the muscle's ability to exert force per unit of time [1]. ****Muscular endurance is the ability of a muscle to continue to perform without fatigue [1] ESM Table 4. Newcastle-Ottawa quality score of prospective cohort studies of cardiorespiratory fitness and incident type 2 diabetes

Study Selection Comparability of cohorts Outcome Stars awarded
Selection of the non-exposed cohort Ascertainment of exposure We chose not to include the "Representativeness of the exposed cohort" item of the original Newcastle-Ottawa Scale [24] since we find this irrelevant for the evaluation of the internal validity of the studies. Thus, a total of 8 stars were achievable. Study quality reflects assessments in relation to the estimates for which we extracted data and not the study per se. We chose not to include the "Representativeness of the exposed cohort" item of the original Newcastle-Ottawa Scale [24] since we find this irrelevant for the evaluation of the internal validity of the studies. Thus, a total of 8 stars were achievable. Study quality reflects assessments in relation to the estimates for which we extracted data and not the study per se.  [55]. b" Feasible minimum risk". c "Plausible minimum risk". d PAFs [58] for low cardiorespiratory fitness were calculated by defining the bottom 25% of the population CRF distribution as unfit (<8.4 METs would be classified as unfit for men whereas women with a CRF <6.0 METs would be classified as unfit) based on the U.S. FRIEND database at 40-59 years of age. We then estimated the proportion of total diabetes cases which could theoretically be prevented by changing the cardiorespiratory fitness level of all unfit adults to the fitness level matching the distribution of the population of "fit" individuals (≥25th percentile). RR's were based on a contrast between the fitness level of the sex-specific 12.5th percentile (the midpoint of the 1st to 25th percentile interval) and the 62.5th percentile (the midpoint of the 25th to 99th percentile) estimated from the restricted cubic spline model. This analysis is comparable to conventional PAF calculations based on eliminating the exposure and "shifting" exposed individuals into matching the distribution of the "non-exposed" reference category (above the sex-specific MET cut-points as specified above). As the PIF is calculated based on a distributional change, rather than complete elimination, it may be preferable over PAFs in the case of a continuous exposure were the minimum risk is achieved at a non-zero exposure level [59]. CRF; cardiorespiratory fitness, PIF; potential impact fraction, PAF; population attributable fraction.  Figure 4. Risk of small-study bias visualized by funnel-plot of cardiorespiratory fitness estimates including control for adiposity ESM Figure 5. Risk of small-study bias visualized by funnel-plot of cardiorespiratory fitness estimates excluding control for adiposity ESM Figure 6. Risk of small-study bias visualized by funnel-plot of muscular strength estimates including control for adiposity ESM Figure 7. Risk of small-study bias visualized by funnel-plot of muscular strength estimates excluding control for adiposity