Skip to main content
NCBI home page
The Journal of Clinical Endocrinology and Metabolism logoLink to The Journal of Clinical Endocrinology and Metabolism
. 2024 Jun 12;110(5):1451–1459. doi: 10.1210/clinem/dgae393

Cardiorespiratory Fitness, Body Composition, Diabetes, and Longevity: A 2-Sample Mendelian Randomization Study

Alisa D Kjaergaard 1,2,✉, Christina Ellervik 3,4,5,6, Niels Jessen 7,8, Sarah J Lessard 9,10
PMCID: PMC12012764  PMID: 38864459

Abstract

Context

Cardiorespiratory fitness, commonly assessed as maximal volume of oxygen consumption (VO2max), has emerged as an important predictor of morbidity and mortality.

Objective

We investigated the causality and directionality of the associations of VO2max with body composition, physical activity, diabetes, performance enhancers, and longevity.

Methods

Using publicly available summary statistics from the largest genome-wide association studies publicly available, we conducted a bidirectional 2-sample Mendelian randomization (MR) study. Bidirectional MR tested directionality, and estimated the total causal effects, whereas multivariable MR (MVMR) estimated independent causal effects. Cardiorespiratory fitness (VO2max) was estimated from a submaximal cycle ramp test (N ≈ 70 000) and scaled to total body weight, and in additional analyses to fat-free mass (mL/min/kg).

Results

Genetically predicted higher (per 1 SD increase) body fat percentage was associated with lower VO2max (β = −0.36; 95% CI: −0.40, −0.32, P = 6 × 10–77). Meanwhile, genetically predicted higher appendicular lean mass (β = 0.10; 95% CI: 0.08 to 0.13), physical activity (β = 0.29; 95% CI: 0.07 to 0.52), and performance enhancers (fasting insulin, hematocrit, and free testosterone in men) were all positively associated with VO2max (all P < .01). Genetic predisposition to diabetes had no effect on VO2max. MVMR showed independent causal effects of body fat percentage, appendicular lean mass, physical activity, and hematocrit on VO2max, as well as of body fat percentage and type 2 diabetes (T2D) on longevity. Genetically predicted VO2max showed no associations.

Conclusion

Cardiorespiratory fitness can be improved by favorable body composition, physical activity, and performance enhancers. Despite being a strong predictor of mortality, VO2max is not causally associated with T2D or longevity.

Keywords: Mendelian randomization analysis; obesity; diabetes mellitus, type 2; cardiorespiratory fitness; longevity; sex hormone–binding globulin

Cardiorespiratory fitness pertains to how effectively the circulatory and respiratory systems work together to supply muscles with oxygen during prolonged physical activity. Direct measurement is obtained as maximal oxygen consumption during an exhaustive maximal exercise test, commonly referred to as VO2max. However, in large-scale studies, cost and safety concerns often necessitate the estimation of VO2max from heart rate responses during submaximal exercise tests.

Cardiorespiratory fitness has emerged as an important predictor of morbidity and mortality (1). Previous studies, including large meta-analyses, have shown that better cardiorespiratory fitness is associated with decreased risk of type 2 diabetes (T2D) and all-cause mortality (2-6). A systematic review and a meta-analysis of 160 randomized controlled trials (RCTs) including 7487 participants, showed that exercise improves VO2max as well as glycemic traits (fasting insulin and glycated hemoglobin A1c [HbA1c]) (7). Additionally, performance enhancers, like anabolic steroids and blood doping, can augment cardiorespiratory fitness by increasing muscle mass or oxygen availability (8, 9). Insulin also plays a role in enhancing performance by promoting glycogen storage in muscles and preventing muscle breakdown when used with anabolic steroids (10). Physical activity and cardiorespiratory fitness are closely related to body composition, impacting fat and muscle distribution (11).

Despite substantial clinical evidence that low VO2max can predict mortality (1), the causality and directionality of the associations of VO2max with body composition, diabetes and glycemic traits, performance enhancers, physical activity, and longevity are unclear. We aimed to address these uncertainties. For this purpose, we employed a 2-sample Mendelian randomization (2SMR) approach.

MR is a powerful method that integrates information on genetic variants (typically single nucleotide polymorphisms [SNPs]) with observational epidemiological data to strengthen causal inference without intervention. The method name, MR, refers to Mendel's laws of heritability and the analogy with RCTs. Because potential confounders are evenly distributed across genotype, which is established at conception (Mendel’s laws of independent assortment and segregation), confounding and reverse causation are largely circumvented (RCT analogy). The MR method assesses the direction and magnitude of the effect of the genetically predicted life-long exposure on the outcome. However, interpreting MR estimates causally relies on the gene–environment equivalence assumption. This assumption stipulates that increasing the exposure levels will have the same effect on the outcome, irrespective of whether the exposure is influenced by genetic variants or environmental factors (or interventions).

In the present bidirectional MR study, using publicly available genome-wide association study (GWAS) summary statistics data from UK Biobank and international consortia, we examined the causality and directionality of the associations of VO2max (estimated from a submaximal exercise test) with body composition, diabetes and glycemic traits, performance enhancers (eg, insulin, hematocrit, and testosterone), physical activity, and longevity. We hypothesized that the observational association between VO2max and longevity is not causal, but explained by the shared causal risk factors, notably adverse body composition and low physical activity levels.

Materials and Methods

We employed a 2SMR method that leverages publicly available summary data (eg, β-coefficients and SEs) on SNP-exposure and SNP-outcome associations from 2 different GWASs. The 3 core assumptions of the method are that the SNPs used as genetic instruments (1) are strongly associated with the exposure, (2) are not associated with confounders, and (3) are associated with outcome only through the exposure.

For the present study, we used summary data from several (European ancestry only) GWAS consortia (eg, GIANT: Genetic Investigation of ANthropometric Traits, DIAGRAM: DIAbetes Genetics Replication And Meta-analysis, MAGIC: Meta-Analyses of Glucose and Insulin-related traits Consortium, BCX: Blood Cell Consortium, and LifeGen), many of which included UK Biobank participants (Table S1 (12)). In addition to sex-combined, we used sex-stratified GWASs when available.

Data Sources

SNPs used as genetic instruments were identified from published studies reporting top independent SNPs in main or supplementary tables, or links to online repositories (Table S1 (12)).

We extracted SNP-exposure summary statistics on SNPs that were robustly associated with the exposure of interest (P < 5 × 10–8) and excluded SNPs in linkage disequilibrium (within a 10-Mb window and R2 cut-off at 0.001, ie, standard MR clumping). However, we relaxed this strictness slightly (to P < 5 × 10–6) for sex-stratified physical activity, due to the small sample size and very few (or none) genetic instruments otherwise.

SNP-outcome summary statistics were then extracted from outcome GWASs. Not all genetic instruments were available in all outcome GWASs, so we used proxies (SNPs in linkage disequilibrium at r2 > 0.8) when possible, which is why the number of SNPs varies slightly for different exposure–outcome combinations.

Body Composition

The body composition measures examined were total body weight, total body fat (percentage and mass) and fat-free mass, trunk fat (percentage and mass) and fat-free mass, body mass index (BMI), waist–hip ratio (WHR), WHR adjusted for BMI (WHRadjBMI, ie, unexpected fat distribution), and appendicular lean mass (hereafter simply referred to as lean mass).

In UK Biobank, body composition was determined using Tanita BC418MA body composition analyzer, which by passing a safe low electrical current through the body can distinguish body fat from muscle due to speed of the current being faster through muscle (higher water percentage) than fat.

Sex-stratified body composition measures were total body weight, total body fat percentage, BMI, WHR, WHRadjBMI, lean mass, and additionally allometric body shape indices: A Body Shape Index (ABSI), Waist–Hip Index (WHI), and Hip Index (HI) (13-15). Allometric refers to the proportional relationships between different dimensions of an organism. Thus, ABSI, WHI, and HI are allometric (scaled to body weight and height) counterparts of waist circumference, WHR and hip circumference, respectively (15). Notably, while high ABSI and WHI represent adverse, high HI represents favorable body composition.

Physical Activity Measures

The largest meta-GWAS (n > 500 000) on physical activity was based on self-reported outcomes: moderate to vigorous physical activity (yes/no) was a binary trait, whereas leisure screen time was a continuous trait of self-reported time spent watching TV, playing videogames, and sitting at the computer, etc. (16).

Because self-reported physical activity is prone to misclassification and bias by recall and awareness of the beneficial effects of physical activity, we also examined device-measured physical activity in a smaller subset (n ≈ 90 000) of UK Biobank participants (17-19). Average physical activity over a 7-day period was derived from an Axivity AX3 wrist-worn activity tracker (18), whereas sedentary time (https://biobank.ndph.ox.ac.uk/showcase/label.cgi? id=1020) was estimated using a machine-learning method on the accelerometer data (17). Sex-stratified physical activity was the accelerometry-based 7-day average physical activity.

Diabetes, Glycemic Traits, and Performance Enhancers

For diabetes, we used GWASs of European ancestry without UK Biobank participants. Type 1 diabetes GWAS included 7467 cases and 10 218 controls (20), and T2D GWAS included 55 005 cases and 400 308 controls (21). GWASs on fasting glucose, glucose tolerance (plasma glucose measured 120 minutes after an intake of 75 g of glucose following an overnight fast), HbA1c, and fasting insulin did not include individuals with diabetes (22, 23).

Hemoglobin and hematocrit were rank-based inverse normalized. Sex-combined summary statistics were from the BCX Consortium (including UK Biobank and several smaller studies), whereas sex-stratified summary statistics were from a UK Biobank GWAS provided by the Neale laboratory (http://www.nealelab.is/uk-biobank/). Total and free (ie, bioavailable) testosterone were from UK Biobank (24). Free testosterone was calculated from the measured levels of total testosterone, sex hormone–binding globulin (SHBG) and albumin by the Vermeulen equation (https://biobank.ndph.ox.ac.uk/showcase/showcase/docs/serum_biochemistry.pdf) (24).

Sex-stratified traits were T2D, fasting glucose, fasting insulin, hemoglobin, hematocrit, and (total and free) testosterone (Table S1 (12)).

Cardiorespiratory Fitness

Cardiorespiratory fitness was assessed as a maximum workload relative to total body weight during a submaximal cycle ramp test (N = 70 783) using a stationary bike and 4-lead electrocardiograph to record heart rate (18). This was used as an estimate of the participants’ maximum rate of oxygen consumption (volume per minute), and scaled to total body weight (in kilograms). The test procedure is provided (https://biobank.ctsu.ox.ac.uk/crystal/crystal/docs/Cardio.pdf), and calculations described elsewhere (18). In addition to sex-combined, sex-stratified summary statistics were also available.

Although scaling VO2max to total body weight allows for comparison between people of different sizes, it does not fully account for different body composition (fat and muscle distribution), which is important since it is primarily the muscle that utilizes the oxygen during physical activity. Unfortunately, there are currently no GWASs with publicly available full summary statistics for alternative VO2max scaling. However, a recent study performed a GWAS on VO2max scaled to fat-free mass (VO2maxFFM), but only published top SNPs (in supplementary data) and not full GWAS summary statistics (6). Thus, we were only able to examine VO2maxFFM as exposure, but not as outcome. Furthermore, that study (6) combined 14 GWAS significant (P < 5 × 10–8) variants associated with VO2maxFFM (in 69 416 UKB participants) with 148 GWAS significant variants associated with resting heart rate (in 452 941 UKB participants) that were also nominally (P < .05) associated with VO2maxFFM (in 69 416 UKB participants), yielding 160 variants in total. After restricting these 14 and 160 genetic variants to SNPs only (Table S2 (12)) and to standard MR clumping, we identified 2 sets of instruments for VO2maxFFM: (1) a strict set of 12 GWAS significant SNPs and (2) a broad set of 104 SNPs with a less robust (P < .05) SNP-VO2maxFFM association. For consistency, we also identified 2 sets of instruments for VO2max (scaled to total body weight): (1) a strict set of 9 GWAS significant SNPs and (2) a broad set of 44 SNPs with a less robust (P < 5 × 10–6) SNP-VO2max association, both subjected to standard MR clumping.

Longevity

Longevity was assessed as parental survival in a sample of 1 012 240 parents (60% deceased) of European ancestry from the UK Biobank and the LifeGen Consortium of 26 cohorts (25). Under the assumption of common effect sizes between fathers and mothers, the β-coefficients were expressed as log hazard protection ratio (ie, negation of log hazard ratio) for carrying 1 copy of the effect allele in self.

Ethics

We used publicly available data only, based on GWASs that have collected relevant ethics approvals.

Statistical Analysis

Analyses were performed in R version 4.0.4 using TwoSampleMR (26, 27) version 0.5.6 and MendelianRandomization (28) version 0.6.0 packages. The latter was used for performing MR-Lasso and calculating I2GX.

We did not correct for multiple comparisons, nor consider P < .05 as the sole evidence of an association. Rather, we performed an overall evaluation of each result by considering the effect size, SE, biological plausibility and consistency across the examined exposures, outcomes, as well as main and sensitivity analyses.

Prior to MR analyses, the data were harmonized by aligning the beta-coefficients of the SNP-exposure and SNP-outcome summary statistics to the effect allele associated with increased level of exposure. Inconsistent and palindromic SNPs with effect allele frequencies close to 50% were excluded. All SNPs except for SNPs for VO2maxFFM had sufficient strength (F > 10) to be used as genetic instruments, calculated as F = β2exposure/SE2exposure. For VO2maxFFM, only 64 out of 104 SNPs had F > 10.

Individual causal estimates, calculated as Wald ratios of the SNP-outcome and SNP-exposure association for each SNP, were meta-analyzed into an overall total causal estimate for each exposure–outcome combination (29). The main MR analysis was the multiplicative inverse variance weighting (IVW) meta-analysis. Heterogeneity was assessed as Cochran's Q and corresponding I2 of the individual causal estimates (30).

Because the IVW MR method assumes that all SNPs are valid genetic instruments, we also performed sensitivity MR analyses with other, more relaxed, assumptions regarding SNP validity and pleiotropy: MR-Lasso, weighted median MR, and MR-Egger regression analyses. MR-Lasso provides IVW estimates after exclusion of SNPs identified as invalid by the lasso procedure (31). Weighted median MR requires that at least 50% of the weight contributed by SNPs comes from valid genetic instruments (32). Meanwhile, MR-Egger is reliable (but imprecise) in cases of substantial pleiotropy (33). We assessed directional pleiotropy by MR-Egger intercept test, and quantified the violation of the “no measurement error” assumption by the I2GX (33). Multivariable MR (MVMR), based on the IVW method, estimated the independent (adjusted) causal effects of pairwise combinations of following exposures: body composition, physical activity, T2D, and performance enhancers, on VO2max and longevity. Specifically, MVMR used SNPs associated with each pair of exposures (subjected to standard MR clumping) to jointly estimate the independent (adjusted) causal effects of each of the exposures on the outcome of interest. For tutorial, please see https://marinalearning.netlify.app/2021/03/22/setting-up-multivariable-mendelian-randomization-analysis/.

Directionality of potential causal associations was examined by bidirectional MR, MR Steiger (true/false), and additional MR analyses were performed after Steiger filtering. While Steiger filtering excludes SNPs that explain more of the variance in outcomes than the exposures, MR Steiger uses all SNPs to assess whether the causal direction of all SNPs combined is true (all SNPs combined explain more variance in the exposure than the outcome) or false (all SNPs combined explain more variance in the outcome than the exposure).

Although, 2SMR methods can be safely used for 1-sample MR performed within large biobanks (34), such as the UK Biobank, it is optimal to use 2 nonoverlapping, but comparable (with regards to sample size, age, sex, and ancestry) population samples. However, additional GWASs without UK Biobank participants were only available for 3 body composition traits (body fat percentage, BMI, and WHR) and diabetes (type 1 diabetes and T2D). Thus, for the 3 body composition traits, the main MR and MVMR included, and the sensitivity and bidirectional MR analyses excluded UK Biobank participants. For diabetes, UK Biobank participants were excluded, as was done in a recent study (6) that found a causal association between cardiorespiratory fitness (VO2maxFFM) and T2D.

Results

Causal Effects of Body Composition on VO2max

Genetically predicted body composition was associated with VO2max. The strongest association was observed for total body fat percentage: 1 SD (≈8.5%) higher genetically predicted body fat percentage was associated with 0.361 mL/kg/min lower VO2max (95% CI −0.408 to −0.314, P = 2 × 10–51, Fig. 1; Table S3 (12)). Corresponding estimates were −0.105 (−0.142 to −0.069, P = 2 × 10–08) per 1 SD (≈16 kg) higher weight, −0.283 (−0.324 to −0.241, P = 5 × 10–41) for 1 SD (8%) higher trunk fat percentage, −0.185 (−0.220 to −0.149, P = 4 × 10–24) for 1 SD (≈4.8 kg/m2) higher BMI, and −0.074 (−0.122 to −0.025, P = .003) for 1 SD (0.09) higher WHR (Fig. 1; Table S3 (12)). Meanwhile, 1 SD higher genetically predicted lean mass (≈3.2 kg) was associated with 0.105 mL/kg/min (0.080 to 0.130, P = 3 × 10–16) higher VO2max (Fig. 1; Table S4 (12)).

Figure 1.

Figure 1.

Open in a new tab

Causal effects of body composition on estimated VO2max. Estimates (β-coefficients and 95% CIs) are from the random-effects inverse variance weighted Mendelian randomization analysis, and expressed in log units per SD increase in the relevant exposure. For each exposure, the number of single nucleotide polymorphisms (SNPs) included in the analysis is shown in parenthesis. (Total body) weight, body fat %, trunk fat %, and lean mass were from the UK Biobank. Body mass index (BMI), waist–hip ratio (WHR) and WHR adjusted for BMI (WHRadjBMI) were from the GIANT (Genetic Investigation of ANthropometric Traits) Consortium, and included UK Biobank participants. Estimated VO2max, a proxy of cardiorespiratory fitness, was based on a submaximal cycle ramp test (in 70 783 UK Biobank participants) and estimated as the participants’ maximum volume of oxygen consumption, per kilogram of body weight, per minute. Sensitivity MR analyses are shown elsewhere (Tables S3-S6; https://github.com/AlisaDK/Cardiorespiratory-fitness).

All these associations, except for WHR, were robust to sensitivity MR analyses (Tables S4-S6 (12)). Sex-stratified and sex-combined analyses were largely concordant (Fig. 1; Fig. S1, Tables S7 and S8 (12)). Collectively, these results show that genetically predicted favorable body composition (low body fat percentage and high muscle mass) is associated with higher VO2max.

However, we observed the opposite association pattern for WHRadjBMI and WHI in women only and for HI in both sexes: genetically predicted higher WHRadjBMI and WHI were associated with higher VO2max in women, and HI with lower VO2max in both sexes (Fig. S1, Tables S7 and S8 (12)). ABSI was not associated with VO2max (Fig. S1, Tables S7 and S8 (12)).

Causal Effects of Diabetes and Glycemic Traits on VO2max

Genetic predisposition to diabetes (type 1 and type 2) and genetically predicted glycemic traits (fasting glucose, glucose tolerance and HbA1c) were not associated with VO2max (Fig. 2 left panel; Tables S3 and S4 (12)).

Figure 2.

Figure 2.

Open in a new tab

Causal effects of glycemic traits (left panel) and performance enhancers (right panel) on estimated VO2max. Estimates (β-coefficients and 95% CIs) are from the inverse variance weighted random-effects Mendelian randomization analysis. For each exposure, the number of single nucleotide polymorphisms (SNPs) included in the analysis is shown in parenthesis. Diabetes (type 1 and 2) cohorts did not include UK Biobank participants. Fasting glucose, glucose tolerance (oral glucose tolerance test, ie, plasma glucose measured 120 minutes after an intake of 75 g of glucose after an overnight fast), HbA1c, and fasting insulin were from the MAGIC (Meta-Analyses of Glucose and Insulin-related traits Consortium), and did not include UK Biobank participants. Hemoglobin, hematocrit, total and free (ie, bioavailable) testosterone were from the UK Biobank. Estimated VO2max, a proxy of cardiorespiratory fitness, was based on a submaximal cycle ramp test (in 70 783 UK Biobank participants) and estimated as the participants’ maximum volume of oxygen consumption, per kilogram of body weight, per minute. Sensitivity MR analyses are shown elsewhere (Tables S3 and S4; https://github.com/AlisaDK/Cardiorespiratory-fitness).

Causal Effects of Performance Enhancers on VO2max

Fasting insulin, hemoglobin, hematocrit, and free testosterone in men (but not in women) were all positively associated with VO2max (Fig. 2 right panel; Tables S3 and S4 (12)).

Causal Effects of Physical Activity on VO2max

As expected, self-reported and accelerometry-measured physical activity was associated with higher VO2max and sedentary behavior with lower VO2max (Fig. S2, Tables S3 and S4 (12)).

Multivariable Mendelian Randomization

Adjusting for body fat percentage, attenuated the causal effects of BMI and WHR on VO2max, but not vice versa (Fig. 1; Table S9 (12)). Interestingly, while genetically predicted higher body weight was associated with decreased VO2max in (univariable) MR analysis (β = −.105, P = 2 × 10–08, Fig. 1), adjusting for body fat percentage in MVMR flipped the direction of the association (β = .167, P = 8 × 10–8, Table S9 (12)). A likely explanation for this is that high body weight adjusted for body fat percentage may reflect high lean mass.

MVMR analyses showed that the causal effects of body fat percentage, lean mass, physical activity and hematocrit on VO2max were independent of each other (Fig. 3; Table S9 (12)). The causal effects of fasting insulin on VO2max were abrogated by adjusting for body fat percentage, but not vice versa (Fig. 3; Table S9 (12)).

Figure 3.

Figure 3.

Open in a new tab

Total and direct causal effects of body composition, diabetes, fasting insulin and physical activity on estimated VO2max and longevity. Estimates (β-coefficients and 95% CIs) are from the inverse variance weighted (IVW) Mendelian randomization (MR) analysis. Univariable (random-effects IVW) MR estimated total (unadjusted) causal effects. Multivariable MR, also based on IVW, estimated direct (adjusted) causal effects of each of the 5 exposures on VO2max (left panel) and longevity (right panel), while accounting for the other 4 exposures, 1 at a time. The 5 exposures were body fat %, lean mass, and moderate to vigorous physical activity (MVPA) from the UK Biobank; type 2 diabetes (T2D) from the DIAGRAM (DIAbetes Genetics Replication And Meta-analysis) Consortium without UK Biobank participants; and fasting insulin from the MAGIC (Meta-Analyses of Glucose and Insulin-related traits Consortium) without UK Biobank participants. For each exposure, the number of single nucleotide polymorphisms (SNPs) included in the analysis is shown in parenthesis. Estimated VO2max, a proxy of cardiorespiratory fitness, was based on a submaximal cycle ramp test (in 70 783 UK Biobank participants) and estimated as the participants’ maximum volume of oxygen consumption, per kilogram of body weight, per minute. Longevity was assessed as parental survival from the LifeGen Consortium, including UK Biobank. Under the assumption of common effect sizes between fathers and mothers, the β-coefficients were expressed as log hazard protection ratio (ie, negation of log hazard ratio) for carrying 1 copy of the effect allele in self. Related MR analyses are shown elsewhere (Table S9; https://github.com/AlisaDK/Cardiorespiratory-fitness).

Longevity

While body fat percentage, lean mass, T2D, and fasting insulin were causally associated with longevity, VO2max was not (Tables S3, S4 S10, and S11 (12); Figs. 3 and 4). Furthermore, MVMR analyses showed that the association with longevity disappeared for lean mass after adjusting for body fat percentage, and for fasting insulin after adjusting for T2D (Fig. 3; Table S9 (12)). The association with moderate to vigorous physical activity was less robust, but supported by sensitivity MR (Tables S2 and S3 (12)) and MVMR analyses (Fig. 3) as well as complementary leisure screen time (Tables S3 and S4 (12)).

Figure 4.

Figure 4.

Open in a new tab

Causal effects of estimated VO2max on body composition, diabetes, fasting insulin, and physical activity. Estimates (β-coefficients and 95% CIs) are from the inverse variance weighted random-effects MR analysis. We used 4 different sets of genetic instruments: a strict and a broad set of single nucleotide polymorphisms (SNPs) for estimated VO2max scaled to total body weight as well as fat-free mass. For each set of instruments, the number of SNPs included in the analysis is shown in parenthesis. The outcomes were body fat %, lean mass, and moderate to vigorous physical activity (PA) from the UK Biobank; type 2 diabetes from the DIAGRAM (DIAbetes Genetics Replication And Meta-analysis) Consortium without UK Biobank participants; fasting insulin from the MAGIC (Meta-Analyses of Glucose and Insulin-related traits Consortium) without UK Biobank participants, and longevity (parental survival) from the LifeGen Consortium, including UK Biobank. Sensitivity MR analyses are shown elsewhere (Tables S10 and S11; https://github.com/AlisaDK/Cardiorespiratory-fitness).

Causal Effects of VO2max

Genetically predicted VO2max was not associated with body composition, diabetes, glycemic traits, performance enhancers, and physical activity (Fig. 4; Tables S10 and S11 (12)). Specifically, we found no robust support for an association with fasting insulin, as reported by a recent study (6) (on VO2maxFFM). Collectively, this indicates that VO2max is more likely a consequence of favorable body composition and physical activity than vice versa.

Discussion

In this 2SMR study, employing publicly available summary statistics from the largest GWASs, we discovered that although VO2max is a strong predictor of both T2D and mortality in observational studies, it is not causally associated with the development of T2D or longevity (Fig. 5). Our findings highlight potential strategies for improving VO2max, including promoting favorable body composition, characterized by low body fat percentage and high muscle mass, increasing levels of physical activity, and the use of performance enhancers. Furthermore, we identified modifiable risk factors such as body fat percentage, diabetes, and physical activity, that demonstrate a causal association with longevity.

Figure 5.

Figure 5.

Open in a new tab

Summary of the main results: causality and directionality of the associations of VO2max with body composition, physical activity, diabetes, performance enhancers, and longevity. Estimated maximum volume of oxygen consumption during a (sub)maximal exercise test, or VO2max, is a common proxy for cardiometabolic fitness, and can be improved by favorable body composition (low body fat % and high lean mass), physical activity, and performance enhancers. Despite being a strong predictor of morbidity and mortality, VO2max is not causally associated with body composition, physical activity, performance enhancers, diabetes or longevity. Modifiable risk factors causally associated with longevity were: body fat %, diabetes, and physical activity.

In contrast to our results, a recent study found that genetically predicted higher VO2maxFFM was associated with decreased risk of T2D, partly mediated by higher SHBG levels (6). This is in concordance with a previous meta-analysis of 15 studies (27 657 T2D cases and 58 481 controls) showing that SHBG rs1799941 A-allele (associated with 0.2 SD higher SHBG levels) was associated with decreased T2D risk (35). However, we could not replicate this finding, as rs1799941 A-allele was not associated with T2D (P > .40) in the largest publicly available European T2D GWASs (>50 000 cases and >400 000 controls) (21, 36). On the other hand, we and others (37) have observed that a moderately correlated (R2 = 0.26) SHBG rs858519 C-allele (associated with 0.1 SD higher SHBG levels) was associated with decreased T2D risk (P < .001). This is possibly due to its association with body composition and linkage disequilibrium with SNPs in the ATP1B2 (38-40). Moreover, in a post hoc bidirectional MR analysis (data not shown), we observed that obesity (assessed as BMI) and T2D were more likely causes of lower SHBG levels than vice versa. Thus, obesity is causally associated with increased T2D risk as well as lower SHBG levels. This may explain why a recent phenome-wide MR study found that genetically predicted higher SHBG adjusted for BMI was associated with decreased T2D risk (41). Collectively, this suggests that the link between SHBG and T2D is obesity. Finally, we found no association (P > .03, and/or IVW and MR-Egger estimates in the opposite directions) between genetically predicted higher VO2max (all 4 SNP sets) and SHBG levels. This was similar to the results from a recent study (6) when the full set of VO2maxFFM SNPs were used (Cai et al (6), Supplementary Table S6 for T2D and Supplementary Table S7 for SHBG as outcomes). However, that study (6) emphasized the results from analyses using radial filtered VO2maxFFM SNPs, which supported causality. It is worth noting, as the authors also point out, that their (radial filtered) results need to be interpreted with caution, as they may be “driven by the selected resting heart rate-associated variants, given both observational and MR studies have found a significant positive association between resting heart rate and type 2 diabetes (42, 43)”. In fact, the referenced GWAS and MR study (43) points to shared etiology between resting heart rate and T2D, as illustrated by significant genetic correlations and bidirectional causality between resting heart rate and T2D. Moreover, the largest meta-GWAS of 100 studies (n = 835 465) found no association between resting heart rate and all-cause mortality (44).

Thus, the collective evidence points to obesity and physical activity, rather than the correlated resting heart rate and VO2max, as the actually modifiable causal risk factors for T2D and longevity. Indeed, it is difficult to imagine an intervention targeting VO2max through pathways independent from body composition and physical activity.

Meanwhile, genetically predicted higher free (ie, not bound to SHBG) testosterone, unlike SHBG (post hoc analysis, data not shown), was associated with higher VO2max. Together with the fact that genetically predicted higher SHBG was associated with lower testosterone levels (post hoc analysis, data not shown), this suggests that the causal pathway from testosterone to VO2max is independent from SHBG.

Causal interpretation of our results seems justified, as we showed that the effects of genetically predicted higher (fasting) insulin, hemoglobin, hematocrit, physical activity, and (free) testosterone on VO2max were equivalent to environmental influences such as doping (performance enhancers) and exercise.

The present study showed internal consistency across correlated/complementary exposures (body composition, diabetes and glycemic traits, hemoglobin and hematocrit, and physical activity), as well as across main and sensitivity analyses.

The unexpected association pattern of WHRadjBMI and allometric body shape indices with VO2max is possibly attributed to the inherent characteristics of VO2max, when scaled to total body weight. Specifically, while 2 individuals may have identical raw VO2max values (mL O2/min), the individual with greater weight would have a lower VO2max when scaled to body weight (mL O2/min/kg). Unfortunately, the exploration of alternative scaling methods for VO2max was not possible due to the unavailability of corresponding GWASs with full summary statistics publicly available, as detailed in “Materials and Methods.”

An inherent limitation of the 2SMR approach is its assumption of a linear relationship between exposure and outcome across varying exposure levels. As we only had access to summary statistics and not individual-level data, we were unable to explore potential nonlinearity. While our use of European ancestry GWASs may have somewhat restricted the generalizability of our findings, it was essential to minimize the risk of population stratification bias. Nevertheless, we advocate for future studies to validate and expand upon our findings across diverse ancestries. Another potential limitation that may have biased our estimates is the substantial overlap among the included GWASs, as most of these included UK Biobank participants. We addressed this issue as outlined in “Materials and Methods.”

In conclusion, our study shows that although VO2max is a strong predictor of both T2D and mortality in observational studies, it is not causally associated with the development of T2D or longevity. The observational association of VO2max with T2D and mortality is likely explained by their shared causal risk factors, namely high body fat percentage and low physical activity levels. These findings emphasize the importance of achieving and maintaining favorable body composition and engaging in regular physical activity for better health outcomes.

Acknowledgments

We thank all the consortia (UK Biobank, GIANT, DIAGRAM, MAGIC, BCX, and LifeGen) and the research community for providing publicly available summary statistics used in this study.

Abbreviations

2SMR

2-sample Mendelian randomization

ABSI

A Body Shape Index

BMI

body mass index

GWAS

genome-wide association study

HbA1c

glycated hemoglobin A1c

HI

Hip Index

IVW

inverse variance weighting

MR

Mendelian randomization

MVMR

multivariable MR

RCT

randomized controlled trial

SHBG

sex hormone–binding globulin

SNP

single nucleotide polymorphism

T2D

type 2 diabetes

VO2max

maximal volume of oxygen consumption

WHI

Waist–Hip Index

WHR

waist–hip ratio

WHRadjBMI

WHR adjusted for BMI

Contributor Information

Alisa D Kjaergaard, Steno Diabetes Center Aarhus, Aarhus University Hospital, 8200 Aarhus, Denmark; Joslin Diabetes Center, Boston, MA 02115, USA.

Christina Ellervik, Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, 2200 Copenhagen, Denmark; Department of Clinical Biochemistry, Zealand University Hospital, 4600 Køge, Denmark; Department of Laboratory Medicine, Boston Children's Hospital, Boston, MA 02115, USA; Department of Pathology, Harvard Medical School, Boston, MA 02115, USA.

Niels Jessen, Steno Diabetes Center Aarhus, Aarhus University Hospital, 8200 Aarhus, Denmark; Department of Biomedicine, Faculty of Health, Aarhus University, 8000 Aarhus, Denmark.

Sarah J Lessard, Joslin Diabetes Center, Boston, MA 02115, USA; Department of Medicine, Harvard Medical School, Boston, MA 02115, USA.

Funding

A.D.K. was funded by the Novo Nordisk Foundation (grant reference number NNF22OC0076023 in the call “Steno North American Fellowships 2022”). C.e. is partly funded by the Laboratory Medicine Endowment Fund of Boston Children’s Hospital. S.J.L. is supported by National Institute of Diabetes and Digestive and Kidney Diseases grants R01 DK124258 and R01 DK129850.

Author Contributions

Conceptualization: A.D.K., C.e., S.J.L. Formal analysis: ADK. Investigation: A.D.K. Data curation: A.D.K. Resources: A.D.K., N.J., S.J.L. Writing—original draft preparation: A.D.K. Writing—review and editing: A.D.K., C.e., N.J., S.J.L. Visualization: A.D.K. Supervision: S.J.L. A.D.K. is the guarantor of this work.

Disclosures

The authors have nothing to disclose. A.D.K. is the chair of the Statistical Review Board Member for The Journal of Clinical Endocrinology & Metabolism and played no role in the Journal's evaluation of the manuscript.

Data Availability

The datasets analyzed for this study are publicly available and accessed as described in the Methods section and Supplementary Table S1 (12).

Prior Presentation

This work was presented as an abstract and a short oral presentation at EASD 2023.

References

  • 1. Ross R, Blair SN, Arena R, et al. Importance of assessing cardiorespiratory fitness in clinical practice: a case for fitness as a clinical vital sign: a scientific statement from the American Heart Association. Circulation. 2016;134(24):e653‐e699. [DOI] [PubMed] [Google Scholar]
  • 2. Ezzatvar Y, Izquierdo M, Núñez J, Calatayud J, Ramírez-Vélez R, García-Hermoso A. Cardiorespiratory fitness measured with cardiopulmonary exercise testing and mortality in patients with cardiovascular disease: a systematic review and meta-analysis. J Sport Health Sci. 2021;10(6):609‐619. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Gonzales TI, Westgate K, Strain T, et al. Cardiorespiratory fitness assessment using risk-stratified exercise testing and dose-response relationships with disease outcomes. Sci Rep. 2021;11(1):15315. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Kodama S, Saito K, Tanaka S, et al. Cardiorespiratory fitness as a quantitative predictor of all-cause mortality and cardiovascular events in healthy men and women: a meta-analysis. JAMA. 2009;301(19):2024‐2035. [DOI] [PubMed] [Google Scholar]
  • 5. Tarp J, Støle AP, Blond K, Grøntved A. Cardiorespiratory fitness, muscular strength and risk of type 2 diabetes: a systematic review and meta-analysis. Diabetologia. 2019;62(7):1129‐1142. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Cai L, Gonzales T, Wheeler E, et al. Causal associations between cardiorespiratory fitness and type 2 diabetes. Nat Commun. 2023;14(1):3904. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Lin X, Zhang X, Guo J, et al. Effects of exercise training on cardiorespiratory fitness and biomarkers of cardiometabolic health: a systematic review and meta-analysis of randomized controlled trials. J Am Heart Assoc. 2015;4(7):e002014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Webb KL, Gorman EK, Morkeberg OH, et al. The relationship between hemoglobin and [formula: see text]: a systematic review and meta-analysis. PLoS One. 2023;18(10):e0292835. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Bhasin S, Storer TW, Berman N, et al. The effects of supraphysiologic doses of testosterone on muscle size and strength in normal men. N Engl J Med. 1996;335(1):1‐7. [DOI] [PubMed] [Google Scholar]
  • 10. Chow LS, Albright RC, Bigelow ML, Toffolo G, Cobelli C, Nair KS. Mechanism of insulin’s anabolic effect on muscle: measurements of muscle protein synthesis and breakdown using aminoacyl-tRNA and other surrogate measures. Am J Physiol Endocrinol Metab. 2006;291(4):E729-736. [DOI] [PubMed] [Google Scholar]
  • 11. Westerterp KR. Exercise, energy balance and body composition. Eur J Clin Nutr. 2018;72(9):1246‐1250. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Kjaergaard AD, Ellervik C, Jessen N, Lessard SJ. Supplementary data to the paper: cardiorespiratory fitness, body composition, diabetes, and longevity: a two-sample Mendelian randomization study. https://github.com/AlisaDK/Cardiorespiratory-fitness. [DOI] [PMC free article] [PubMed]
  • 13. Krakauer NY, Krakauer JC. A new body shape index predicts mortality hazard independently of body mass index. PLoS One. 2012;7(7):e39504. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Krakauer NY, Krakauer JC. An anthropometric risk index based on combining height, weight, waist, and hip measurements. J Obes. 2016;2016:8094275. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Christakoudi S, Evangelou E, Riboli E, Tsilidis KK. GWAS of allometric body-shape indices in UK Biobank identifies loci suggesting associations with morphogenesis, organogenesis, adrenal cell renewal and cancer. Sci Rep. 2021;11(1):10688. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Wang Z, Emmerich A, Pillon NJ, et al. Genome-wide association analyses of physical activity and sedentary behavior provide insights into underlying mechanisms and roles in disease prevention. Nat Genet. 2022;54(9):1332‐1344. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Doherty A, Smith-Byrne K, Ferreira T, et al. GWAS identifies 14 loci for device-measured physical activity and sleep duration. Nat Commun. 2018;9(1):5257. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Hanscombe KB, Persyn E, Traylor M, et al. The genetic case for cardiorespiratory fitness as a clinical vital sign and the routine prescription of physical activity in healthcare. Genome Med. 2021;13(1):180. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Doherty A, Jackson D, Hammerla N, et al. Large scale population assessment of physical activity using wrist worn accelerometers: the UK Biobank study. PLoS One. 2017;12(2):e0169649. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Inshaw JRJ, Sidore C, Cucca F, et al. Analysis of overlapping genetic association in type 1 and type 2 diabetes. Diabetologia. 2021;64(6):1342‐1347. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Mahajan A, Taliun D, Thurner M, et al. Fine-mapping type 2 diabetes loci to single-variant resolution using high-density imputation and islet-specific epigenome maps. Nat Genet. 2018;50(11):1505‐1513. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Chen J, Spracklen CN, Marenne G, et al. The trans-ancestral genomic architecture of glycemic traits. Nat Genet. 2021;53(6):840‐860. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Lagou V, Mägi R, Hottenga JJ, et al. Sex-dimorphic genetic effects and novel loci for fasting glucose and insulin variability. Nat Commun. 2021;12(1):24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Ruth KS, Day FR, Tyrrell J, et al. Using human genetics to understand the disease impacts of testosterone in men and women. Nat Med. 2020;26(2):252‐258. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Timmers PR, Mounier N, Lall K, et al. Genomics of 1 million parent lifespans implicates novel pathways and common diseases and distinguishes survival chances. Elife. 2019;8:e39856. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Hemani G, Tilling K, Davey Smith G. Orienting the causal relationship between imprecisely measured traits using GWAS summary data. PLoS Genet. 2017;13(11):e1007081. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Hemani G, Zheng J, Elsworth B, et al. The MR-base platform supports systematic causal inference across the human phenome. Elife. 2018;7:e34408. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Broadbent JR, Foley CN, Grant AJ, Mason AM, Staley JR, Burgess S. MendelianRandomization v0.5.0: updates to an R package for performing Mendelian randomization analyses using summarized data. Wellcome Open Res. 2020;5:252. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Hartwig FP, Davies NM, Hemani G, Davey Smith G. Two-sample Mendelian randomization: avoiding the downsides of a powerful, widely applicable but potentially fallible technique. Int J Epidemiol. 2016;45(6):1717‐1726. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Greco MF, Minelli C, Sheehan NA, Thompson JR. Detecting pleiotropy in Mendelian randomisation studies with summary data and a continuous outcome. Stat Med. 2015;34(21):2926‐2940. [DOI] [PubMed] [Google Scholar]
  • 31. Rees JMB, Wood AM, Dudbridge F, Burgess S. Robust methods in Mendelian randomization via penalization of heterogeneous causal estimates. PLoS One. 2019;14(9):e0222362. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol. 2016;40(4):304‐314. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Bowden J, Del Greco MF, Minelli C, Davey Smith G, Sheehan NA, Thompson JR. Assessing the suitability of summary data for two-sample Mendelian randomization analyses using MR-egger regression: the role of the I2 statistic. Int J Epidemiol. 2016;45(6):1961‐1974. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Minelli C, Del Greco MF, van der Plaat DA, Bowden J, Sheehan NA, Thompson J. The use of two-sample methods for Mendelian randomization analyses on single large datasets. Int J Epidemiol. 2021;50(5):1651‐1659. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Perry JR, Weedon MN, Langenberg C, et al. Genetic evidence that raised sex hormone binding globulin (SHBG) levels reduce the risk of type 2 diabetes. Hum Mol Genet. 2010;19(3):535‐544. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Mahajan A, Spracklen CN, Zhang W, et al. Multi-ancestry genetic study of type 2 diabetes highlights the power of diverse populations for discovery and translation. Nat Genet. 2022;54(5):560‐572. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Ghanbari F, Yazdanpanah N, Yazdanpanah M, Richards JB, Manousaki D. Connecting genomics and proteomics to identify protein biomarkers for adult and youth-onset type 2 diabetes: a two-sample Mendelian randomization study. Diabetes. 2022;71(6):1324‐1337. [DOI] [PubMed] [Google Scholar]
  • 38. Rooney MR, Chen J, Echouffo-Tcheugui JB, et al. Proteomic predictors of incident diabetes: results from the Atherosclerosis Risk in Communities (ARIC) study. Diabetes Care. 2023;46(4):733‐741. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Gudmundsdottir V, Zaghlool SB, Emilsson V, et al. Circulating protein signatures and causal candidates for type 2 diabetes. Diabetes. 2020;69(8):1843‐1853. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Wang Q, Kangas AJ, Soininen P, et al. Sex hormone-binding globulin associations with circulating lipids and metabolites and the risk for type 2 diabetes: observational and causal effect estimates. Int J Epidemiol. 2015;44(2):623‐637. [DOI] [PubMed] [Google Scholar]
  • 41. Yuan S, Wang L, Sun J, et al. Genetically predicted sex hormone levels and health outcomes: phenome-wide Mendelian randomization investigation. Int J Epidemiol. 2022;51(6):1931‐1942. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Aune D BÓH, Vatten LJ. Resting heart rate and the risk of type 2 diabetes: a systematic review and dose–response meta-analysis of cohort studies. Nutr Metab Cardiovasc Dis. 2015;25(6):526‐534. [DOI] [PubMed] [Google Scholar]
  • 43. Guo Y, Chung W, Zhu Z, et al. Genome-wide assessment for resting heart rate and shared genetics with cardiometabolic traits and type 2 diabetes. J Am Coll Cardiol. 2019;74(17):2162‐2174. [DOI] [PubMed] [Google Scholar]
  • 44. van de Vegte YJ, Eppinga RN, van der Ende MY, et al. Genetic insights into resting heart rate and its role in cardiovascular disease. Nat Commun. 2023;14(1):4646. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The datasets analyzed for this study are publicly available and accessed as described in the Methods section and Supplementary Table S1 (12).

Articles from The Journal of Clinical Endocrinology and Metabolism are provided here courtesy of The Endocrine Society

RESOURCES