Article Text

Download PDFPDF

Original research article
Impaired oxygen uptake kinetics in heart failure with preserved ejection fraction
  1. Christopher M Hearon Jr1,2,
  2. Satyam Sarma1,2,
  3. Katrin A Dias1,2,
  4. Michinari Hieda1,2,
  5. Benjamin D Levine1,2
  1. 1 Department of Internal Medicine, University of Texas Southwestern Medical Center, Dallas, Texas, USA
  2. 2 Texas Health Presbyterian Hospital Dallas, Institute for Exercise and Environmental Medicine, Dallas, Texas, USA
  1. Correspondence to Dr. Benjamin D Levine; BenjaminLevine{at}


Objective The time needed to increase oxygen utilisation to meet metabolic demand (V̇O2 kinetics) is impaired in heart failure (HF) with reduced ejection fraction and is an independent risk factor for HF mortality. It is not known if V̇O2 kinetics are slowed in HF with preserved ejection fraction (HFpEF). We tested the hypothesis that V̇O2 kinetics are slowed during submaximal exercise in HFpEF and that slower V̇O2 kinetics are related to impaired peripheral oxygen extraction.

Methods Eighteen HFpEF patients (68±7 years, 10 women) and 18 healthy controls (69±6 years, 10 women) completed submaximal and peak exercise testing. Cardiac output (acetylene rebreathing, Q̇c), ventilatory oxygen uptake (V̇O2, Douglas bags) and arterial-venous O2 difference (a-vO2 difference) derived from Q̇c and V̇O2 were assessed during exercise. Breath-by-breath O2 uptake was measured continuously throughout submaximal exercise, and V̇O2 kinetics was quantified as mean response time (MRT).

Results HFpEF patients had markedly slowed V̇O2 kinetics during submaximal exercise (MRT: control: 40.1±14.2, HFpEF: 65.4±27.7 s; p<0.002), despite no relative impairment in submaximal cardiac output (Q̇c: control: 8.6±1.7, HFpEF: 9.7±2.2 L/min; p=0.79). When stratified by MRT, HFpEF with an MRT ≥60 s demonstrated elevated Q̇c, and impaired peripheral oxygen extraction that was apparent during submaximal exercise compared with HFpEF with a MRT <60 s (submaximal a-vO2 difference: MRT <60 s: 9.7±2.1, MRT ≥60 s: 7.9±1.1 mL/100 mL; p=0.03).

Conclusion HFpEF patients have slowed V̇O2 kinetics that are related to impaired peripheral oxygen utilisation. MRT can identify HFpEF patients with peripheral limitations to submaximal exercise capacity and may be a target for therapeutic intervention.

  • heart failure with preserved ejection fraction
  • exercise capacity
  • oxygen uptake kinetics

Statistics from

Request Permissions

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


V̇O2 kinetics reflect the speed at which oxygen delivery and utilisation rise to meet the metabolic demand of contracting skeletal muscle during the transition from rest to steady state exercise, which may be quantified as the mean response time (MRT).1 Impaired V̇O2 kinetics are a common characteristic of heart failure with reduced ejection fraction (HFrEF) that is closely associated with the severity of cardiac dysfunction.2–5 Impairments along every step of the oxygen transport and utilisation cascade have been identified in HFrEF, each contributing mechanistically to impaired V̇O2 kinetics and exercise intolerance. However, a distinguishing feature of HFrEF is reduced cardiac output (Q̇c) at rest that remains depressed throughout submaximal and maximal exercise, such that oxygen uptake is partially limited by reduced bulk oxygen delivery to working skeletal muscle.6 7 HFrEF and HFpEF share many pathophysiological features that contribute to impaired V̇O2 kinetics including reduced skeletal muscle diffusive and oxidative capacity.8 However, a critical difference between these HF patients is the presence of an adequate Q̇c response to submaximal exercise in heart failure with preserved ejection fraction (HFpEF) compared with HFrEF patients7 9 10 despite significant central limitations during maximal exercise in HFpEF.11

Similar to HFrEF, peak Q̇c and impaired peripheral oxidative capacity limit peak exercise performance in HFpEF patients.7 9–17 While peakc is blunted due to chronotropic incompetence and impaired stroke volume reserve during maximal exercise,12 it is unclear to what extent HFpEF patients experience impaired haemodynamic and metabolic function during submaximal exercise, similar to activities of daily living. In contrast to HFrEF, HFpEF patients consistently demonstrate normal augmentation of Q̇c during low-intensity submaximal exercise.7 9 10 18 Thus, the pathophysiology of exercise intolerance during submaximal exercise may be distinct from HFrEF patients. However, to date, no studies have investigated how the unique pathophysiology observed in HFpEF impacts V̇O2 kinetics. Therefore, we tested the hypothesis that HFpEF patients would demonstrate slow V̇O2 kinetics (longer MRT) compared with age matched controls and that this abnormality may be primarily related to reduced peripheral oxidative capacity.


Subject characteristics

A total of 18 patients with a hospital discharge diagnosis of CHF were enrolled in studies at the Institute for Exercise and Environmental Medicine between 2007 and 2016. Haemodynamic data from 11 HFpEF patients in this investigation have been reported previously18 and are combined with data from seven subsequent HFpEF patients whose data have not been published previously. Detailed inclusion and exclusion criteria are provided in online supplementary materials. All subjects signed an informed consent form approved by the institutional review boards of UTSW Medical Center and Presbyterian Hospital, Dallas, Texas.

Supplemental material

Cardiopulmonary exercise testing (CPET)

Subjects completed two modified Astrand-Saltin incremental treadmill protocols to determine peak oxygen uptake (V̇O2 peak) as reported previously.18 On the first occasion, maximal exercise testing was used for screening and familiarisation (day 1). At least 72 hours later, exercise testing was repeated to assess steady-state haemodynamics, V̇O2 and V̇O2 kinetics during submaximal exercise, followed by haemodynamics and V̇O2 during maximal exercise (day 2). Resting haemodynamics and ventilatory V̇O2 were measured during 3 min of quiet standing on a treadmill. Following resting measurements, a low-intensity steady-state protocol was initiated at the desired treadmill velocity, enabling a rapid transition into the steady state workload. The workload was chosen for each individual to be below ventilatory threshold (~40%–50% predicted V̇O2 peak, calculated from day 1 CPET) for 6 min. After a minimum of 10 min of rest, peak exercise was assessed a second time using the same modified Astrand-Saltin incremental treadmill protocol as performed on day 1. Atrioventricular nodal blockers were withheld for 48 hours prior to avoid confounding effects of β-blockade on oxygen uptake kinetics (see online supplementary material S1). Other antihypertensive medications were continued.

Heart rate (HR) was monitored continuously via electrocardiogram (Schiller AT-10; Welch Allyn Inc, Skaneateles Falls, New York, USA), and blood pressure was measured using electrosphygmomanometry (Suntech Tango+; Morrisville, North Carolina, USA). Measurements of ventilatory gas exchange were made on a breath-by-breath basis using an infrared turbine flow metre (VMM, Interface Associates; Laguna Niguel, California, USA) and mass spectrometry for rapid gas analysis via proprietary software. In addition, breath-by-breath ventilatory gas exchange was also verified at steady state by collection of expired gas in a Douglas bag for 1 min and analysed via mass spectrometry. Steady-state ventilatory volume was measured by use of a 120 L Tissot spirometer (W.E. Collins P-1700; Braintree, Massachusetts, USA). During maximal exercise, peak V̇O2 was defined as the highest oxygen uptake measured from at least a 30 s Douglas bag collection. Q̇c was measured using a modified acetylene rebreathing technique that is well validated, reproducible and equivalent to invasive measures of Q̇c during maximal exercise.18 Arterial-venous oxygen difference (a-vO2 difference) was calculated from the ratio between Q̇c and oxygen uptake using the Fick equation.


V̇O2 kinetics were measured during the transition from rest to steady-state exercise (~40%–50% V̇O2 peak) as described above. Ventilatory V̇O2 data were linearly interpolated to 1 s intervals, time averaged into 5 s bins and fit using first-order kinetics (see online supplementary materials S2 and S3), with the time delay set to the start of exercise and therefore inclusive of both the cardiodynamic phase I and phase II rise in V̇O2. MRT was defined as the exponential time constant τ, which represents the time needed for oxygen uptake to rise to ~63% of steady state V̇O2. Oxygen deficit was calculated using Simpson’s method as the difference between cumulative oxygen demand and cumulative oxygen utilisation.

Statistical analysis

The sample was selected from all of the CPET tests collected as part of research studies at the Institute for Exercise and Environmental medicine between 2007 and 2016 (18 and online supplementary materials S5 and S6) to maximise sample size. Inclusion/exclusion of all tests was determined prior to statistical analysis, and the sample represents all tests with adequate breath by breath and haemodynamic data available at IEEM for this particular analysis. The typical error of VO2 uptake kinetics (specifically MRT) in our laboratory is 8%–13% (or 5–8 s; see online supplementary material S7). Various physiological variables are reported as descriptive variables of the experimental intervention in both groups. Primary comparisons of MRT and a-vO2 difference between control and HFpEF patients was made using unpaired Student’s t-tests. As an exploratory post hoc analysis, HFpEF patients were stratified into fast and slow kinetics based on an MRT of 60 s (MRT <60 s and MRT ≥60 s, respectively). Comparison of haemodynamics were made separately during rest, the V̇O2 kinetics protocol and the graded exercise testing protocol between control, MRT <60 s and MRT ≥60 s using analysis of variance and Newman-Keuls post hoc tests, which is robust up to three comparison groups but does not adjust for multiple comparisons. Fishers’ exact test was used for categorical data. Significance was set a priori at p<0.05. Data for all primary comparisons are expressed in text as mean difference with 95% CIs, and all data were expressed in tabular form as mean±SD.


Subject characteristics

Clinical characteristics of control subjects (n=18), HFpEF patients (n=18) and HFpEF patients stratified by MRT (MRT <60 s: n=9; MRT ≥60 s: n=9) are presented in table 1.

Table 1

Clinical characteristics of healthy controls and patients with HFpEF

V̇O2 kinetics and haemodynamics during submaximal exercise: HFpEF versus control

During submaximal exercise, there was no evidence of chronotropic incompetence, and the rise in Q̇c was not different between HFpEF and controls (∆Qc: control: 4.61±1.02 vs HFpEF: 4.82±1.62 L/min; p=0.65). Submaximal a-vO2 difference (mL/100 mL) was not different between HFpEF and controls (mean difference: 0.96 mL/100 mL, 95% CI −0.286 to 2.204; p=0.18). There were no significant differences in oxygen demand (V̇O2) or the amplitude of V̇O2 kinetics (ΔV̇O2), defined as the rise in V̇O2 from baseline to steady state (control: 0.61±0.15 vs HFpEF: 0.63±0.15 L/min; p=0.70). However, MRT was much longer in HFpEF patients (mean difference: −25.3 s, 95% CI 10.4 to 40.2; p=0.002) nearly doubling the oxygen deficit (control: 0.38±0.23 vs HFpEF: 0.63±0.28 L; p=0.007) (figures 1 and 2).

Figure 1

(A) Representative breath-by-breath ventilatory oxygen uptake kinetics in a HFpEF patient (blue) and a healthy control subject (black). Breath-by-breath oxygen utilisation (V̇O2) was measured during submaximal treadmill exercise, interpolated linearly at 1 s intervals and fitted using on-transient monoexponential curve fitting. (B) Mean response curves for controls HFpEF patients and HFpEF patients stratified by mean response time (MRT) modelled using group average amplitude (Δ V̇O2=steady state V̇O2 – baseline V̇O2) and MRT, defined as the exponential time constant of V̇O2 onset kinetics, which reflects the time needed for oxygen uptake to rise to~63% of steady state V̇O2. HFpEF, heart failure with preserved ejection fraction.

Figure 2

Mean response time (MRT) is slower and oxygen deficit is greater in HFpEF patients. The MRT in seconds (s) of oxygen uptake kinetics (A) and accumulated O2 deficit (B) during submaximal treadmill exercise in HFpEF patients and age-matched healthy controls. The MRT of oxygen uptake kinetics were slower, and O2 deficit was exaggerated in HFpEF patients compared with controls. *P<0.01 versus controls; n=18, 8M:10F for each group, respectively. HFpEF, heart failure with preserved ejection fraction.

Haemodynamics at peak exercise: HFpEF versus control

At peak exercise, absolute V̇O2 was slightly but variably lower in HFpEF compared with controls (control: 1.65±0.44 vs HFpEF: 1.44±0.48 L/min; p=0.19). Peak V̇O2 was consistently and significantly lower in HFpEF patients when scaled to body weight (control: 22.2±4.0 vs HFpEF: 14.6±3.1 mL/min/kg; p<0.001). There was no difference in peak Q̇c between groups (control: 13.4±4.0 vs HFpEF: 13.9±3.5 L/min; p=0.69) (table 4). In contrast to submaximal exercise, significant impairments in peripheral a-vO2 difference were apparent at peak exercise in HFpEF patients (mean difference: −2.41 mL/100 mL, 95% CI −0.59 to −4.231; p=0.01) consistent with previous studies performed in the upright posture.7 10 12 19 Peak heart rate was lower in HFpEF patients compared with controls (control: 158±13 vs HFpEF: 132±23 bpm; p=0.01), and there were no significant differences between groups in blood pressure (systolic, diastolic and mean arterial pressure), peak cardiac index or peak stroke volume index.

Table 4

Haemodynamics at peak exercise in controls and HFpEF patients

Haemodynamics in HFpEF patients with MRT <60 s and MRT ≥60 s

To better understand the impact of slowed kinetics on exercise performance, HFpEF patients were stratified using an MRT of 60 s,2 4 creating two subgroups: HFpEF with normal MRT (MRT <60 s) and HFpEF with slowed MRT (MRT ≥60 s). Previous investigations have demonstrated that MRTs ≥60 s in HFrEF patients are predictive of reduced survival, transplant free survival and survival free of hospitalisation.4–6 Subject characteristics, haemodynamics at rest, submaximal and peak exercise are presented in tables 1–4. During submaximal exercise, there was no difference in baseline V̇O2, steady state V̇O2 and V̇O2 amplitude between the HFpEF cohorts (table 3). In HFpEF patients with MRT ≥60 s, peripheral oxygen extraction was reduced during submaximal exercise compared with control and MRT <60 s (a-vO2 difference: control: 9.8±1.8; MRT <60 s: 9.7±2.1; MRT ≥60 s: 7.9±1.1 mL/100 mL; p<0.05). These patients also required a greater Q̇c for the same absolute V̇O2 (Control: 8.6±1.7; MRT <60 s: 8.7±2.0; MRT ≥60 s: 10.7±2.1 L/min; p<0.05). With the exception of blood pressure (MAP: MRT <60 s: 105±14 vs MRT ≥60 s: 122±13 mm Hg), there were no other significant differences in haemodynamics at peak exercise between MRT <60 s and MRT ≥60 s.

Table 2

Resting haemodynamics in healthy controls and patients with HFpEF

Table 3

V̇O2 kinetics and submaximal exercise haemodynamics in controls and HFpEF patients


This investigation found that patients with HFpEF have significantly slowed V̇O2 kinetics during submaximal exercise compared with age-matched sedentary controls. The impairment in V̇O2 kinetics is similar in magnitude to that reported in HFrEF despite preserved Q̇c during submaximal exercise. When HFpEF patients were stratified by MRT, those with an MRT ≥60 s showed an impairment in peripheral oxidative capacity earlier during submaximal exercise, similar in intensity to activities of daily living, which may result in rapid accumulation of oxygen deficit. Together, these findings suggest that oxygen transport and utilisation is severely compromised during submaximal exercise in HFpEF patients that may contribute significantly to exercise intolerance and fatigue.

V̇O2 kinetics in heart failure (HF)

In HF and other clinical populations, slower kinetics extend the time required for oxygen delivery to meet metabolic demand resulting in a large ‘oxygen deficit’ and fatigue.20 In this cohort of HFpEF patients, MRT was markedly (>50%) impaired compared with the MRT of age-matched controls resulting in greater than 50% increase in oxygen deficit (figure 2). Interestingly, the magnitude of impairment in MRT in HFpEF is similar to MRT values reported previously for HFrEF patients.2 4 21–23 The presence of impaired MRT despite preserved submaximalc, a common finding in HFpEF patients,9 10 14 18 underscores extensive impairments in vascular and metabolic capacity in HFpEF patients. The physiological consequence of such severely slowed V̇O2 kinetics is fatigue related to the rapid depletion of intramuscular high-energy phosphates. Indeed, assessment of intramuscular energetic parameters using 31P magnetic resonance spectroscopy during plantar flexion exercise in HFpEF and HFrEF patients revealed a greater rate of phosphocreatine decline in HFpEF patients, which correlate with skeletal muscle fatigue (plantar flexion), peak V̇O2 and 6 min walk performance.24

To further explore the relationship between MRT and submaximal exercise haemodynamics, we stratified HFpEF patients into two groups based on an MRT cut-off of 60 s. Previous studies investigating the predictive value of MRT indicate that HFrEF patients with an MRT ≥60 s have lower rates of survival, transplant free survival and survival free of hospitalisation compared with HFrEF patients with an MRT <60 s.2 3 25 Additionally, HFrEF patients with an MRT ≥60 s have significantly worse indices of cardiac function including lower ejection fraction and elevated pulmonary arterial pressure during exercise.4 21 26 In this investigation, HFpEF patients with normal MRT experienced similar reductions in peak exercise capacity as patients with MRT ≥60 s (peak V̇O2: control: 22.2±4, MRT <60 s: 14.6±3.1, MRT ≥60 s: 14.7±2.5 mL/kg/min; both p<0.05 vs control), demonstrating that slowed MRT is not merely a generalised consequence of deconditioning.

Rather, slow MRT identified a unique phenotype within HFpEF patients characterised by disproportionately impaired peripheral oxygen extraction. Clinically, patients from each ‘subtype’ of HFpEF (slower vs faster MRT) would have appeared to be identical based solely on peak V̇O2 measured during standard CPET. However, incorporation of V̇O2 kinetics was able to further distinguish between these HFpEF phenotypes. The ‘peripherally limited’ HFpEF phenotype with impaired MRT may be particularly susceptible to exercise intolerance and fatigue given the presence of impaired oxidative capacity during submaximal exercise intensities compared with HFpEF with normal MRT. Assessment of MRT may identify those patients that are well suited for training interventions targeted at improving skeletal muscle function.

Potential mechanisms for reduced kinetics in HFpEF

In healthy individuals, V̇O2 kinetics are largely independent of oxygen delivery and instead constrained by mitochondrial capacity.1 However, in the case of pathophysiological limitations of Q̇c, like that observed in HFrEF, V̇O2 kinetics may become oxygen delivery dependent. Recently, Chatterjee et al 4 assessed both central and peripheral determinants of oxygen uptake concurrently with MRT to identify the factors related to slow MRT in patients with HFrEF. They report that slowed kinetics (MRT ≥60 s) are associated with reduced Q̇c and exaggerated peripheral oxygen extraction during submaximal and peak exercise. The presence of reduced Q̇c and exaggerated peripheral extraction is consistent with a significant limitation of Q̇c and bulk oxygen delivery to contracting skeletal muscle, which is partially compensated by elevated peripheral oxygen extraction in HFrEF patients.27 In contrast to HFrEF, impaired MRT in HFpEF was present despite a normal or exaggerated Q̇c response to submaximal exercise and was associated with an attenuated rise in a-vO2 difference (Δ a-vO2 diff: control: 4.1±1.1, HFpEF: 2.5±1.7 mL/100 mL; p=0.002) suggestive of a primary contribution of impaired skeletal muscle oxidative capacity.

Importantly, the finding of preserved submaximalc and reduced a-vO2 difference during exercise is not isolated to this cohort of HFpEF patients. A recent meta-analysis indicates that HFpEF patients have normal or slightly smaller changes in a-vO2 difference at peak exercise in the supine position, whereas during exercise in the upright posture (the position of most activities of daily living), a-vO2 difference at peak exercise is consistently lower.12 In regards to Q̇c, all studies to date have found that HFpEF patients display a preserved or exaggerated Q̇c response during submaximal exercise.9 10 14 18 Specifically, the majority of these studies indicate that the rise in Q̇c in response to a change in V̇O2 (Q̇c/V̇O2 relationship) is preserved in HFpEF patients compared with healthy controls.10 15–18 28 29 These findings indicate that at sufficiently mild exercise intensities when Q̇c is not limited, skeletal muscle and vascular maladaptations may be important for understanding the impairment in V̇O2 kinetics and exercise intolerance in HFpEF.

Furthermore, Kitzman and colleagues8 show reduced capillary-to-fibre ratio, and a shift towards glycolytic type II muscle fibres in skeletal muscle of HFpEF patients. Greater percentage of type II fibres has been tied mechanistically to slower V̇O2 kinetics in healthy individuals (see online supplementary material). Several studies also report a smaller change in V̇O2 in response to a given change in external work rate in HFpEF,10 14 28 consistent with impaired oxidative capacity and greater prevalence of type 2 fibres. These skeletal muscle fibre type, and structural abnormalities may be compounded further by dysregulated mitochondrial function related to obesity and excessive skeletal muscle fat deposition24 30 that result in cellular metabolic impairment, rapid depletion of high energy phosphates and muscular fatigue.24 Emerging evidence suggests that these pathophysiological skeletal muscle abnormalities may be primary features of an obese HFpEF phenotype.15 In some HFpEF patients, skeletal muscle abnormalities are present during submaximal exercise intensities and manifest prominently during metabolic transitions.7 24 30 Future studies using more direct measures of skeletal muscle oxidative function will be important to determine the mechanisms and extent of skeletal muscle impairments in HFpEF patients. Significant limitations may also exist upstream of skeletal muscle in the oxygen cascade including ventilatory gas exchange, vascular convective oxygen delivery, the distribution of blood flow to, and within active skeletal muscle, and diffusive capacity of oxygen from blood to skeletal muscle.

Limitations and considerations

HFpEF patients demonstrate reduced chronotropy at peak exercise (table 4),12 and it is possible that slower HR kinetics during the transition from rest to steady state could contribute to lower Q̇c and thus slower V̇O2 kinetics. This study is not able to exclude a possible contribution of impaired Q̇c to slowed uptake kinetics. However, it is unlikely that chronotropic incompetence can explain the observed decrements in V̇O2 kinetics considering there was no evidence of lower HR during submaximal exercise in this study, and even at peak exercise, the lower HR did not reduce peak Q̇c. However, further studies are warranted to fully characterise the relative contribution of on-transient cardiac responses to impairments in MRT in HFpEF patients. Finally, larger prospective studies will be necessary to validate the phenotypic findings of the MRT >60 s subanalysis in this investigation.


The MRT of V̇O2 kinetics, an independent predictor of hospitalisation and mortality in HF, is significantly prolonged in HFpEF patients and may be related to impairments in peripheral oxidative capacity. Furthermore, stratification by MRT identified a subset of HFpEF patients with particularly severe peripheral oxidative limitations to submaximal exercise capacity. Given the complex multifactorial nature of HFpEF, the primary mechanism of exercise intolerance is likely to be heterogeneous among patients. In contrast to a-vO2 difference, MRT can be assessed without measurement of Q̇c and may be useful clinically in conjunction with CPET to determine the ‘physiological phenotype’ of an individual HFpEF patient and identify patients who may benefit from exercise interventions tailored to improve skeletal muscle function. Finally, utilising MRT as a gauge of intervention efficacy may identify strategies that improve the integrative cardiovascular response to submaximal exercise intensities that are in line with activities of daily living, which may be missed when focusing solely on changes in peak V̇O2.

Key messages

What is already known on this subject?

  • O2 kinetics are an integrative marker of the efficiency of the cardiovascular system and skeletal muscle to adapt to changes in metabolic demand. Addition of V̇O2 kinetics analysis to maximal exercise testing results improves prognostic capacity in heart failure with reduced ejection fraction.

What might this study add?

  • In heart failure with preserved ejection fraction (HFpEF), V̇O2 kinetics are considerably slowed despite preserved cardiac output during submaximal exercise and are related to impairments in peripheral oxygen utilisation.

How might this impact on clinical practice?

  • O2 kinetics assessed during submaximal exercise intensities can be incorporated into standard cardiopulmonary exercise testing to provide a more integrative assessment of cardiovascular and skeletal muscle function in addition to peak oxygen uptake. In HFpEF, V̇O2 kinetics may be used to phenotype patients with significant peripheral limitations to exercise capacity and inform decisions regarding treatment and exercise rehabilitation.



  • Contributors All authors provided substantial contributions to the conception or design of the work or the acquisition, analysis or interpretation of data for the work; drafting the work or revising it critically for important intellectual content; final approval of the version to be published; and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

  • Funding This work was supported by the National Institutes of Health (NIH R01 AG17479, NIH 1F32HL137285-0) and the American Heart Association (AHA-14SFRN20600009-02).

  • Competing interests None declared.

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

  • Patient consent for publication Not required.

Linked Articles