Neither LVEF nor LV-GLS was independent associated with impaired exercise capacity. In contrast, while all the phenotypes of HF have similar pattern of DD and elevated LV filling pressure, DD majorly determined the peak VO2. The increase in peak VO2 also depended on the blunt increase in medial E/e’ along with the increased cardiac output during exercise. The study results may suggest that the medial E/e’ normalized by cardiac output as a surrogate of dynamic left-sided filling pressures during incremental exercise was closely related to impaired exercise capacity. (Graphical abstract) In addition, LV systolic function, indexed by GLS but not LVEF was associated with VE/VCO2 slope. And the increase in peak LV-GLS during exercise may also contribute to the ventilatory efficiency. LV-GLS could reflect the degree of the extent of myocardial fibrosis and attenuate the LV relaxation and contractility[23]. However, the definite mechanism of how the LV-GLS affect ventilatory efficiency was worth more exploration.
The exercise capacity and ventilatory efficiency in heart failure with different phenotypes
Although HF could be categorized by LVEF into different phenotypes, patients with HF may all experience equivalent risks of hospitalization for HF and the same loss of lifespan[24]. Even though they have been characterized with distinct LV systolic function, their exercise capacity might not be different. Pugliese et al. have reported 169 subjects with either HFrEF, HFmrEF or HFpEF had similar peak VO2[9]. In this study, we also demonstrated ambulatory patients with HF may have similar exercise performance, in terms of peak VO2, OUES, and VE/VCO2, regardless of the phenotypes. Given the study only recruited subjects with limited symptoms and eligible for exercise tests, the results may not be generalized to patients with poor daily activity. In addition, the study further extended our understandings that patients with HFrecEF also have comparable exercise capacity and ventilatory efficacy with the other phenotypes of HF.
Associations of Left ventricular systolic and diastolic functions with exercise capacity
Although the exercise capacity was similar between HF phenotypes[8] [9], neither LV-GLS nor LVEF was independently predictive of impaired exercise capacity. While the common feature across all the phenotypes of HF is elevated LV end-diastolic pressure, which could be indirectly assessed by echocardiographic parameters, including LV volume, TRV and E/e’ ratio, the indices of DD. In addition, the indices of DD independently correlated with peak VO2. The results may support the major determinant of exercise capacity was LV diastolic but not systolic function. In fact, none of the patients with DD could achieve a peak VO2 of > 14 ml/kg/min in this study. In contrast, only 17.2% of patients with normal diastolic function had a peak VO2 < 14 ml/kg/min. (Table S1)
The performance of exercise medial E/e’ to CO ratio and LV-GLS with adjustment of cardiac output
During exercise, a normal LV would increase the trans-mitral flow by facilitating LV relaxation to augment cardiac output, which may limit the elevation in LA pressure[25]. In contrast, a failing heart is characterized by elevated LV filling pressure, which would be further enhanced during exercise[26]. The elevated of LV end-diastolic pressure would be transmitted to pulmonary system and activate the receptor of dyspnea[27], resulting in cessation of exercise. Exercise-related PCWP change in relation to cardiac output has been an indicator to evaluate hemodynamic changes in the face of the stress load[28]. In this study, we used the medial E/e' to CO ratio to correct the augmentation in LV filling pressures in response to the increase in cardiac output. Therefore, the peak medial E/e' to CO ratio would reflect the speed at which medial E/e' raised during the exercise. In contrast, the peak LV-GLS but not Peak LV-GLS to CO ratio better correlated with VE/VCO2 slope. This result may support diastolic filling pressure rather than systolic function was better associated with LV volume changes, while LV-GLS but not medial E/e’ a volume-independent indicator.
The AVO2diff was a measure of oxygen was extracted from blood circulates into the body, which was a derivative of cardiac output and VO2. AVO2diff was considered as the major determinant of exercise capacity, especially for HFmrEF and HFpEF subtypes but not in HFrEF[9, 29]. In this study, though peak AVO2diff was crudely associated with reduced exercise capacity. (OR [95%CI] : 0.741[0.550–0.997]) and lower AUCs (0.711)(not shown in Fig. 3), the discriminatory ability of AVO2diff was lower than peak medial E/e’ ratio or its adjusted value with CO. This difference from previous studies may be due to the different heart failure phenotypes, and to investigate peripheral oxygen utilization one should consider methods other than the indirect calculation of peak AV O2 difference, such as near-infrared spectroscopy (NIRS). LVEF, as an important parameter for heart failure phenotyping, did not show an independent predictive ability for Peak VO2 or VE/VCO2 slope in our study. This is similar to observations made in the past HF-ACTION trial.[8]
The different impacts of medial E/e’ to CO ratio and GLS on peak VO 2 and VE/VCO2 slope
As mentioned above, the change of LV filling pressure would transmit to LA pressure and then to pulmonary system, which cause dyspnea and affect exercise performances[27]. This mechanism was compatible with our finding and diastolic dysfunction, at rest or during exercise were the major determinants of reduced exercise capacity. LV-GLS was a surrogate of LV contractility, less load-dependent and associated with clinical adverse events, such hospitalized HF or death in all HF phenotypes[30, 31]. VE/VCO2 slope as an indicator of ventilatory inefficiency had comparable and even superior prognostic value in comparison with peak VO2[2]. Traditionally, the elevation of VE/VCO2 slope was resulted from decreased pulmonary perfusion caused by high LV filling pressure[4]. However, the association between LV-GLS, either resting or peak, and VE/VCO2 slope was more predominant than diastolic dysfunction or medial E/e’ ratio. The mechanism of the association between GLS and VE/VCO2 slope was worth more exploration, and whether peak LV-GLS has a better prognostic value than resting LV-GLS needed to be further investigated.
Study limitations
Given the sample of the study was comparatively small, there were some limitations. First, the study participants were relatively young and stable. Therefore, the results may not be generated to the sicker and older population, when the muscle strength could be a significant confounder. Second, even exercise intolerance was an important surrogate of the clinical outcomes, a longitudinal cohort study was needed to support the superior prognostic role of CPET in patients with HF. Third, the study was limited by the relatively small sample and single-center, retrospective observational nature of the study. Therefore, there was inevitable selection bias and type II error in this observation study and may impact the generalizability of our results. For example, the difference in peak VO2 between HFrEF and HFrecEF may be statistically significant if the sample size was larger. A case-matched study could provide substantial improvement in the generalizability of the study results.