This was a prospective study that included patients with a diagnosis of HFpEF according to the criteria of the European Society of Cardiology12 and New York Heart Association functional class II-III/IV between 2 periods: June 2012 to May 2013 and June 2015 to May 2016. The study was conducted in a single third-level center in Spain. All patients provided signed informed consent before participation. The protocol was approved by the research ethics committee of our center in accordance with the principles of the Declaration of Helsinki and national regulations.

Candidate patients were selected from the outpatient HF unit.13 All patients met the following inclusion criteria: a) previous history of symptomatic HF (New York Heart Association functional class ≥ II); b) normal left ventricular ejection fraction (ejection fraction > 0.50 by the Simpson method and end-diastolic diameter < 60mm); c) structural heart disease (left ventricle hypertrophy/left atrial enlargement) and/or diastolic dysfunction estimated by 2-dimensional echocardiography; d) previous admission for acute HF; and e) clinical stability, without hospital admissions in the past 3 months. Patients were excluded if they could not perform a valid baseline exercise test or showed any previous medical condition such as: unstable angina, myocardial infarction or cardiac surgery within the previous 3 months; chronic metabolic, orthopedic, infectious disease or pulmonary disease (including pulmonary arterial hypertension, chronic thromboembolic pulmonary disease or chronic obstructive pulmonary disease); steroid, hormone, or cancer therapy; acute HF decompensation; any other comorbidity with a life expectancy of less than 1 year. The patient flowchart is depicted in Figure 1.

Figure 1.

Flowchart for patient inclusion and follow-up. HF, heart failure; LVEF, left ventricular ejection fraction.

(0.29MB).

After signing the informed consent form, a comprehensive medical history, physical examination, anthropometry and examination tests were performed by 2 trained cardiologists.

Study procedures included electrocardiogm, echocardiography, cardiopulmonary exercise testing, and blood samples for a panel of baseline biomarkers. All of them were performed on the same day.

The total number of all-cause unplanned hospitalizations was selected as the primary endpoint. Cardiovascular admissions were selected as the secondary endpoint. Cardiovascular admissions were considered as those occurring due to acute HF, acute coronary syndrome, arrhythmias, stroke, or other cardiovascular causes, such as rupture of an aneurysm, peripheral ischemia, or aortic dissection. Hospitalizations were identified from the clinical records of patients in the HF unit and hospital wards and from electronic medical records. Fatal events were identified from the clinical records of the HF unit, hospital wards, emergency room, and general practitioners and by contacting the patient's relatives. All patients included were foowed up until September 2016. The minimum duration of patient follow-up was 3 months.

Continuous and categorical variables are presented as the mean ± standard deviation, median [interquartile range] or percentages, as appropriate. Crude rates (number of events per 10 person-year) are presented for pp-peak VO2 quartiles (pp-peak VO2Q). Bivariate negative binomial regression was used to assess the independent association between pp-peak VO2 with the endpoint.17 This method simultaneously models the number of admissions with the mortality event and uses a shared frailty to account for the correlation between these 2 outcomes. Thus, the estimates for recurrent admissions are internally adjusted by mortality as a terminal event (informative dropout). To account for differences in the observation time per patient, we used the natural logarithm of the observation time to offset the model for admission endpoints as well as for mortality. Estimates are reported as incidence rate ratio (IRR). All variables listed in Table 1 were evaluated for prognostic purposes. A backward stepwise selection, with the Akaike information criterion as the stopping criterion, was used to achieve a parsimonious model. The linearity assumption for continuous variables was simultaneously tested and transformed, if appropriate, with fractional polynomials. The pp-peak VO2 was modeled using a fractional polynomial transformation of 0 (natural logarithm), per decrease in 10%, and as pp-peak VO2Q. The covariates included in the final models for all-cause and cardiovascular rehospitalization are included in the legend of Figure 2. A 2-sided P-value of < .05 was considered to be statistically significant. All analyses were performed using Stata 14.0.

Figure 2.

Rates of all-cause, CV and AHF hospitalizations according to quartiles of percentage of predicted peak exercise oxygen uptake. AHF, acute heart failure; CV, cardiovascular.

Q1 = 26.1% to 46.4% of predicted peak exercise oxygen uptake.

Q2 = 47% to 60% of predicted peak exercise oxygen uptake.

Q3 = 60% to 67.5% of predicted peak exercise oxygen uptake.

Q4 = 67.7% to 102.8% of predicted peak exercise oxygen uptake.

(0.13MB).

In a univariate setting, pp-peak VO2 was significantly and inversely associated with the risk of recurrent all-cause admissions. Thus, risk estimates attributable to pp-peak VO2 evaluated as continuous (per decrease in 10%) and as a natural logarithm (per decrease in 1 log) were (IRR, 1.30; 95% confidence interval [95%CI], 1.06-1.61; P = .014 and IRR, 3.94; 95%CI, 1.40-11.05; P = .009), respectively.

In a multivariate scenario, including well-established prognosticators and potential confounders (age, atrial fibrillation, systolic blood pressure, heart rate, E/e’ ratio, hemoglobin and N-terminal pro-B-type natriuretic peptide) in the risk model, the inverse association between pp-peak VO2 and the risk of repeat hospitalizations remained significant (Table 2). The risk gradient showed an inverse and almost linear relationship between pp-peak VO2 and the risk of all-cause readmissions (Figure 3). When pp-peak VO2 was modeled as continuous, a 10% decrease of pp-peak VO2 increased the risk of recurrent hospitalizations by 32% (IRR, 1.32; 95%CI, 1.03-1.68; P = .028) (Table 2). In a sensitivity analysis, forcing in the multivariate analysis other important cardiopulmonary exercise prognosticators such as the VE/VCO2 slope, ppVO2 (per decrease of 10%) remained significantly associated with this endpoint (IRR, 1.31; 95%CI, 1.05-1.65; P = .020) and the VE/VCO2 slope was borderline associated with the risk of recurrent hospitalizations (IRR, 1.03; 95%CI, 0.99-1.06; P = .088).

Figure 3.

Relationship between pp-peak VO2 and the risk of all-cause readmission after multivariate adjustment. pp-peak VO2, percentage of predicted peak exercise oxygen uptake. Adjusted for age, atrial fibrillation, systolic blood pressure, heart rate, E/e’ ratio, hemoglobin and N-terminal pro-B-type natriuretic peptide.

(0.07MB).

Univariate estimates showed an inverse relationship between pp-peak VO2 and cardiovascular admissions. The pp-peak VO2 (per 10% decrease) and logarithm of pp-peak VO2 (per decrease by 1 log) were inversely associated with a higher risk of cardiovascular hospitalizations (IRR, 1.28; 95%CI, 0.99-1.66; P = .063 and IRR, 3.73; 95%CI, 1.05-13.24; P = .042, respectively). After a multivariate adjustment, this inverse association was strengthened (Table 2). Regarding acute HF hospitalizations, multivariate analysis showed that pp-peak VO2 (per 10% decrease) and the logarithm of pp-peak VO2 (per decrease by 1 log) were not significantly related to the risk of repeat acute HF admissions (IRR, 1.10; 95%CI, 0.25-4.73; P = .583 and IRR, 1.55; 95%CI, 0.36-6.69; P = .556, respectively). Similarly, quartiles of pp-peak VO2 were not related to the risk of repeat acute HF hospitalizations (Table 2).

This study has some limitations. First, this is a single center observational study in which there may have been many potential confounders; second, the small number of adverse events registered, such as mortality and acute HF hospitalizations, precluded evaluation of the independent association between pp-peak VO2 and those events; third, these findings cannot be directly extrapolated to patients with milder forms of the disease; and fourth, the use of a bicycle exercise protocol rather than treadmill exercise testing could have led to underestimation of functional capacity in some patients.14