The natural history of heart failure (HF) is different in women and men. In women, HF which develops later in life, is more often attributable to hypertension and less frequent to coronary disease and is more often associated with preserved ventricular function than that in men [1] . Although women with HF have been reported to survive longer [2] -[4] , they are admitted to hospital more frequently [5] [6] and present with more severe symptoms than men [3] .

The cardiopulmonary exercise test (CPET) provides simultaneous assessment of the performance of the heart, lungs, muscles and cellular respiration during physical exercise and is one of the most reliable tools to select candidates for heart transplantation. Beyond this established role for risk stratification in HF patients, the long- term prognostic value of CPET in women suffering from HF has not yet been deeply explored.

Our study aimed to assess the prognostic value of ventilatory variables for a composite of cardiovascular mortality and HF admissions in a cohort of elderly women with HF and reduced or preserved systolic function.

We studied female patients with a clinical diagnosis of HF based on European Society of Cardiology [7] criteria who were consecutively evaluated at the CPET laboratory of our institution between 2006 and 2011. Patients with both reduced (HFrEF) and preserved (HFpEF) left ventricular ejection fraction (LVEF) were included; patients with preserved (>50%) LVEF had to show signs of congestion on chest X-ray or a BNP value > 100 pg/ml and echocardiographic evidence of diastolic dysfunction [7] .

Patients were in NYHA class II-III, on stable pharmacological treatment since at least three months and none of them was a heart transplant candidate. The indication to CPET was risk stratification and clinical decision- making for follow-up planning. Patients were included in the study if they were able to perform a CPET test with a respiratory exchange ratio > 1.00, indicating the achievement of anaerobic exercise conditions. Among 293 considered patients, 82 were ineligible for clinical contraindications to exercise (severe aortic stenosis (n = 10); severe pulmonary disease (n = 6); orthopedic (n = 65) problems; recent onset atrial fibrillation (n = 1)); 80 were excluded for a respiratory exchange ratio < 1.00.

All patients received routine follow-up care. Outcome data of patients who did not attend their scheduled appointments were obtained by telephone interview of the patient or her family: during the follow-up time, none of these patients had been admitted to the hospital. The study was approved by the Institutional Review Board. Subjects signed a written informed consent to the use of their data for scientific purposes.

Standard symptom-limited CPET was performed by cycle ergometry ramping [8] . Analysis was performed with a metabolic cart (Sensor Medics Vmax 29, Yorba Linda, Calif). Before each test, the equipment was calibrated in standard fashion with reference gases. Standard 12-lead ECGs were obtained at rest, each minute during exercise, and for at least 5 minutes during the recovery phase. Blood pressure was measured with a standard cuff sphygmomanometer. Minute ventilation (VE), oxygen uptake (VO₂), carbon dioxide output (VCO₂), and other cardiopulmonary variables were acquired on a breath-by-breath basis and averaged over 10-second intervals. Peak VO₂ (PVO₂) and peak RER were expressed as the highest averaged samples obtained during the exercise test. VE and VCO₂ values were acquired from the initiation of exercise to peak exercise. For the calculation of the slope of minute ventilation/carbon dioxide production linear relationship (VE/VCO₂ slope), we measured all data points from the beginning to the end of exercise [9] . Exercise oscillatory breathing (EOB) was assessed using criteria previously reported by Leite et al. [10] .

Standard M-mode and two-dimensional echocardiography and Doppler blood flow measurements were performed in agreement with the American Society of Echocardiography guidelines [11] [12] . Early and late diastolic mitral peak flow velocity (E, A) and E-wave deceleration time were assessed by pulsed-wave Doppler echocardiography. Mitral regurgitation was quantitated on a 4-grade (absent, mild, moderate, severe) scale by color-Doppler assessment. In case of at least moderate regurgitation, the PISA method quantification was attempted and insufficiency was classified as severe in the presence of regurgitant surface area ≥ 0.25 cm² and/or ≥ 40 mL regurgitant blood volume [11] [12] .

We studied 131 stable female HF outpatients after implementation of recommended drug therapy including renin-angiotensin system inhibitors in 131 (100%) and beta-blockers in 64 (50%). Cardiac resynchronization therapy and a cardioverter defibrillator had been previously implanted in 5 patients (4%) who met guideline indications for device therapy.

Clinical characteristics and CPET parameters by LV function are summarized in Table 1. Patients with HFpEF and HFrEF had similar age, symptom severity, body mass index, prevalence of atrial fibrillation, achieved workload and VE/VCO₂ slope. Patients with HFrEF showed a higher prevalence of ischemic etiology, left bundle branch block, beta-blocker treatment, moderate-to-severe mitral regurgitation, EOB than subjects with HFpEF.

During a median follow-up of 18 [interquartile range 8 - 54] months, 20 patients died, 18 (14%) of cardiovascular causes (pump failure 8, sudden death 6, acute myocardial infarction 6) and 2 of cancer, whereas 41 subjects were admitted for worsening HF. Using a time-to-first event approach, overall 39 patients (29.7%) met the combined end-point of cardiovascular mortality or HF admission. Variables significantly associated to the combined endpoint by univariable analysis (Table 2) were dichotomized and sequentially entered into multivariable Cox models first combining clinical and echocardiographic variables (model 1) and then adding CPET variables (model 2) (Table 3). A Cox regression of the log hazard ratio on a covariate with a standard deviation of 0.444 based on a sample of 130 patients achieves, at a 0.05 significance level, 86% power to detect a regression coefficient equal to 1.22 and achieves 89% power to detect a regression coefficient equal to 1.283. The sample size was adjusted since a multiple regression of the main variables of interest on the other covariates (VE/VO₂ slope, atrial fibrillation, LVEF, NYHA class, watts achieved) is expected to have an R-Squared of 0.1130.

In adjusted model 2, the hazard ratio for coexistent EOB and VE/VCO₂ slope > 36 was 8.08 (95% CI 3.01 to 21.7). Adjusted survival rates free from HF admission for patients with neither EOB nor VE/VCO₂ slope > 36, either or both ventilatory abnormalities are shown in Figure 1.

We tested the incremental predictive value of adding CPET variables (model 2) to the model including clinical and echocardiographic parameters (model 1): AUC values of model 1 and 2 were 0.634 (95% CI 0.536 to 0.725), and 0.755 (95% CI 0.662 to 0.832), respectively (p = 0.016).

We compared the prognostic value of CPET variables in female patients with HFrEF vs HFpEF. To the best of our knowledge, this is the first report on a large population of elderly HF women to compare the additive predic-

Data are presented as Median [interquartile range] or numbers (%). A, Transmitral A Velocity; E, Transmitral E Velocity; EOB = Exercise Oscillatory Breathing; LBBB = Left Bundle Branch Block; LVEDVI = Left Ventricular End Diastolic Volume Index; LVEF = Left Ventricular Ejection Fraction; LVESVI = Left Ventricular End Systolic Volume Index; MR = Mitral Regurgitation; PVO₂ = Peak Oxygen Uptake; % PVO₂ = Percent Predicted PVO₂; VE/VCO₂ Slope = Slope of the Regression Line Relating VE to Carbon Dioxide Production.

tive value of CPET for prognosis stratifying by ventricular function. We found that ventilatory parameters, EOB and VE/VCO₂ slope, improved outcome prediction, independently of LV function, beyond established clinical and echocardiographic predictors atrial fibrillation and mitral regurgitation.

Our study describes the prognostic impact of ventilatory parameters in a “real world” female HF population, as demonstrated by the mean age (72 years) and proportion with HFpEF (54%) generally neglected in HF trials.

In our female series, atrial fibrillation and mitral regurgitation were key prognostic determinants, consistently with previous reports, irrespective of sex [1]">]">5] . HF and atrial fibrillation share the same risk factors and together have a negative synergic impact on prognosis [1]">6] , increasing mortality and morbidity both in hospitalised [1]">7] and community patients [1]">6] [1]">8] , irrespective of LV function [10] [20] . The unfavourable effect of atrial fibrillation on outcome may be even stronger in women [1]">]">9] and in HFpEF [20] .

On the other hand, global and local LV remodelling in HF is frequently associated with mitral regurgitation which in turn increases left atrial dilation and remodelling with enhanced risk of atrial fibrillation and clinical events [21] .

CPET represent an important risk stratification tool in daily clinical practice even in elderly HF patients not eligible for heart transplant [22] , in whom it is safe, feasible and prognostically reliable [23] , provided that they are physically able to perform a maximal exercise test. Previous studies addressed gender differences among relatively young populations of heart transplant candidates. Elmariah et al. [24] examined the gender effects on PVO₂ and survival in 594 ambulatory HF patients (28% women) and concluded that women had a significantly

For abbreviations see Table 1.

CPET = Cardiopulmonary exercise test. For other abbreviations see Table 1.

lower PVO₂ than men, but had better survival at all levels of exercise capacity and suggested to re-examine the current practice of uniform application of PVO₂ as an aid for transplant timing. This study focused on relatively young female transplant candidates and did not take into account novel ventilatory parameters, such as ventilatory efficiency assessed by the VE/VCO₂ slope during CPET, and EOB. Corrà et al. [2]">]">5] evaluated gender differences in PVO₂ in 529 HF patients (22% women) concluding that in women with HF cut-off PVO₂ values useful for risk stratification should be redefined. Guazzi et al. [2]">6] demonstrated that the prognostic value of PVO₂ and VE/VCO₂ slope is similar in men and women with HF and that in both genders the prognostic power of VE/VCO₂ slope is greater than that of PVO₂. Both studies involved women younger than ours (mean age 60 years and 55 years, respectively) and did not consider the incidence and the prognostic weight of EOB.

In a former study among 370 elderly HF patients (29% female), we found a higher EOB prevalence than previously reported in middle-aged cohorts. EOB and the ratio of VE/VCO₂ slope to PVO₂ emerged as the strongest ventilatory predictors of all-cause mortality, independently of ventricular function [2]">7] .

EOB is a cyclic breathing pattern that reflects a severe derangement of the ventilatory control system [2]">7] . In previous studies in middle-aged male patient cohorts [2]">8] -[30] , prevalence of EOB ranged from 12% to 35% In the only study conducted in patients with diastolic HF [31] , EOB was as common as in subjects with systolic dysfunction. In our population, the prevalence of EOB, which has not been previously specifically explored in female HF patients, was 40%, distinctly higher than values reported in transplant candidates (20% - 30%).

The novelty of our study resides first in the population analysed: while women represent over one half of HF patients in the community and show a peculiar phenotype with predominantly preserved LVEF, risk stratification using CPET has been seldom described in a sizable cohort for this patient group. The study by Corrà et al. [30] included only 8 women, whereas other reports analysed female cohorts that, although larger, were definitely younger than our patients [2]">8] [2]">]">9] . Furthermore none of these studies [2]">8] -[30] compared, within the female group, patient with preserved and reduced LVEF. Our findings demostrate that ventricular function does not independently predict prognosis when compaed with ventilatory variables.

We confirmed that prognosis, even for stable female outpatients, is poor: in the medium term one third of patients either died for cardiovascular reasons or was admitted for HF. In agreement with previous data [2]">8] -[31] , the ventilatory parameters EOB and VE/VCO₂ slope, emerged as the strongest CPET predictors of outcome in our series, both contributing incremental prognostic value independently of LVEF. We confirmed in elderly women the lack of independent prognostic value of PVO₂ [2]">6] [2]">7] [30] [31] . CPET-derived variables showed an incremental prognostic power when added to clinical and echocardiographic parameters, with very good predictive accuracy as described by the C statistic of 0.755.

The authors declare no conflict of interest.