What is the true normal: comparison of different equations for estimating peak oxygen uptake for outcome prediction in heart failure

Abstract Background and Objectives One of the strongest prognostic parameters obtained during Cardiopulmonary exercise testing (CPET) in patients with heart failure (HF) is the peak oxygen uptake (pVO2), as a standalone or as a percentage of the predicted peak oxygen uptake (ppVO2) reached. However, traditional standards by which ppVO2 is calculated have limitations, having been derived from small cohorts, lacking portability and poorly represented by women. Thus, we aim to compare the FRIEND Registry equation to previously used methods of calculating ppVO2 as it relates to prognostic value imparted by the percentage of ppVO2 (%ppVO2) reached during CPET and portability to the Portuguese HF population. Methods Single centre retrospective study of patients with HF with LVEF < 50% who underwent CPET between 2015-2021. Patients who did not reach RER≥ 1.05 were excluded from our analysis. All tests were evaluated and ventilatory thresholds determined by 3 independent, experienced operators. ppVO2 was calculated according to 1) the Wasserman equation; 2) the Wasserman equation using ideal body weight; 3) the Neder equation, 4) the SHIP equation and 5) the FRIEND equation. Our primary endpoint was a composite of CV death, urgent transplant or left ventricular assist device implantation and HF hospitalization. Results We included 238 patients (mean age 59 ± 11 years, 83% males). The HF aetiology was mostly ischemic in nature (68%), with LVEF of 34 ± 9% and median NTproBNP of 776 (2566 – 2342) pg/mL. Most patients (67%) were in class NYHA I-II and with high prevalence of GDMT (93% ACEi/ARB; 97% beta-blockers and 64% on MRA). Mean BMI in this group was 27.3 ± 4.7 kg/min. Patients performed CPET on a treadmill with a protocol adjusted to their predicted exercise capacity. Mean pVO2 was 18.1 ± 6.2mL/kg/min, corresponding to a median %ppVO2 that ranges from 51 ± 16% to 71 ± 22%, depending on the equation that is used. Mean VE/VCO2 Slope was 42 ± 13 and 48 % (n = 98) of patients showed exercise oscillatory ventilation (EOV). Mean ppVO2 as per the FRIEND and Wasserman equations was 18.9 ± 6.6 and 21.2 ± 4.5 mL/kg/min, with no statistically significant difference between both. %ppVO2 as calculated through all equations was an independent predictor of prognosis on multivariate analysis adjusted for LVEF, NTproBNP and VE/VCO2 Slope (p < 0.001 for values derived from all equations). However, the FRIEND and Wasserman equations showed a slight advantage over the remaining 2 equations, with higher AUC on ROC curve analysis (see figure). Conclusions In our population of Portuguese patients with HF, both the FRIEND registry and Wasserman equations for predicting pVO2 yielded strong prognostic power. However, these equations provide different absolute predicted VO2 values for the same patients and thus should not be used interchangeably. Standardisation of the type ppVO2 equation used should be recommended by local scientific societies.