Background: Heart failure with preserved ejection fraction (HFpEF) is common, yet there is currently no consensus on how to define HFpEF according to various society and clinical trial criteria. How clinical and hemodynamic profiles of patients vary across definitions is unclear. We sought to determine clinical characteristics, as well as physiologic and prognostic implications of applying various criteria to define HFpEF. Methods: We examined consecutive patients with chronic exertional dyspnea (New York Heart Association class II to IV) and ejection fraction ≥50% referred for comprehensive cardiopulmonary exercise testing with invasive hemodynamic monitoring. We applied societal and clinical trial HFpEF definitions and compared clinical profiles, exercise responses, and cardiovascular outcomes. Results: Of 461 patients (age 58±15 years, 62% women), 416 met American College of Cardiology/American Heart Association (ACC/AHA), 205 met European Society of Cardiology (ESC), and 55 met Heart Failure Society of America (HFSA) criteria for HFpEF. Clinical profiles and exercise capacity varied across definitions, with peak oxygen uptake of 16.2±5.2 (ACC/AHA), 14.1±4.2 (ESC), and 12.7±3.1 mL·kg ^-1 ·min ^-1 (HFSA). A total of 243 patients had hemodynamic evidence of HFpEF (abnormal rest or exercise filling pressures), of whom 222 met ACC/AHA, 161 met ESC, and 41 met HFSA criteria. Over a mean follow-up of 3.8 years, the incidence of cardiovascular outcomes ranged from 75 (ACC/AHA) to 298 events per 1000 person-years (HFSA). Application of clinical trial definitions of HFpEF similarly resulted in distinct patient classification and prognostication. Conclusions: Use of different HFpEF classifications variably enriches for future cardiovascular events, but at the expense of not including up to 85% of individuals with physiologic evidence of HFpEF. Comprehensive phenotyping of patients with suspected heart failure highlights the limitations and heterogeneity of current HFpEF definitions and may help to refine HFpEF subgrouping to test therapeutic interventions.