reviewed by Stephan Müller
Almost 15 years ago, the Exercise in Diastolic Heart Failure (Ex-DHF) pilot study demonstrated in a group of 64 patients with heart failure and preserved ejection fraction (HFpEF) that 3 months of supervised endurance/resistance training improved peak oxygen uptake (VO2) 1. The effect was quite impressive: peak VO2 increased with 2.6 mL/kg/min in patients randomized to training, while it reduced with -0.7 mL/kg/min in controls. Subsequently, larger trials were designed to confirm this finding. The Optimizing Exercise Training in HFpEF (OptimEx-Clin) study was published in 2021, and found a smaller increase in peak VO2 in patients randomized to moderate continuous endurance training (+1.6 mL/kg/min) or high-intensity interval training (+1.1 mL/kg/min) after 3 months, versus -0.6 mL/kg/min in controls 2. The difference was statistically significant; however, it did not meet a prespecified minimal clinically important difference of +2.5 mL/kg/min.
Around the same time, a follow-up trial to Ex-DHF was designed including 322 patients randomized to endurance + resistance training across 11 centres in Germany and Austria. The primary endpoint was defined as the modified Packer score, which consists of a hierarchical composite endpoint of all-cause mortality, hospitalizations potentially related to HF or exercise, and a “clinical score”. The latter included change in peak VO2, change in diastolic function (E/e’), change in New York Heart Association (NYHA) class, and change in Global Self Assessment score (a clinical questionnaire). While ‘last patient out’ was in 2017, the study results were published in January 2025 in Nature Medicine 3. At 12 months, the Packer score was improved in 20.5% of patients randomized to training, versus 8% randomized to usual care; this difference was not statistically significant (p=0.17).
So what conclusions should we draw regarding exercise training in HFpEF from this study? Fortunately, looking at this complex outcome in more detail, there are still important signals of benefit. All-cause mortality did not improve, neither did the number of HF hospitalizations. The improvements in Packer score were driven by a larger proportion of patients improving their ‘clinical score’, in turn driven by better peak VO2 (p<0.001) and better NYHA class (p<0.001). Around 35% of patients randomized to training increased their peak VO2, even though the defined margin for improvement was quite large (+15%), and 33% improved NYHA class. Also, a more ‘modern’ approach to hierarchical outcomes using generalized pairwise comparisons (“win ratio”) was introduced only after the design of Ex-DHF 4. When analysing the trial according to this method, exercise training had significantly more “wins” than usual care (win ratio: 1.32, 95% confidence interval 1.01-1.73, p=0.04).
The increase in peak VO2 demonstrated a plateau after 9 months, emphasizing the need for progressive training, including repeated re-assessment and intensity adaptation 5. Hospitalizations were numerically higher in the training group (33 versus 23 patients), potentially due to increased monitoring rather than adverse effects (11 events were considered related to exercise, but 7 of these were detected before exercise was started, which may be considered beneficial). Diastolic function (E/e′) remained unchanged, aligning with previous studies demonstrating mostly peripheral adaptations to exercise in HFpEF 6. Finally, adherence was low: only 48% of patients performed at least 2 sessions per week overall. In a post hoc analysis, higher adherence to exercise training was significantly associated with better results of the modified Packer score (test for trend in proportions: P=0.002).
So, let’s not throw out the baby with the bathwater. Ex-DHF underscores the importance of structured training in HFpEF, but at the same time highlights challenges in adherence, endpoint selection, and disease progression. Future trials should apply refined composite endpoints, differentiate training effects, and explore strategies to enhance participation.