How Do Patients With HF and Preserved EF Die?
Although several studies have examined outcomes in HFPEF, few have reported specific causes or modes of death. Why is it important to drill down to specifics? As discussed above, multiple age-related co-morbidities may co-exist in patients with HFPEF, and each co-morbidity may impart a mortality risk. Knowledge of all-cause mortality alone will not allow discernment of risk related to the co-morbidity vs. risk associated with HFPEF itself. Knowledge of cause-specific mortality, on the other hand, will aid our understanding of the pathophysiology and natural history of HFPEF as a distinct syndrome, and allow better discernment of risk that may be prevented or treated, vs. that which cannot. This knowledge then forms the basis for planning and predicting the impact of interventional strategies. Take, for instance, a patient with HFPEF who suffers an acute myocardial infarction [cause of death (COD)] leading to sudden death [mode of death (MOD)], vs. one with severe infective exacerbation of concomitant chronic lung disease (COD) leading to respiratory collapse (MOD). An intervention such as an implantable cardiac defibrillator would be expected to prevent death in the former, but not the latter, and knowledge of the relative proportions of deaths from each cause/mode in HFPEF would guide decisions on whether defibrillators should be considered or tested as a therapeutic strategy in this patient population.
Comprehensive clinical data are needed for the accurate classification of death in HFPEF. A previously published ACME system for death in heart failure is a useful guideline for the extent of information required to draw meaningful inferences: A for activity and place of death (outpatient or in-hospital), C for COD (e.g. myocardial infarction, ventricular dysfunction, or pneumonia), M for MOD (e.g. sudden death or circulatory failure), and E for events associated with death (e.g. recent hospitalizations or de-compensations). Without a priori planning, such detailed information is rarely available to allow accurate classification of death.
Causes of death are more commonly reported in epidemiological studies, where data are extracted from death certificates, medical records, and autopsy findings. Reliability of the data is increased when death ascertainment is carried out by autopsy findings or a chief medical examiner; however, this is available on a large scale in only few communities. MODs are more readily available in RCTs, where there is regular surveillance and formal adjudication of deaths by appointed outcome review committees using pre-specified criteria.
There is a lack of uniformity in definitions used to classify deaths in HFPEF in previous publications. To illustrate, mortal events have been grouped under subheadings of sudden death (MOD) and acute myocardial infarction (COD) within the same table, without distinguishing MOD vs. COD. Some other studies have also reported events according to the underlying vs. immediate causes of death, rather than COD and MOD.
Notwithstanding the variability of definitions used in different studies, the following conclusions can be derived from published outcomes studies in HFPEF (Figure 1). First, the majority of deaths in HFPEF across all studies are cardiovascular deaths. Secondly, the proportion of cardiovascular deaths varies with the type of study, ranging from 51–60% in epidemiological studies to ~70% in RCTs. The only exception was the study by Adabag et al. of HF hospitalizations in St Paul, MN, where 5-year post-discharge mortality in HFPEF was classified based on death certificates as non-cardiovascular deaths in 61% of cases. Lack of autopsy data and selection bias due to large amounts of missing EF data may have influenced these results. Nonetheless, the higher proportion of non-cardiovascular deaths in this observational study is consistent with the trend for more non-cardiovascular deaths in epidemiological studies, compared with RCTs, in HFPEF.
(Enlarge Image)
Figure 1.
Distribution of mortal events in studies of heart failure with preserved ejection fraction. Bars and numbers indicate the approximate percentage of total deaths.
Specific MODs in HFPEF have been examined in the CHARM-Preserved and I-PRESERVE clinical trials (Figure 2). Despite some differences in the populations studied, remarkable consistency was observed in MODs reported: among the ~70% majority of cardiovascular deaths, sudden death was the most common cardiac MOD (26–28% of all deaths), followed by heart failure deaths (14–21% of all deaths).
(Enlarge Image)
Figure 2.
Distribution of modes of death of patients with heart failure with preserved ejection fraction from the I-PRESERVE trial (top) and CHARM-Preserved trial (bottom). Pie charts on the left display the proportion of CV, non-cardiovascular (non-CV), and unknown deaths with the respective percentages of total death. Further breakdown of CV deaths into sudden cardiac death, heart failure deaths, stroke, myocardial infarction, and other CV deaths are shown in a separate pie chart on the right with the respective percentages of CV death.
In epidemiological studies, CODs are described albeit inconsistently. In the large Olmsted County cohort, COD was ascertained from death certificates (>75% of which were completed by a coroner or Mayo staff) and reported under the categories of non-cardiovascular deaths, coronary heart deaths and other cardiovascular deaths. Results showed that the predominant single COD in this study was non-cardiovascular death. However, cardiovascular CODs as a group were more common than non-cardiovascular CODs, and the most prominent cardiovascular COD was coronary heart disease (29% of all deaths).
Both COD and MOD were specifically reported and directly compared in the recent TIME-CHF study, where outcomes were classified according to the ACME criteria after an 18-month follow-up.
Analysis of COD vs. MOD in TIME-CHF showed differences in the proportions of cardiovascular-related events when classified by COD (~40%) vs. MOD (~66%) (Figure 3). Careful inspection of the data revealed that this difference was primarily attributed to a larger proportion of 'unknown' assignments for COD, with less 'unknown' and more frequent assignment of 'cardiovascular' for MOD, and otherwise comparable classification of non-cardiovascular events by COD or MOD. These results highlight the challenges in classification of death and the limitations of comparisons across studies using different methods, and provide a possible source of discordance between observational studies (mainly reporting COD) and clinical outcome trials (mainly reporting MOD).
(Enlarge Image)
Figure 3.
Effect of classification by causes (left) vs. modes (right) of death in heart failure with preserved ejection fraction (adapted from TIME-CHF). All numbers represent the approximate percentage of total deaths. Under causes of death, cardiac deaths are further separated into acute myocardial infarction (AMI) and LV dysfunction. Under modes of death, cardiac deaths are further separated into sudden cardiac death or circulatory failure.
Cardiovascular deaths constitute the majority of deaths in both HFPEF and HFREF (Figure 4). However, the proportion of total deaths that are cardiovascular related is higher in HFREF than in HFPEF, whether considering data from RCTs (~80% vs. ~70%) or from community-based studies (~60% vs. ~50%). Conversely, non-cardiovascular deaths constitute a larger proportion of deaths in HFPEF than in HFREF (~30% vs. ~15% from RCTs; ~50% vs. ~30% in community-based studies).
(Enlarge Image)
Figure 4.
Distribution of deaths in heart failure with preserved ejection fraction (HFPEF) vs. heart failure with reduced ejection fraction (HFREF) from studies examining outcomes in HFPEF and HFREF.
Specific MODs differ in their distribution between HFPEF and HFREF (Table 2). Sudden death and heart failure deaths constitute a larger proportion of deaths in HFREF compared with HFPEF. In fact, the leading single MOD in HFREF RCTs was sudden death (~42%), whereas that in HFPEF was non-cardiovascular death (~30%) followed closely by sudden death (26–28%). In epidemiological studies from Olmsted County, the proportion of deaths attributed to coronary heart deaths was larger in HFREF (43%) than in HFPEF (29%). Of note, the lower proportion of coronary heart deaths in HFPEF appeared to account for the lower proportion of cardiovascular deaths, or the higher proportion of non-cardiovascular deaths, in HFPEF compared with HFREF. The Olmsted County study also showed that the proportion of cardiovascular deaths decreased from 69% in 1979–1984 to 40% in 1997–2002 (P = 0.007) in HFPEF, in contrast to a modest trend in HFREF (from 77% to 64%, P = 0.08). These observations raise the intriguing notion that HFPEF patients may be increasingly spared from coronary deaths, only to succumb eventually to non-cardiovascular deaths. The role of coronary artery disease (CAD) as a risk factor for death is discussed in further detail later.
In summary, cardiovascular-related deaths comprise the majority of mortality events in HFPEF patients, with greater predominance seen in RCTs compared with epidemiological studies. Among cardiovascular deaths, sudden death and heart failure death are the leading cardiac MODs in HFPEF clinical trials. Compared with HFREF, the proportions of cardiovascular deaths, sudden death, and heart failure deaths are lower in HFPEF. Conversely, non-cardiovascular deaths constitute a higher proportion of deaths in HFPEF than in HFREF, particularly in epidemiological studies, where this difference appears to be primarily related to fewer coronary heart deaths in HFPEF.
How Do Patients With Heart Failure With Preserved Ejection Fraction Die?
Importance of Defining how Patients With Heart Failure With Preserved Ejection Fraction Die
Although several studies have examined outcomes in HFPEF, few have reported specific causes or modes of death. Why is it important to drill down to specifics? As discussed above, multiple age-related co-morbidities may co-exist in patients with HFPEF, and each co-morbidity may impart a mortality risk. Knowledge of all-cause mortality alone will not allow discernment of risk related to the co-morbidity vs. risk associated with HFPEF itself. Knowledge of cause-specific mortality, on the other hand, will aid our understanding of the pathophysiology and natural history of HFPEF as a distinct syndrome, and allow better discernment of risk that may be prevented or treated, vs. that which cannot. This knowledge then forms the basis for planning and predicting the impact of interventional strategies. Take, for instance, a patient with HFPEF who suffers an acute myocardial infarction [cause of death (COD)] leading to sudden death [mode of death (MOD)], vs. one with severe infective exacerbation of concomitant chronic lung disease (COD) leading to respiratory collapse (MOD). An intervention such as an implantable cardiac defibrillator would be expected to prevent death in the former, but not the latter, and knowledge of the relative proportions of deaths from each cause/mode in HFPEF would guide decisions on whether defibrillators should be considered or tested as a therapeutic strategy in this patient population.
Challenges in the Classification of Death in Heart Failure With Preserved Ejection Fraction
Comprehensive clinical data are needed for the accurate classification of death in HFPEF. A previously published ACME system for death in heart failure is a useful guideline for the extent of information required to draw meaningful inferences: A for activity and place of death (outpatient or in-hospital), C for COD (e.g. myocardial infarction, ventricular dysfunction, or pneumonia), M for MOD (e.g. sudden death or circulatory failure), and E for events associated with death (e.g. recent hospitalizations or de-compensations). Without a priori planning, such detailed information is rarely available to allow accurate classification of death.
Causes of death are more commonly reported in epidemiological studies, where data are extracted from death certificates, medical records, and autopsy findings. Reliability of the data is increased when death ascertainment is carried out by autopsy findings or a chief medical examiner; however, this is available on a large scale in only few communities. MODs are more readily available in RCTs, where there is regular surveillance and formal adjudication of deaths by appointed outcome review committees using pre-specified criteria.
There is a lack of uniformity in definitions used to classify deaths in HFPEF in previous publications. To illustrate, mortal events have been grouped under subheadings of sudden death (MOD) and acute myocardial infarction (COD) within the same table, without distinguishing MOD vs. COD. Some other studies have also reported events according to the underlying vs. immediate causes of death, rather than COD and MOD.
Cardiovascular vs. Non-cardiovascular Deaths in Heart Failure With Preserved Ejection Fraction
Notwithstanding the variability of definitions used in different studies, the following conclusions can be derived from published outcomes studies in HFPEF (Figure 1). First, the majority of deaths in HFPEF across all studies are cardiovascular deaths. Secondly, the proportion of cardiovascular deaths varies with the type of study, ranging from 51–60% in epidemiological studies to ~70% in RCTs. The only exception was the study by Adabag et al. of HF hospitalizations in St Paul, MN, where 5-year post-discharge mortality in HFPEF was classified based on death certificates as non-cardiovascular deaths in 61% of cases. Lack of autopsy data and selection bias due to large amounts of missing EF data may have influenced these results. Nonetheless, the higher proportion of non-cardiovascular deaths in this observational study is consistent with the trend for more non-cardiovascular deaths in epidemiological studies, compared with RCTs, in HFPEF.
(Enlarge Image)
Figure 1.
Distribution of mortal events in studies of heart failure with preserved ejection fraction. Bars and numbers indicate the approximate percentage of total deaths.
Mode of Death in Heart Failure With Preserved Ejection Fraction
Specific MODs in HFPEF have been examined in the CHARM-Preserved and I-PRESERVE clinical trials (Figure 2). Despite some differences in the populations studied, remarkable consistency was observed in MODs reported: among the ~70% majority of cardiovascular deaths, sudden death was the most common cardiac MOD (26–28% of all deaths), followed by heart failure deaths (14–21% of all deaths).
(Enlarge Image)
Figure 2.
Distribution of modes of death of patients with heart failure with preserved ejection fraction from the I-PRESERVE trial (top) and CHARM-Preserved trial (bottom). Pie charts on the left display the proportion of CV, non-cardiovascular (non-CV), and unknown deaths with the respective percentages of total death. Further breakdown of CV deaths into sudden cardiac death, heart failure deaths, stroke, myocardial infarction, and other CV deaths are shown in a separate pie chart on the right with the respective percentages of CV death.
Cause of Death in Heart Failure With Preserved Ejection Fraction
In epidemiological studies, CODs are described albeit inconsistently. In the large Olmsted County cohort, COD was ascertained from death certificates (>75% of which were completed by a coroner or Mayo staff) and reported under the categories of non-cardiovascular deaths, coronary heart deaths and other cardiovascular deaths. Results showed that the predominant single COD in this study was non-cardiovascular death. However, cardiovascular CODs as a group were more common than non-cardiovascular CODs, and the most prominent cardiovascular COD was coronary heart disease (29% of all deaths).
Cause of Death vs. Mode of Death in Heart Failure With Preserved Ejection Fraction
Both COD and MOD were specifically reported and directly compared in the recent TIME-CHF study, where outcomes were classified according to the ACME criteria after an 18-month follow-up.
Analysis of COD vs. MOD in TIME-CHF showed differences in the proportions of cardiovascular-related events when classified by COD (~40%) vs. MOD (~66%) (Figure 3). Careful inspection of the data revealed that this difference was primarily attributed to a larger proportion of 'unknown' assignments for COD, with less 'unknown' and more frequent assignment of 'cardiovascular' for MOD, and otherwise comparable classification of non-cardiovascular events by COD or MOD. These results highlight the challenges in classification of death and the limitations of comparisons across studies using different methods, and provide a possible source of discordance between observational studies (mainly reporting COD) and clinical outcome trials (mainly reporting MOD).
(Enlarge Image)
Figure 3.
Effect of classification by causes (left) vs. modes (right) of death in heart failure with preserved ejection fraction (adapted from TIME-CHF). All numbers represent the approximate percentage of total deaths. Under causes of death, cardiac deaths are further separated into acute myocardial infarction (AMI) and LV dysfunction. Under modes of death, cardiac deaths are further separated into sudden cardiac death or circulatory failure.
Classification of Death in Heart Failure With Preserved Ejection Fraction vs. Heart Failure With Reduced Ejection Fraction
Cardiovascular deaths constitute the majority of deaths in both HFPEF and HFREF (Figure 4). However, the proportion of total deaths that are cardiovascular related is higher in HFREF than in HFPEF, whether considering data from RCTs (~80% vs. ~70%) or from community-based studies (~60% vs. ~50%). Conversely, non-cardiovascular deaths constitute a larger proportion of deaths in HFPEF than in HFREF (~30% vs. ~15% from RCTs; ~50% vs. ~30% in community-based studies).
(Enlarge Image)
Figure 4.
Distribution of deaths in heart failure with preserved ejection fraction (HFPEF) vs. heart failure with reduced ejection fraction (HFREF) from studies examining outcomes in HFPEF and HFREF.
Specific MODs differ in their distribution between HFPEF and HFREF (Table 2). Sudden death and heart failure deaths constitute a larger proportion of deaths in HFREF compared with HFPEF. In fact, the leading single MOD in HFREF RCTs was sudden death (~42%), whereas that in HFPEF was non-cardiovascular death (~30%) followed closely by sudden death (26–28%). In epidemiological studies from Olmsted County, the proportion of deaths attributed to coronary heart deaths was larger in HFREF (43%) than in HFPEF (29%). Of note, the lower proportion of coronary heart deaths in HFPEF appeared to account for the lower proportion of cardiovascular deaths, or the higher proportion of non-cardiovascular deaths, in HFPEF compared with HFREF. The Olmsted County study also showed that the proportion of cardiovascular deaths decreased from 69% in 1979–1984 to 40% in 1997–2002 (P = 0.007) in HFPEF, in contrast to a modest trend in HFREF (from 77% to 64%, P = 0.08). These observations raise the intriguing notion that HFPEF patients may be increasingly spared from coronary deaths, only to succumb eventually to non-cardiovascular deaths. The role of coronary artery disease (CAD) as a risk factor for death is discussed in further detail later.
In summary, cardiovascular-related deaths comprise the majority of mortality events in HFPEF patients, with greater predominance seen in RCTs compared with epidemiological studies. Among cardiovascular deaths, sudden death and heart failure death are the leading cardiac MODs in HFPEF clinical trials. Compared with HFREF, the proportions of cardiovascular deaths, sudden death, and heart failure deaths are lower in HFPEF. Conversely, non-cardiovascular deaths constitute a higher proportion of deaths in HFPEF than in HFREF, particularly in epidemiological studies, where this difference appears to be primarily related to fewer coronary heart deaths in HFPEF.