Author + information
- Received August 5, 2014
- Revision received December 9, 2014
- Accepted December 23, 2014
- Published online June 1, 2015.
- Peter E. Carson, MD∗∗ (, )
- Inder S. Anand, MD, DPhil (Oxon)†,
- Sithu Win, MD†,
- Thomas Rector, PhD†,
- Markus Haass, MD‡,
- Jose Lopez-Sendon, MD§,
- Alan Miller, MD||,
- John R. Teerlink, MD¶,
- Michel White, MD#,
- Robert S. McKelvie, MD, PhD∗∗,
- Michel Komajda, MD††,
- Michael R. Zile, MD‡‡,
- John J. McMurray, MD§§ and
- Barry Massie, MD¶
- ∗Washington Veterans Affairs Medical Center and Georgetown University, Washington, District of Columbia
- †Veterans Affairs Health Care System and University of Minnesota, Minneapolis, Minnesota
- ‡Theresienkrankenhaus, Mannheim, Germany
- §Hospital Gregorio Maranon, Madrid, Spain
- ||University of Florida–Jacksonville, Jacksonville, Florida
- ¶San Francisco Veterans Affairs Medical Center and University of California, San Francisco, San Francisco, California
- #Montreal Heart Institute Research Center, Montreal, Quebec, Canada
- ∗∗Population Health Research Institute and McMaster University, Hamilton, Ontario, Canada
- ††Université Paris, Pitie Salpetriere Hospital, Paris, France
- ‡‡Ralph H. Johnson Veterans Affairs Medical Center and Medical University of South Carolina, Charleston, South Carolina
- §§British Heart Foundation Glasgow Cardiovascular Research Center, Glasgow, United Kingdom
- ↵∗Reprint requests and correspondence:
Dr. Peter E. Carson, Washington Veterans Affairs Medical Center, Department of Cardiology, 151D, 50 Irving Street, NW, Washington, DC 20422-0001.
Objectives The aim of this study was to investigate prognosis in patients with heart failure (HF) with preserved ejection fraction and the causes of hospitalization and post-hospitalization mortality.
Background Although hospitalizations in patients with HF with preserved ejection fraction are common, there are limited data from clinical trials on the causes of admission and the influence of hospitalizations on subsequent mortality risk.
Methods Patients (n = 4,128) with New York Heart Association functional class II to IV HF and left ventricular ejection fractions >45% were enrolled in I-PRESERVE (Irbesartan in Heart Failure and Preserved Ejection Fraction). A blinded events committee adjudicated cardiovascular hospitalizations and all deaths using predefined and standardized definitions. The risk for death after HF, any-cause, or non-HF hospitalization was assessed using time-dependent Cox proportional hazard models.
Results A total of 2,278 patients had 5,863 hospitalizations during the 49 months of follow-up, of which 3,585 (61%) were recurrent hospitalizations. For any-cause hospitalizations, 26.5% of patients died during follow-up, with an incident mortality rate of 11.1 deaths per 100 patient-years (PYs) and an adjusted hazard ratio of 5.32 (95% confidence interval: 4.21 to 6.23). Overall, 53.6% of hospitalizations were classified as cardiovascular and 43.7% as noncardiovascular, with 2.7% not classifiable. HF was the largest single cause of initial (17.6%) and overall (21.1%) hospitalizations, although, after HF hospitalization, a substantially higher proportion of readmissions were due to primary HF causes (40%). HF hospitalization occurred in 685 patients, with 41% deaths during follow-up, an incident mortality rate of 19.3 deaths per 100 PYs. The adjusted hazard ratio was 2.93 (95% confidence interval: 2.40 to 3.57) relative to patients who were not hospitalized for HF and was greater in those with longer durations of hospitalization. There were 1,593 patients with only non-HF hospitalizations, 21% of whom died during follow-up, with an incident mortality rate of 8.7 deaths per 100 PYs and an adjusted hazard ratio of 4.25 (95% confidence interval: 3.27 to 5.32). The risk for death was highest in the first 30 days and declined over time for all hospitalization categories. Patients not hospitalized for HF or for any cause had observed incident mortality rates of 3.8 and 1.3 deaths per 100 PYs, respectively.
Conclusions In I-PRESERVE, HFpEF patients hospitalized for any reason, and especially for HF, were at high risk for subsequent death, particularly early. The findings support the need for careful attention in the post-discharge time period including attention to comorbid conditions. Among those hospitalized for HF, the high mortality rate and increased proportion of readmissions due to HF (highest during the first 30 days), suggest that this group would be an appropriate target for investigation of new interventions.
Although heart failure (HF) with preserved ejection fraction (HFpEF) represents a substantial proportion of patients with HF, there remain gaps in knowledge and also controversy regarding this syndrome and its prognosis (1–3). There is agreement that hospitalizations in patients with HFpEF are common. However, although cardiovascular (CV) hospitalizations have been reported to account for the majority of events, and HF is the most frequent primary cause, other CV causes have not been well characterized (4,5). Furthermore, although recurrent hospitalizations in patients with HF are an important clinical concern, as well as a performance metric, there are no comprehensive reports of these events in patients with HFpEF.
The controversy over prognosis involves mortality in patients with HFpEF. Large registries and databases, including a recent meta-analysis, have reported that patients with HFpEF are at high mortality risk (6–8). However, randomized clinical trials in patients with HFpEF have noted markedly lower mortality rates (4,5,9,10), though still higher than those noted in older subjects without HF enrolled in clinical trials of hypertension or diabetes (11). How can these varying event rates be explained? One explanation may be that many HF registries enroll patients during acute HF hospitalizations (6,7,12,13), whereas randomized clinical trials of HFpEF have enrolled stable outpatients. The influence of HF hospitalization on mortality in clinical trials of HFpEF has not previously been reported. Furthermore, although many patients with HFpEF have important comorbidities and are commonly hospitalized for reasons other than for HF, there are no data on the influence of non-HF hospitalizations on prognosis.
Currently, because there is no proven therapy to decrease the hospitalization burden or mortality in patients with HFpEF, better understanding of the composition of events may aid in designing therapeutic strategies and clinical trials.
The I-PRESERVE (Irbesartan in Heart Failure and Preserved Ejection Fraction) clinical trial is the largest study of HFpEF to date (9,14). To evaluate the prognosis in HFpEF, we examined the composition of first and recurrent CV hospitalizations. We also report on the influence of hospitalizations on subsequent mortality.
The I-PRESERVE study design and results have been previously described (9,14). Briefly, 4,128 patients with HF symptoms and left ventricular ejection fractions of at least 45% were randomized to receive either irbesartan or placebo. Patients eligible for enrollment were required to have hospitalizations for HF within the previous 6 months and at least New York Heart Association functional class II symptoms or corroborative evidence suggesting diastolic dysfunction and New York Heart Association functional class III or IV symptoms. A clinical end point committee adjudicated all deaths and CV hospitalizations; details of the mortality adjudications have been published previously (15), and both these definitions as well as those for hospitalizations are available in the Online Appendix. The end point committee chair reviewed all hospitalizations that were classified as non-CV in serious adverse event reports, and possible CV events were sent to the full committee. Hospitalizations were defined as admissions of at least 24 h in duration or with calendar date changes if times of admission or discharge were unavailable, and those assessed as HF hospitalizations required specific intravenous therapy or augmented oral medications for that disease. Deaths were adjudicated as CV, non-CV, or of unknown cause. The major CV death subcategories were sudden cardiac death and pump failure. For the purposes of this analysis, other CV deaths included those due to myocardial infarction, stroke, CV procedures, vascular causes, and other cardiac causes.
Baseline characteristics of patients were compared using 2-sample Student t tests for continuous variables, and cross tabulations with chi-square tests for categorical variables between patients admitted for any cause or with no hospitalization, or for an admission for HF or no hospitalization. Similar comparisons were made between the same groups of patients by vital status. Data are presented as mean ± SD or percent. Echocardiographic variables are from a 745-patient substudy. The effect of the first hospitalization for any reason and for HF on the risk for mortality was determined by Cox regression analyses with the first hospitalization as a time-dependent variable. A similar analysis was done for the non-HF hospitalizations, and this group was composed of patients who did not experience HF admissions during the study. A 30-day readmission was defined as rehospitalization for any reason within 30 days of discharge from any HF hospitalization. An overall (time) unadjusted hazard ratio (HR) and an HR adjusted for established predictors of deaths in the I-PRESERVE cohort were estimated for each hospitalization end point (16). Similarly, because a previous study (17) had indicated that the risk for dying decreases with time post-discharge, a Cox model was used to estimate HRs for several discrete periods of time after discharge. Furthermore, separate analyses were done for different causes of death and the length of stay for the first hospitalization. Other types of deaths were censored in these analyses. The person-time mortality rates within each discrete time period were calculated to complement the HRs. In separate analyses, the first hospitalization for either HF or any cause was considered the index hospitalization. Subsequent hospitalizations of any type after discharge from the index hospitalization were considered recurrent events. Patients were censored at the time of death, study termination, or last known follow-up date.
A 2-sided p value ≤0.05 was considered to indicate statistical significance in this secondary analysis of the I-PRESERVE database. Stata version 13 (StataCorp LP, College Station, Texas) was used for all analyses.
During a mean follow-up period of 49.5 months, 2,278 of 4,128 patients (55%) were hospitalized at least once, and 1,850 patients were never hospitalized. The patients who were hospitalized differed from those not hospitalized in many baseline parameters, including that they were older; were more likely to have ischemic causes of HF, histories of diabetes mellitus, or myocardial infarctions; had recent HF hospitalizations; had higher N-terminal pro–brain natriuretic peptide levels, heart rates, and quality-of-life scores; and lower left ventricular ejection fractions and estimated glomerular filtration rates (Table 1). A total of 126 patients died during the initial hospitalization, and among the 2,152 survivors, 1,346 (63%) were readmitted. There were 685 patients (16.6%) hospitalized at least once for a primary diagnosis of HF, while 3,443 patients were not. Patients hospitalized for HF differed from those not hospitalized in being older and more likely to have histories of previous HF hospitalization, diabetes mellitus, and chronic obstructive pulmonary disease; they also had higher N-terminal pro–brain natriuretic peptide levels but lower left ventricular ejection fractions and estimated glomerular filtration rates (Table 2). A total of 31 patients died during the first admission for HF. Among the 654 surviving patients, 450 (69%) experienced at least 1 recurrent admission for any cause.
There were a total of 5,863 any-cause hospitalizations in I-PRESERVE. Of these, 2,278 (39%) were first hospitalizations and 61% (3,585 of 5,863) were recurrent hospitalizations. As seen in Figure 1A, 1,226 (53.8%) of the initial hospitalizations were due to CV causes, 997 (43.8%) were non-CV hospitalizations, and 55 (2.4%) were for unknown causes. A primary admission for worsening HF was the largest subcategory of first hospitalizations (400 admissions, 17.6% of all first admissions). CV procedures were the next largest CV subcategory (6.1%). Total hospitalizations by category are seen in Figure 1B. As noted above, there were 5,863 admissions for any cause: 3,141 (53.6%) CV, 2,564 (43.7%) non-CV, and 158 (2.7%) unknown. Worsening HF was the largest single subcategory, with 1,236 (21.1%) of all hospitalizations. These results are also seen in Online Table 2.
The distribution of all sequential recurrent hospitalizations without regard to the cause of the initial hospitalization is seen in Online Table 2. For the 1,346 first recurrent hospitalizations, 725 (53.9%) were for a CV cause, and 574 (42.6%) were for a non-CV cause, and 47 (3.5%) were for unknown causes. Worsening HF was the most common subcategory, 268 (19.9%). Subsequent recurrent hospitalizations had a similar distribution of cause of readmission except for worsening HF which tended to be increased.
There were 685 patients with hospitalizations for HF. Among these patients, there were 551 readmissions, for a total of 1,236 HF hospitalizations. As seen in Figure 1C, 306 (68%) of the first recurrent hospitalizations were for primary CV diagnoses, while 138 (30%) were for primary non-CV causes, with 6 (1.3%) of unknown cause. Worsening HF was the primary diagnosis in 189 (42%) of these first recurrent events, and they accounted for approximately the same proportion of second and subsequent recurrent events (Online Table 3). There were 1,371 recurrent hospitalizations for any cause after an initial HF event, 866 (63.2%) were CV, 482 (35.2%) were non-CV, and 1.7% were unknown. There were 551 (40.2%) HF rehospitalizations (Figure 1D). The distribution of subsequent recurrent hospitalizations subcategories is seen in Online Table 3 with a similar pattern of events.
Of the 1,236 HF hospitalizations, 30-day follow-up was available for 1,111 events, because of either death or loss to follow-up. There were 205 (18%) readmissions within 30 days of HF hospitalizations. Of these, 150 readmissions (73%) were due to CV causes, with 99 (48%) being due to worsening HF (Online Table 4, Online Figure 1).
Mortality risk post-hospitalization
Mortality risk post–any-cause hospitalization
There were 2,278 patients with hospitalizations for any reason: 126 (5.5%) died during the initial hospitalization, and another 603 (26.5%) died by the end of the trial, resulting in 32% deaths (729 of 2,278) in this group and 28% (603 of 2,152) among those who were discharged. Among the 1,850 patients who did not have hospitalizations during the study, 152 (8.2%) died. The complete list of baseline characteristics of patients hospitalized versus never hospitalized and vital status is shown in Online Table 1A. Patients who died after hospitalization differed from those who did not in a number of characteristics associated with worse outcomes.
After discharge, the unadjusted HR was 8.53 (95% confidence interval [CI]: 7.06 to 10.31) and the adjusted HR was 5.32 (95% CI: 4.21 to 6.73) for all-cause mortality compared with those never hospitalized for any reason (Table 3). The overall death rate in the patients after any hospitalization was 11.1 deaths/100 patient-years [PYs], while in those never hospitalized, it was 1.3 deaths/100 PYs. The risk for mortality and the incident mortality rate were highest in the first month and decreased thereafter, although they remained significantly elevated throughout the entire follow-up period in both the adjusted and unadjusted analyses (Table 4).
Mortality risk post–HF hospitalization
During I-PRESERVE, among the 685 patients with at least 1 hospitalization for HF, 31 (4.5%) died during the initial hospitalization and another 270 (39.4%) died by the end of the trial, resulting in 43.9% (301 of 685) deaths in this group and 41% (270 of 654) among those who were discharged. In contrast, there were 3,443 patients without HF hospitalizations, and 580 of these patients (16.8%) died during the study. Those hospitalized for HF differed in many baseline characteristics from those not hospitalized, as indicated above, and those who died differed from those alive in both subgroups (Online Table 1B). Among patients hospitalized for HF, those who died differed in a number of characteristics, including that they were older, had higher N-terminal pro–brain natriuretic peptide levels, had higher heart rates, had lower left ventricular ejection fractions, and had worse renal function compared with those who survived.
The overall (time) unadjusted HR for subsequent all-cause mortality in patients discharged after HF hospitalization, compared with a concurrent group that was never hospitalized for HF, was 4.86 (95% CI: 4.19 to 5.76) and was 2.93 (95% CI: 2.40 to 3.57) when adjusted for baseline predictors of mortality (16) (Table 3). The overall mortality rate post–HF hospitalization was 19.3 deaths/100 PYs. The adjusted mortality risk and incident mortality rate were highest in the first month (9.39 and 46 per 100 PYs, respectively) with a subsequent decrease, although the risk for dying remained 2 to 3 times higher until the end of follow-up (Table 4). For those not hospitalized for HF, the mortality rate was 3.8 deaths/100 PYs.
Longer duration of the initial HF hospitalization was associated with increased risk for mortality post-discharge compared with shorter hospitalizations (Figure 2). Patients with HF hospitalizations were at particularly high risk for subsequent HF death. The HRs for different modes of death are graphically displayed in Figure 3. The full analysis on mode of death is available in the Online Appendix.
Mortality risk post–non-HF hospitalizations
Because the entire any-cause hospitalization cohort included patients with HF admissions but the majority of patients were hospitalized for non-HF reasons, we analyzed for mortality after non-HF events and excluded patients who had HF hospitalizations at any point. There were 1,593 patients with only non-HF hospitalizations: 109 (6.8%) died during the initial hospitalization, and another 319 died by the end of the trial, resulting in 27% deaths (428 of 1,593) in this group and 21% (319 of 1,484) among those who were discharged.
After discharge from the first hospitalization for a non-HF reason, the unadjusted HR was 6.55 (95% CI: 45.32 to 8.07), and the adjusted HR was 4.25 (95% CI: 3.27 to 5.32) for all-cause mortality compared with the group that was never hospitalized for a non-HF cause (Table 3). The overall death rate in the patients after non-HF hospitalization was 8.73 deaths/100 PYs. The risk for mortality and the incident mortality rate were again highest in the first month and decreased by 6 months, although they remained significantly elevated throughout the entire follow-up period in both the adjusted and unadjusted analyses (Table 4).
Cumulative mortality post HF, non-HF, and no hospitalization
Figure 4 displays a Kaplan-Meier curve for cumulative mortality incidence for subjects after HF or non-HF hospitalization and also for those never hospitalized since randomization.
In I-PRESERVE, we also found that 55% of patients experienced at least 1 hospital admission, and the majority of hospitalized patients who survived the admission (63%) were readmitted. CV hospitalizations were more common than those for non-CV causes. Whereas HF hospitalizations were the most common type of CV event, they were a minority of all hospitalizations (∼20%), whether they were the first or subsequent hospitalizations. However, after HF hospitalization, more recurrent events were for HF (∼40%). During the 30 days after discharge, nearly one-half of readmissions were for HF.
In examining post-hospitalization mortality, we found that patients with HFpEF were at high risk for mortality after any hospitalization. The risk was highest early after discharge and was independent of measured prognostic variables available in this database. We observed the highest post-hospitalization mortality event rates in those admitted primarily for HF, but those hospitalized for non-HF reasons also experienced substantial mortality risk. Patients who were not hospitalized for any reason had a low mortality rate.
Hospitalizations in HFpEF
While it is known that HF patients are frequently hospitalized, the primary reasons for these events have not been well defined, particularly in HFpEF. HF registries and other observational databases have reported on mortality and cause of hospitalizations during follow-up but do not provide detailed data on adjudicated causes of hospitalizations (12,13,18). Clinical trials are often designed to look at hospitalizations, particularly for HF as part of a composite endpoint, as a time to first event analysis. Thus, other causes of hospitalization are often not reported and subsequent hospitalizations not adjudicated. Further most previous reports on the hospitalization burden in HF have been largely from patients with a reduced ejection fraction (HFrEF) which also found that the most hospitalizations were for CV causes, with the majority of these for HF (19–22). However, only 2 HFrEF studies, EVEREST (Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study with Tolvaptan) and COPERNICUS (Carvedilol Prospective Randomized Cumulative Survival) have presented detailed description of CV hospitalizations (21,22). Not surprisingly in these advanced HFrEF patients, HF hospitalizations were the predominant component of CV hospitalizations: In EVEREST, HF hospitalizations were 46% of any cause and 76% of CV; while in COPERNICUS (placebo arm) HF hospitalizations were 53% of any cause and 64% of CV. Other CV causes were infrequent. Treatments that have reduced hospitalizations in studies of HFrEF have showed significant reduction in cardiovascular hospitalizations, which in turn has been due largely to a reduction in the common HF hospitalizations.
As noted in this paper in a HFpEF cohort, CV hospitalizations were also the majority of events and there was a wide range of CV causes with HF hospitalizations a lower proportion than in HFrEF. The only previously published data describing causes of hospitalization in an HFpEF cohort are the unadjudicated results from the ancillary DIG (Digitalis Investigation Group) study that only described the first hospitalization of each type, and reported a wide range of CV hospitalizations consistent with the current data (4). The population enrolled in I-PRESERVE was, as in HFpEF epidemiologic databases, an older cohort with multiple CV risk factors, particularly hypertension, as well as multiple non-CV comorbidities. Not surprisingly, hospitalized patients had more comorbidities and more unfavorable physiologic variables than those not hospitalized, including those previously reported to be associated with worse outcomes in I-PRESERVE (9). In considering the composition of the hospitalizations, it is notable that many CV causes were infrequent despite the age and comorbidities present. For example, while it might be expected that events related to cardiac ischemia would be infrequent given the modest numbers with a previous myocardial infarction or ischemic etiology, it is perhaps surprising that hospitalizations due to atrial arrhythmias or cerebrovascular accidents were also relatively uncommon. CV procedures were the most frequent non HF admission subcategory and these were <6% of recurrent events. The wide range of CV hospitalizations and common non-CV hospitalizations, together with a lower proportion of HF events, make total or CV hospitalization reduction difficult to achieve in a cohort with HFpEF.
Nonetheless, HF was the most common single cause of hospitalization in the overall I-PRESERVE cohort, with a proportion similar to reports from CHARM (Candesartan in Heart Failure Assessment of Reduction in Mortality and Morbidity) Preserved (5) as well as the ancillary DIG study (4). These events would be a target for therapy. However, in this overall cohort of patients, who were clinically stable outpatients when enrolled, over a mean follow-up of more than 49 months, only 16% experienced HF hospitalizations, which represented only 21% of total events. Therefore, an intervention to reduce HF admissions would require a large clinical trial with a long follow-up time. A recent approach to the challenge of such clinical trials has been statistical analyses, which have included recurrent hospitalizations (22,23). Alternatively, patients hospitalized for HF might provide an attractive target for an intervention, given that a greater proportion of HF events as suggested by the present data in which 40% of rehospitalizations were for this cause.
This study also examined recurrent hospitalizations in the 30-day period after HF admission. This has become an important hospital performance measure. There have been no previous reports from a clinical trial or registry to provide insight on 30-day readmission rates in a HFpEF population and only 1 from a HFrEF clinical trial: the A-HeFT (African-American Heart Failure Trial); 23.6% 30-day readmission rate after the first HF hospitalization in the placebo group (24). The I-PRESERVE data also indicates a slightly lower readmission rate (18%) but there still was a high proportion of 30-day readmissions due to CV and HF cause.
Post-hospitalization mortality in HFpEF
Our findings demonstrate that patients with HFpEF enrolled as stable outpatients in a randomized clinical trial have substantial mortality risk after hospitalization for any cause, particularly after HF admission. There are no comparable data from a randomized clinical trials involving patients with HFpEF, but there have been 2 previous reports from predominantly or entirely HFrEF studies that have used a similar time-dependent statistical methodology. Both SHIFT (Systolic Heart Failure Treatment With the IF Inhibitor Ivabradine Trial) (25) and CHARM (26) found that mortality risk was highest early after HF hospitalization and decreased over time, though it remained substantially elevated compared with patients never hospitalized or yet to be hospitalized. The methodology in the present analysis differs slightly in that the patient time before a hospitalization, so-called immortal time, was not included in the calculation of risk ratios, but the pattern of risk over time is consistent with these previous studies. In considering post–HF hospitalization mortality rates, those in I-PRESERVE are substantially lower than the unadjusted event rates reported from the HFrEF cohort in SHIFT, a finding that would be consistent with the data from the recent MAGGIC (Meta-Analysis Global Group in Chronic Heart Failure) meta-analysis of cohort studies and clinical trial databases (8). The CHARM Program publication did not provide separate mortality rates for its HFpEF cohort (37% of total patients) or provide event rates for the overall cohort (18). The only previous reports of post–HF hospitalization mortality results in a HFpEF cohort have come from registries. For example, in the DIAMOND (Danish Investigations of Arrhythmia and Mortality) registry, the post–HF hospitalization mortality rate was 19% at 1 year (27).
Why were patients hospitalized for HF in I-PRESERVE at such high risk? The baseline characteristics of those hospitalized for HF, including both comorbidities and indicators of more severe disease, differed from those not hospitalized, and those who died differed further from those who did not. However, even after adjustment for the baseline predictors associated with mortality risk in the overall I-PRESERVE database, the HF hospitalization event remained independently associated with a worse outcome. This may seem surprising, because HF hospitalization could be due to progression of the disease but also could be due to other causes, such as nonadherence, poor follow-up, or inadequate therapy. Although the present analysis does not provide a good explanation for the high post–HF hospitalization mortality risk, the mode-of-death analysis showed a particularly high risk for HF death early after discharge, suggesting worsening of the disease process in these subjects. The hospital experience itself may contribute, as noted below. The finding that an HF hospitalization is a salient event in a patient’s course, which requires careful post-discharge attention, has been reported in numerous HFrEF databases (4,23,26), but our data extend these findings to patients with HFpEF, indicating a need for careful attention early after discharge, although specific therapies for this period of high risk have not been established.
We also found a high risk for death after a nonelective hospitalization for any cause, with a similar pattern to that observed after HF admission. However, these findings need to be interpreted with caution, because both initial and subsequent HF hospitalizations were included in all-cause hospitalizations and therefore would affect the risk of the time intervals and event rates of the all-cause events. For this reason, we have included an analysis of patients who were hospitalized for any cause but not for HF. We observed a lower risk for mortality after these hospitalizations than that observed after HF hospitalization, but the event rate for these patients is still substantially higher than that for patients without hospitalizations. The excess risk associated with any hospitalization, HF or not, may be related to factors associated with the in-hospital experience. As noted by Krumholz (28), these include alterations in cognitive and physiologic function, including deconditioning and malnutrition, which then also affect the early post-discharge phase. These factors are suggested in the data of the current study, in which prolonged HF hospitalization was associated with the highest early post-discharge mortality risk. An older population is particularly vulnerable to what has been described as the “post-hospital syndrome.” This excess risk also points to a need for careful attention to the treatment of comorbidities in HFpEF, as suggested by Ather et al. (29) as part of a strategy to reduce overall mortality risk.
Our study also identified a group of patients at lower mortality risk who were not hospitalized for HF and an even lower risk group who were never hospitalized for any reason. The patients not hospitalized for HF differed from those hospitalized for HF in many characteristics, as did the group hospitalized for any reason compared with those never hospitalized. The finding that most patients in I-PRESERVE were not hospitalized for HF, and many were not hospitalized at all, indicates that some subjects may have a mild variant of the disease, and others may not have HF at all. Although the I-PRESERVE inclusion criteria required the presence of signs and symptoms as well as structural and functional abnormalities, many of these features, such as shortness of breath and ankle edema, are nonspecific, underscoring the difficulty in diagnosing HFpEF (30).
The finding that hospitalized patients are at higher risk than those not hospitalized is not surprising, although this has not been previously quantitated for non-HF hospitalization or in the HFpEF population. This underscores the need to focus more on the clinical care of such hospitalized patients. These data further support the importance of enrolling patients during acute decompensated HF hospitalizations in clinical trials of patients with HFpEF, as the enrollment of stable outpatients is more likely to be dominated by low-risk subjects. This problem is magnified in HFpEF because of the lack of specificity of the clinical symptoms and findings that are often present. Our data suggest that enrolling subjects during their HF hospitalizations would result in a cohort with a high subsequent risk for HF hospitalizations and mortality. It is important to note that the HF hospitalizations in the I-PRESERVE database were carefully adjudicated by an event committee. It should also be emphasized that when HF hospitalization is used as an inclusion criterion in clinical trials, the definition of HF hospitalization needs to be carefully considered. This might avoid the difficulties encountered in the TOPCAT (Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist) study, in which a low-risk cohort was reported in those entry criteria pathway that was a history of hospitalization with a “a major component of the care provided” (10,31).
Strengths and limitations
We studied well-characterized patients in the I-PRESERVE trial, the largest study of patients with HFpEF, which had adjudicated data and excellent follow-up data. In addition, we were able to adjust for prognostic variables previously described in this population. However, this was a secondary analysis of data from a randomized controlled trial, and patients were defined according to events occurring after enrollment. Furthermore, some of the baseline characteristics may have changed by the time of the hospitalization. The inclusion and exclusion criteria, including a history of hospitalization within 6 months and limitations on non-CV comorbidities, may have contributed to the distribution of hospitalizations over the duration of the study, although the proportion of hospitalizations for HF is similar to that in a recent community-based report of hospitalizations after a new diagnosis of HF (17). Judgments about the mode of death can be influenced by a recent HF hospitalization. Non-CV hospitalizations did not undergo full committee review, as these events were initially assessed by the sponsor, on the basis of serious adverse events reports, and then reviewed by the events committee chair, who sent potential CV hospitalizations to review.
In the HFpEF population of I-PRESERVE, hospitalizations were common, and the majority of patients were readmitted. CV causes for hospitalization were most common, but non-CV hospitalizations were frequent, as both first and recurrent events. After an initial HF hospitalization, a greater proportion of recurrent admissions were due to HF, particularly during the initial 30-day interval. Additionally, patients hospitalized for any reason but particularly for HF were at high risk for subsequent death. Substantial post-hospitalization mortality risk was present also in patients with non-HF hospitalizations, suggesting a need for careful attention to comorbid conditions. These findings support the need for intensive follow-up in the early post-discharge time period. The outcomes after a HF hospitalization also suggest that this would be an optimal group in which to test interventions in future HFpEF clinical trials.
COMPETENCY IN MEDICAL KNOWLEDGE: Hospitalizations for HF represent a minority of the overall hospitalization burden in patients with HFpEF. However, there is a high post-hospitalization mortality risk after these events, and subsequent hospitalizations are more likely to be for HF. These patients need careful attention post-hospitalization, particularly early, when risk is highest. Additionally, patients admitted for non-HF causes also are at substantial risk for subsequent mortality and therefore also require careful follow-up with attention to comorbidities. Finally, patients with the diagnosis of HFpEF but not requiring hospitalization are at low mortality risk.
TRANSLATIONAL OUTLOOK: Clinical trials in HFpEF have been hampered by low event rates, due partly to large proportions of enrolled subjects who do not experience outcomes such as hospitalizations and death. This may indicate a mild version of the disease or an inaccurate diagnosis. A clinical trial composed of subjects admitted for HF would have the advantage of greater certainty in the diagnosis of a condition that can be difficult to diagnose correctly and a high event rate, including death and recurrent HF hospitalizations. Better understanding of patients with HFpEF who do not experience clinical events is necessary in considering large clinical trials that enroll stable outpatients.
Bristol-Myers Squibb sponsored the I-PRESERVE trial. The views expressed herein do not necessarily represent the views of the U.S. Department of Veterans Affairs or the U.S. government. Dr. Rector was supported by resources and facilities of the Minneapolis Veterans Affairs Health Care System. Drs. Carson, Anand, Haass, Lopez-Sendon, Miller, Teerlink, White, McKelvie, Komajda, Zile, McMurray, and Massie were consultants to Bristol-Myers Squibb. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- heart failure
- heart failure with preserved ejection fraction
- heart failure with reduced ejection fraction
- hazard ratio
- Received August 5, 2014.
- Revision received December 9, 2014.
- Accepted December 23, 2014.
- American College of Cardiology Foundation
- Vasan R.S.,
- Larson M.G.,
- Benjamin E.J.,
- et al.
- Hogg K.,
- Swedberg K.,
- McMurray J.J.
- Packer M.
- Ahmed A.,
- Rich M.W.,
- Fleg J.L.,
- et al.
- Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC)
- Campbell R.T.,
- Jhund P.S.,
- Castagno D.,
- Hawkins N.M.,
- Petrie M.C.,
- McMurray J.J.
- Yancy C.W.,
- Lopatin M.,
- Stevenson L.W.,
- et al.
- Fonarow G.C.,
- Stough W.G.,
- Abraham W.T.,
- et al.
- Zile M.R.,
- Anand I.S.,
- et al.,
- for the I-PRESERVE Investigators
- Komajda M.,
- Carson P.E.,
- Hetzel S.,
- et al.
- Dunlay S.M.,
- Redfield M.M.,
- Weston S.A.,
- et al.
- Hjalmarson A.,
- Goldstein S.,
- Fagerberg B.,
- et al.,
- for the MERIT-HF Study Group
- O’Connor C.M.,
- Miller A.B.,
- Blair J.E.,
- et al.
- Packer M.,
- Fowler M.B.,
- Roecker E.B.,
- et al.
- Rogers J.K.,
- McMurray J.J.,
- Pocock S.J.,
- et al.
- Anand I.S.,
- Sithu W.,
- Rector T.S.,
- et al.
- Solomon S.D.,
- Dobson J.,
- Pocock S.,
- et al.,
- for the Candesartan in Heart Failure,
- Assessment of Reduction in Mortality and Morbidity (CHARM) Investigators
- Gustafsson F.,
- Torp-Pedersen C.,
- Brendorp B.,
- Seibaek M.,
- Burchardt H.,
- Kober L.,
- for the Diamond Study Group
- Ather S.,
- Chan W.,
- Bozkurt B.,
- et al.
- Campbell R.T.,
- McMurray J.J.
- Pfeffer M.A.,
- Clagget B.,
- Assmann S.F.,
- et al.