Author + information
- Received August 17, 2015
- Revision received September 4, 2015
- Accepted September 5, 2015
- Published online January 1, 2016.
- Warren K. Laskey, MD, MPH∗∗ (, )
- Jingjing Wu, MS†,
- Phillip J. Schulte, PhD†,
- Adrian F. Hernandez, MD, MHS†,
- Clyde W. Yancy, MD, MS‡,
- Paul A. Heidenreich, MD, MS§,
- Deepak L. Bhatt, MD, MPH‖ and
- Gregg C. Fonarow, MD¶
- ∗Division of Cardiology, Department of Medicine, University of New Mexico School of Medicine, Albuquerque, New Mexico
- †Duke Clinical Research Institute, Duke University, Durham, North Carolina
- ‡Division of Cardiology, Department of Medicine, Northwestern University Feinberg School of Medicine, Chicago, Illinois
- §Division of Cardiology, Department of Medicine, Veterans’ Affairs Palo Alto Health Care System, Palo Alto, California
- ‖Brigham and Women’s Hospital Heart and Vascular Center and Harvard Medical School, Boston, Massachusetts
- ¶Division of Cardiology, Department of Medicine, David Geffen School of Medicine, University of California at Los Angeles, Los Angeles, California
- ↵∗Reprint requests and correspondence:
Dr. Warren Laskey, Division of Cardiology, Department of Internal Medicine, University of New Mexico School of Medicine, MSC10-5550, 1 University of New Mexico, Albuquerque, New Mexico 87131.
Objectives This study assessed the association between pulse pressure (PP) and adverse outcomes at 1 year in patients hospitalized for heart failure (HF).
Background PP has been shown to be predictive of the development of HF. The value and utility of PP assessment in patients with prevalent HF is less clear.
Methods We conducted a retrospective cohort study from clinical registry data linked to Medicare claims for 40,421 HF patients entered in the Get With the Guidelines-HF program. Cox proportional hazards models were used to estimate the association between discharge PP and all-cause mortality and the composite outcome of all-cause mortality/readmission by 1 year.
Results A nonlinear association between PP and mortality (expressed as hazard ratio [HR] per 10-mm Hg increase in PP) was observed in patients with HF and reduced (<0.40) ejection fraction (EF). Risk decreased as PP increased up to 50 mm Hg (adjusted HR: 0.946; 95% confidence interval [CI]: 0.900 to 0.995; p = 0.03). When PP was ≥50 mm Hg, risk increased as PP increased (adjusted HR: 1.091; 95% CI: 1.050 to 1.135; p < 0.001). In patients with HF and preserved EF (≥0.40), there was a significant association between PP and mortality with risk increasing as PP increased, although the magnitude of the risk was significantly impacted by systolic blood pressure (SBP). Qualitatively similar observations were obtained for the composite outcome and use of EF ≥0.50 to define HF with preserved EF.
Conclusions The association between PP at hospital discharge and 1-year outcomes is a function of HF phenotype, SBP, and absolute PP.
Arterial pulse pressure (PP) has been shown to be associated with all-cause (1) as well as cardiovascular mortality in cardiovascular disease-free (2) and older hypertensive (3) populations. PP has also been shown to be a strong determinant of incident coronary heart disease (4,5), atrial fibrillation (6), and recurrent events after myocardial infarction (7,8).
Prospective cohort studies in patients without heart failure (HF) at baseline have demonstrated that PP is strongly and positively associated with subsequent development of HF (9,10). However, an inverse association between PP and adverse cardiovascular outcomes has been described in other HF patient populations (11–16). There is also a growing awareness of the impact of left ventricular (LV) ejection fraction (EF) on this association (16).
The purpose of the present study was to assess the association between arterial PP at hospital discharge and clinical outcomes by 1 year in a broad spectrum of HF patients receiving contemporary, guideline-based therapy.
The Get With The Guidelines-Heart Failure (GWTG-HF) program is among the largest quality improvement initiatives focusing on patients hospitalized with HF and has been previously described (17–20). Patients hospitalized with new or worsening HF as primary diagnosis or patients who developed significant HF symptoms such that HF was the primary discharge diagnosis were included in the registry starting January 1, 2005. Patients were enrolled in the program regardless of EF. Hospitals from all regions of the United States are represented, and a variety of institutions participated, from community hospitals to large tertiary medical centers.
Data collected for each patient included demographics, medical/surgical history, admission medications, admission and discharge vital signs, physical examination findings, rhythm on admission, serum laboratory tests, pharmacological and nonpharmacological interventions, in-hospital outcomes, and discharge status information. Trained personnel (e.g., dedicated research coordinators and/or quality improvement/assurance personnel) entered chart-abstracted data, using standardized definitions. All participating hospitals were required to submit GWTG-HF protocol to their institutional review board for approval. Because data collected were used for hospital quality improvement, sites were granted a waiver of informed consent under the common rule. Quintiles are the data collection coordination center for the American Heart Association/American Stroke Association Get With The Guidelines programs. The Duke Clinical Research Institute served as the data analysis center and had an agreement to analyze de-identified data for research purposes.
We obtained clinical data from the GWTG-HF registry and Medicare claims data from the Centers for Medicare and Medicaid Services (CMS). Medicare data included inpatient claims and corresponding denominator files from January 1, 2005, through December 30, 2010. We linked data from the GWTG-HF registry to the research identifiable inpatient claims data with the use of indirect identifiers admission date, discharge date, sex, and age or date of birth (21). Combinations of these identifiers are almost always unique, enabling identification of registry hospitalizations in Medicare claims data. For patients with multiple linked hospitalizations in the registry, we selected the first hospitalization for analysis.
From January 1, 2005, to December 30, 2010, there were 53,484 CMS-matched patients discharged alive from 265 fully participating hospitals. From these patients, we excluded: 1) 2,144 patients who did not have eligible fee-for-service Medicare at the time of index HF hospitalization discharge; 2) 1,811 patients who died in-hospital; 3) 4,719 patients (8.8%) missing discharge blood pressure information; 4) 8 patients who were using LV-assist devices; 5) 1,891 patients who left against medical advice or were discharged to short-term hospital or hospice service; and 6) 2,490 patients whose EF status was missing. The final sample size for analysis was 40,421 patients from 258 sites.
Pulse pressure is defined as the difference between the brachial arterial systolic and diastolic pressures obtained, in conformance with the local standard of care for assessment of vital signs. In this study we used the PP on the basis of the systolic and diastolic blood pressures recorded at or closest to hospital discharge as the covariate of interest.
The primary outcome variable was all-cause mortality by 1 year. A secondary outcome was all-cause mortality or readmission by 1 year.
Overall baseline patient characteristics were compared across PP quartiles. Medians and 25th to 75th percentiles were determined for continuous variables and compared using Kruskal-Wallis tests. Percentages are reported for categorical variables and comparisons made with chi-square tests.
We fitted models separately for HF patients with reduced EF (i.e., EF <0.40 [HFrEF]) and HF patients with preserved EF (i.e., EF ≥0.40 [HFpEF]) to allow for different relationships among patient and hospital characteristics and outcome(s) in each group. The association of PP with each outcome for HFrEF patients was assessed using unadjusted and adjusted cause-specific Cox proportional hazards regression models for 1-year follow-up. The functional form of PP was assessed by fitting a flexible restricted cubic spline (RCS) transformation and comparing to a linear fit. Both the HFrEF and HFpEF groups suggested a nonlinear relationship between PP and outcomes. Inspection of the plotted RCS fit suggested that a linear spline with 1 knot would fit the data well while providing easier interpretation. We tested a number of knot points over the observed range of PP and selected 50 mm Hg as the 1 maximizing model likelihood. A likelihood ratio test suggested the linear spline fit the data well.
Proportional hazard assumptions were assessed using plotted Schoenfeld residuals versus time, and no violations were detected. Colinearity between predictor variables in the final model was assessed using variance inflation factors. Variance inflation factor values >5 were examined, and in the presence of a strong correlation between 2 covariates, 1 would be dropped from the model.
Adjusted models included age, sex, race (white vs. other), insurance status, anemia, ischemic history, cardiovascular accident/transient ischemic attack, atrial fibrillation/flutter, diabetes, hyperlipidemia, hypertension, chronic obstructive pulmonary disease or asthma, peripheral artery disease, renal insufficiency, smoking, systolic blood pressure (SBP) on admission, heart rate, SBP at discharge (to check interaction with discharge PP), serum sodium, blood urea nitrogen concentration, hospital region (Midwest, South, West vs. Northeast), hospital type (teaching/nonteaching), number of beds, rural versus nonrural hospital, defect-free compliance score (defined as the frequency of patients with 100% compliance with all GWTG-HF–defined quality performance measures and included evidence-based pharmacotherapy). Model covariates were chosen on the basis of their clinical relevance or known association with the exposure of interest (PP) and the specific outcome. Single imputation of patient characteristics was used to reduce “missingness” in models. Missing PP values (principal analysis) were not imputed. Missing values for categorical variables were imputed to the most common category; continuous variables were imputed to the mean. To determine whether SBP might serve as an effect modifier of the PP-outcome relationship, an interaction term was included in adjusted models. When significant, the relationship between PP and outcome is described as values of SBP; when not significant, a reduced model further adjusts for SBP. The analysis was repeated in similar fashion for HFpEF patients.
Analyses were conducted in which the continuous variable of interest, PP, was transformed to a categorical variable. However, results are reported herein using the RCS transformation, thereby maintaining the interval properties of a continuous variable as well as statistical power (22).
Similar analyses were conducted for HFpEF, defined as EF >0.50 and HFrEF <0.50.
Two-sided p values of <0.05 were considered statistically significant for all tests. Analyses were performed using SAS version 9.3 software (SAS Institute, Cary, North Carolina).
Pulse pressure appeared normally distributed in this sample of 40,421 patients, with 25th, 50th, and 75th percentile values of 46, 57, and 70 mm Hg, respectively, and of 58.1 ± 17.6 mm Hg. Distribution of demographic, medical historical, and clinical variables across PP quartiles exhibited a number of important differences (Online Table 1). Patients in the upper quartiles tended to be older, more likely female, and of non-white race. A history of diabetes, hypertension, and renal insufficiency was more frequent in the upper quartiles. Conversely, patients in the lower quartiles tended to have lower body mass index and lower EF. Admission and discharge SBP were also lower in these quartiles.
Association between PP and outcomes at 1 year in patients with HFrEF
As presented in Table 1 and Figure 1, for PP <50 mm Hg, PP was significantly associated with a more favorable outcome (adjusted hazard ratio [HR]: 0.946; 95% confidence interval [CI]: 0.900 to 0.995) or a 5.4% decrease in risk of all-cause mortality per 10-mm Hg increase in PP, up to a PP of 50 mm Hg. For PP ≥50 mm Hg, increased PP was significantly associated with an increased risk of death (adjusted HR: 1.091; 95% CI: 1.050 to 1.135) or a 9.1% increase in risk of all-cause mortality per 10-mm Hg increase in PP. Similar findings were observed for the association between PP and risk of death or rehospitalization at 1 year (Table 1). Of note, the reported HRs are adjusted for the use of evidence-based pharmacological therapy at the time of discharge and included beta-blockers, angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, as well as the patient’s pre-existing antihypertensive therapy.
Analysis of the association between PP and all-cause mortality and all-cause mortality/all-cause hospitalization in HFrEF, defined as EF < 0.50, did not substantially change the aforementioned conclusions (Table 2).
Analysis of the association between PP analyzed as a categorical variable and an outcome did not suggest significant relationships (Online Table 2), in contrast to the analysis of PP as a continuous variable. However, point estimates of the adjusted HRs were in the same “U” shape as the continuous results shown previously. There was no significant interaction between categorically defined PP and SBP, also in contrast to the analysis of PP as a continuous variable.
Association between PP and outcomes at 1 year in patients with HFpEF
There was no significant association between PP and risk of all-cause mortality at 1 year for increasing PP up to 50 mm Hg (Table 3, Online Figure 1). However, when PP was ≥50 mm Hg, there was a significant association between increasing PP and mortality and a significant interaction between SBP and PP (p for interaction = 0.0041). Higher PP was associated with progressively higher and significant increments in the risk of all-cause mortality at 1 year (4.0% to 12.1% increase in risk of death at 1 year with increase in SBP from 130 to 200 mm Hg) (Online Figure 2).
When HFpEF was defined as EF ≥0.50 and PP <50 mm Hg, the association of mortality with PP was not significant, but when PP was ≥50 mm Hg and SBP was >140 mm Hg, higher PP had an unfavorable association (Table 4).
Analysis of the association between PP, analyzed as a categorical variable, and all-cause mortality failed to identify a significant relationship (Online Table 2). There was no significant interaction between categorically defined PP and SBP, in contrast to the analysis of PP as a continuous variable.
For the combined outcome of death/rehospitalization (Online Table 3), there was a significant interaction between PP and SBP when PP was <50 mm Hg (p interaction = 0.0097) as well as when PP was ≥50 mm Hg (p interaction = 0.0001). After adjustment in patients with HFpEF, when PP was <50 mm Hg, higher PP was unfavorable (i.e., HR >1.0 when SBP was >120 mm Hg and became progressively more unfavorable at higher deciles of SBP).
When HFpEF was defined as EF ≥0.50 and when PP was <50 mm Hg, higher PP was unfavorable when SBP was ≥130 mm Hg. When PP was >50 mm Hg, higher PP was unfavorable when SBP was ≥140 mm Hg (Table 5).
Arterial PP at the time of hospital discharge in over 40,000 patients ≥65 years of age with a principal discharge diagnosis of HF was significantly associated with adverse outcomes at 1 year. The magnitude and direction of the association with all-cause mortality and all-cause mortality/readmission were generally similar for patients with discharge PP of ≥50 mm Hg in both HFrEF and HFpEF patients, although the magnitude of the effect in HFpEF patients varied with SBP. At a PP <50 mm Hg, there was an inverse association between PP and outcome in patients with HFrEF. These associations were consistently observed regardless of the cut point used to define HFpEF.
Pulse pressure has conventionally been viewed as a surrogate for arterial compliance (23,24). Although this concept has endured to the present time, PP is, in fact, a complex entity reflecting both ventricular and systemic arterial function and should take into account the hemodynamic circumstances under which it is determined (25). Under conditions where depressed LV systolic function is a dominant feature (e.g., HFrEF), PP is more likely to reflect diminished LV output and would decrease in magnitude, whereas under conditions where decreased arterial distensibility is a dominant feature and LV function is preserved (e.g., HFpEF), the PP would more likely reflect changes in arterial distensibility.
Other investigators have observed a nonlinear pattern to the association between outcome and PP in patients with HFrEF (11–16). Approximately 50% of the HFrEF patients in the present study had PP <50 mm Hg. We surmise that patients at this lower portion of the observed range of PP are at increased risk for adverse outcomes due to a decrease in LV output (26). At higher PPs (and reduced LV stroke output), the failing LV encounters decreased arterial compliance, increased ventricular afterload, and a further decrease in LV performance and likely adverse outcomes.
The dominant effect on outcomes of an increased PP in HFpEF reflects afterload excess (i.e., decreased arterial compliance) (27–29). In a recent systematic review of randomized trials and observational heart failure studies, no association was found between PP and outcome in HFpEF patients after adjusting for multiple potential confounders, although the unadjusted association was significant (16). HFpEF patients made up only 25% of the patient sample in that analysis, whereas in the current study, HFpEF patients represented more than 60% of the sample, thereby providing greater power to detect an effect if one truly existed. Additionally, PP <50 mm Hg was infrequent in HFpEF and was observed in less than 25% of HFpEF patients in our study. A more detailed comparison between the present analysis and the above-mentioned systematic review can be found in Table 6 and highlights differences in patient populations, underlying mortality risk and analytic methodology.
Although the stroke volume/brachial artery PP ratio has long been considered a surrogate for arterial compliance (23,24), this approximation is accurate only in patients with minimal amplification of the pressure pulse from the aorta to the periphery and must be used cautiously in subjects exhibiting significant pulse wave amplification. However, in elderly patients, the extent of PP amplification and, therefore, overestimation of the central aortic PP is likely to be minimal (30,31). The absence of stroke volume determination precludes our ability to estimate arterial compliance. Our underlying assumption regarding low PP as a surrogate for low stroke output is, however, consistent with conclusions reached in previous studies of similar patient phenotypes (11–16,26). It is also possible that low PP is associated with other measures of severe circulatory compromise or noncardiac measures of a poor prognosis for which we could not account. We note that the lower PP quartile was populated by patients with lower body weight and lower body mass index, lower SBP, lower EF, and higher rates of discharge to hospice facilities. These characteristics are frequently observed among frail patients (32) and may help to explain the “reverse epidemiology” of the association between low PP and outcome. Direct measures of arterial stiffness or aortic distensibility (33) may help to clarify the significance of low PP. Such measures were not performed in this registry, and conflicting data exist with respect to the prognostic value of measurements of aortic (34–36) or arterial (37,38) stiffness in patients with HF.
Limitations of a single PP assessment encompass the accuracy, validity, and reliability of the assessment. The recording of vital signs is a standardized function performed in hospitalized patients by trained individuals. Inaccuracies in blood pressure determination in any given patient or across patients in a sample as large as ours are unlikely to result in a systematic bias favoring over- or underestimation of the true blood pressure. Information bias of this sort, if it is present, would tend to mitigate the true association between PP and outcome and reduce the estimated effect toward the null. That we observed a statistically significant measure of effect argues against a significant degree of such misclassification. The magnitude of the association between PP and outcome is relatively small notwithstanding the level of statistical significance. However, numerous known predictors of mortality in younger populations are often attenuated in magnitude in older age groups (39). The higher prevalence of increased PP in elderly patients has important public health implications given the higher population attributable risk for this growing segment of the population (40). Finally, this was a retrospective analysis from a prospectively designed and conducted registry. No adjustments were made for multiple comparisons, and the results are considered hypothesis generating. Data were collected by chart review and are dependent on the quality and accuracy of data collection. Hospitals voluntarily participating in GWTG-HF may not be representative of all hospitals in the United States, although a previous study has shown that GWTG-HF patients and hospitals have characteristics similar to those of hospitals nationwide. We restricted the analysis to fee-for-service Medicare beneficiaries ≥65 years of age in order to allow for assessment of post-discharge outcomes by linkage of GWTG-HF records with those from Medicare. Most patients hospitalized with HF in the United States are ≥65 years of age. Characteristics and outcomes of Medicare beneficiaries in previous HF registries were similar to those in the broader Medicare population with HF (41).
In summary, arterial PP at the time of hospital discharge in patients with HF is associated with adverse outcomes at 1 year. A nonlinear association was observed in HFrEF patients such that, at PP < 50 mm Hg, risk increases with decreasing PP and that at PP ≥ 50 mm Hg, risk increases with increasing PP. In HFpEF patients, risk increases with increased PP, and the magnitude of the association is significantly impacted by SBP. The proper interpretation of PP in HF patients necessitates an understanding of the hemodynamic determinants of PP and the clinical characteristics of the patient.
COMPETENCY IN MEDICAL KNOWLEDGE: Arterial pulse pressure, a function of both ventricular output and arterial stiffness, in patients with heart failure is associated with all-cause mortality and the composite of all-cause mortality or readmission at 1 year in patients with either HFrEF or HFpEF. The magnitude of association is strongly affected by systolic blood pressure in HFpEF.
TRANSLATIONAL OUTLOOK: Further studies of the impact of targeted modification of pulse pressure or measures of aortic stiffness in patients with heart failure may provide deeper understanding of ventricular-arterial interactions in patients with heart failure.
Get With the Guidelines-Heart Failure, sponsored by the American Heart Association, has been supported by Medtronic, GlaxoSmithKline, Ortho-McNeill, and the American Heart Association Pharmaceutical Roundtable. Dr. Hernandez has received research support from Amgen, Bristol-Myers Squibb, GlaxoSmithKline, and Novartis and honoraria from Amgen, Bristol-Myers Squibb, Novartis, and Janssen. Dr. Bhatt is on the advisory boards of Cardax, Elsevier Practice Update Cardiology, Medscape Cardiology, Regado Biosciences; and on the board of directors of Boston Veterans’ Affairs Research Institute and Society of Cardiovascular Patient Care; serves as Chair, American Heart Association’s Get With the Guidelines-Heart Failure Steering Committee; is a member of the Data Monitoring Committees for Duke Clinical Research Institute, Harvard Clinical Research Institute, and Mayo Clinic Population Health Research Institute; has received honoraria from American College of Cardiology as Senior Associate Editor, Clinical Trials and News, Belvoir Publications as Editor in Chief, Harvard Heart Letter, and from Duke Clinical Research Institute, serving on clinical trial steering committees, Harvard Clinical Research Institute clinical trial steering committee, HMP Communications as Editor in Chief, Journal of Invasive Cardiology, Journal of the American College of Cardiology as Associate Editor and as Section Editor, Pharmacology, Population Health Research Institute clinical trial steering committee, Slack Publications as Chief Medical Editor, Cardiology Today’s Intervention, WebMD CME steering committees, and Deputy Editor of Clinical Cardiology; has received research funding from Amarin, AstraZeneca, Bristol-Myers Squibb, Eisai, Ethicon, Forest Laboratories, Ischemix, Medtronic, Pfizer, Roche, Sanofi, The Medicines Company; and support from FlowCo, PLx Pharma, and Takeda. Dr. Fonarow has received research support from Agency for Healthcare Research and Quality, and U.S. National Institutes of Health; and is a consultant for Amgen, Bayer, Gambro, Janssen, Novartis, and Medtronic. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- Centers for Medicare and Medicaid Services
- ejection fraction
- Get With The Guidelines
- heart failure
- left ventricle
- pulse pressure
- systolic blood pressure
- Received August 17, 2015.
- Revision received September 4, 2015.
- Accepted September 5, 2015.
- American College of Cardiology Foundation
- Franklin S.S.,
- Khan S.A.,
- Wong N.D.,
- Larson M.G.,
- Levy D.
- Domanski M.J.,
- Mitchell G.F.,
- Norman J.E.,
- Exner D.V.,
- Pitt B.,
- Pfeffer M.A.
- Mitchell G.F.,
- Moye L.A.,
- Braunwald E.,
- et al.
- Voors A.A.,
- Petrie C.J.,
- Petrie M.C.,
- et al.
- Jackson C.E.,
- Castagno D.,
- Maggioni A.P.,
- et al.,
- on behalf of the Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC)
- Remington J.W.,
- Nobach C.B.,
- Hamilton W.F.,
- Gold J.J.
- Chemla D.,
- Hebert J.-L.,
- Coirault C.,
- et al.
- Denardo S.J.,
- Nandyala R.,
- Freeman G.L.,
- Pierce G.L.,
- Nichols W.W.
- Chirinos J.A.,
- Kips J.G.,
- Jacobs D.R. Jr..,
- et al.
- Redheuil A.,
- Wu C.O.,
- Kachenoura N.,
- et al.
- O’Rourke M.F.,
- Blazek J.V.,
- Morreels C.L.,
- Krovetz L.T.
- Afilalo J.,
- Alexander K.P.,
- Mack M.J.,
- et al.
- Laurent S.,
- Cockcroft J.,
- Van Bortel L.,
- et al.,
- for the European Network for Non-invasive Investigation of Large Arteries
- Pepine C.J.,
- Nichols W.W.,
- Conti C.R.
- Laskey W.K.,
- Kussmaul W.G.,
- Martin J.L.,
- Kleaveland J.P.,
- Hirshfeld J.W. Jr..,
- Shroff S.
- Laskey W.K.,
- Parker H.G.,
- Ferrari V.A.,
- Kussmaul W.G.,
- Noordergraff A.
- Chirinos J.A.,
- Khan A.,
- Bansal N.,
- et al.,
- for the CRIC Study Investigators
- Wilson P.W.,
- D’Agostino R.B.,
- Levy D.,
- Belanger A.M.,
- Silbershatz H.,
- Kannel W.B.