Cardiac and Non-Cardiac Dysfunction as Precursors of Heart Failure with Reduced and Preserved Ejection Fraction in the Community

Abstract

Background

Heart failure (HF) is a clinical syndrome characterized by signs and symptoms involving multiple organ systems. Longitudinal data demonstrating that asymptomatic cardiac dysfunction precedes overt HF are scarce, and the contribution of non-cardiac dysfunction to HF progression is unclear. We hypothesized that subclinical cardiac and non-cardiac organ dysfunction would accelerate the manifestation of HF.

Methods and Results

We studied 1038 participants of the Framingham Heart Study original cohort (mean age 76±5 years; 39% men) with routine assessment of left ventricular (LV) systolic and diastolic function. Major non-cardiac organ systems were assessed using serum creatinine (renal), serum albumin (hepatic), ratio of forced expiratory volume in 1 second (FEV1) to forced vital capacity (FVC) (pulmonary), hemoglobin concentration (hematologic/oxygen carrying capacity) and white blood cell count (systemic inflammation). On follow-up (mean 11 years), there were 248 incident HF events (146 in women). Adjusting for established HF risk factors, antecedent LV systolic (hazards ratio 2.33; 95%CI, 1.43–3.78) and diastolic (hazards ratio 1.32; 95%CI, 1.01-1.71) dysfunction were associated with increased HF risk. Adjusting for cardiac dysfunction, higher serum creatinine, lower FEV1:FVC ratios and lower hemoglobin concentrations were associated with increased HF risk (all P<0.05); serum albumin and white blood cell count were not. Subclinical dysfunction in each non-cardiac organ system was associated with a 30% increased risk of HF (P=0.013).

Conclusions

Antecedent cardiac and non-cardiac organ dysfunction are associated with increased incidence of HF, supporting the notion that HF is a progressive syndrome and underscoring the importance of non-cardiac factors in its occurrence.

INTRODUCTION

Heart failure (HF) is a clinical syndrome characterized by a constellation of signs and symptoms involving multiple organ systems such as the heart (classic ‘pump failure’), the lungs (dyspnea), the kidneys (salt and water retention) and the liver (congestion). Current HF guidelines^{1, 2} emphasize the importance of asymptomatic cardiac dysfunction as a preceding stage in the progression to clinically overt HF. Cross-sectional studies have demonstrated the presence of asymptomatic systolic or diastolic LV dysfunction in the community in individuals at risk of, but without HF,³ and an even higher prevalence of these abnormalities in patients with overt HF.⁴ However, to demonstrate a prospective association between these structural precursors and future HF, longitudinal studies are needed.

Importantly, the current HF staging system¹ does not specifically acknowledge the potential association of non-cardiac dysfunction with the occurrence of HF. Since the syndrome of HF involves multiple organ systems, even mild functional derangement of a non-cardiac organ system may accelerate the manifestation of overt HF, particularly when other organ systems are also involved. Indeed, emerging evidence suggests that subclinical renal impairment,⁵ hypoalbuminemia,^{6, 7} decline in pulmonary function,^{8, 9} anemia¹⁰ and systemic inflammation¹¹ may all contribute to HF progression. Of note, the prevalence of non-cardiac co-morbidities is high among patients with overt HF, and these co-morbidities are major determinants of mortality after the onset of HF.^{12, 13}

The relations of antecedent cardiac and non-cardiac dysfunction (i.e. present before onset of overt HF) to the incidence of HF have not been studied comprehensively in the community. In a previous report,¹⁴ we described the prevalence and prognosis of asymptomatic LV systolic dysfunction in the community but that investigation did not examine the association of LV diastolic dysfunction or of non-cardiac major organ system dysfunction with the risk of HF. Accordingly, we aimed to prospectively determine the association of cardiac and non-cardiac dysfunction with the incidence of HF among older adults without HF in the community. To achieve this, we harnessed the unique availability of longitudinal data and routine surveillance in the Framingham Heart Study. We hypothesized that subclinical dysfunction in both cardiac and non-cardiac organ systems would accelerate the manifestation of HF. Further recognizing potential mechanistic differences between HF with reduced ejection fraction (HFREF) versus HF with preserved ejection fraction (HFPEF), we also hypothesized that the types of antecedent subclinical organ system dysfunction may differ according to the type of incident HF (HFREF versus HFPEF).

METHODS

Participants

The Framingham Heart Study is a longitudinal community-based cohort study that began in 1948.¹⁵ The original cohort has been under continuous surveillance and participants are examined at the Heart Study clinic approximately every 2 years. In the present investigation, we included participants attending the 20^th biennial examination with routine assessment by Doppler echocardiography, but without prevalent HF (Supplementary Figure 1). Since our focus was on subclinical organ dysfunction, we excluded participants with overt organ dysfunction such as those with overt renal failure (defined as a serum creatinine>2 mg/dL [176.8 μmol/l]; N=12). All participants provided written informed consent and the study protocol was approved by the Institutional Review Board of the Boston University Medical Center.

Definition of cardiac dysfunction

Established HF risk factors modeled as covariates included age, sex, body mass index, systolic blood pressure, hypertension treatment, cholesterol, diabetes mellitus, prior myocardial infarction, and valvular heart disease. LV systolic dysfunction was assessed by echocardiographic ejection fraction estimated visually.¹⁴ LV diastolic dysfunction was defined on the basis of LV filling pattern as any abnormal relaxation, pseudonormal filling or restrictive filling. Abnormal relaxation (mitral E/A<0.5, deceleration time>280 ms) or restrictive filling (mitral E/A>2.0, deceleration time<120 ms) was classified based on mitral inflow patterns.¹⁶ In the absence of tissue Doppler imaging, pseudonormal LV filling was distinguished from normal LV diastolic function by the presence of any of the following: left atrial size ≥ sex-specific 80^th percentile, LV mass ≥ sex-specific 80^th percentile, or any atrial fibrillation. These criteria closely parallel recommendations from the European Society of Cardiology (ESC) for the diagnosis of HFPEF,¹⁷ and use the upper sex-specific quintiles of left atrial size and LV mass to characterize atrial enlargement and LV hypertrophy,¹⁸ respectively, in our elderly cohort. Both LV systolic and diastolic dysfunction were modeled as binary variables (presence versus absence).

Definition of non-cardiac major organ system dysfunction

We evaluated non-cardiac major organ systems that could accelerate the manifestations of HF (dyspnea, fluid retention/pedal edema, and exertional fatigue). Participants underwent routine spirometry and phlebotomy. Measurement variables used to define non-cardiac function included:^{5-7, 10, 11}

1. Renal system: serum creatinine.

2. Hepatic system: serum albumin concentration.

3. Pulmonary system: ratio of forced expiratory volume in 1 second (FEV1) to forced vital capacity (FVC) expressed as % predicted for age and sex.

4. Hematologic system/oxygen carrying capacity: hemoglobin concentration.

5. Systemic inflammation: white blood cell count.

These non-cardiac function variables were modeled as both continuous and as binary variables (see below). The sample size for analyses of non-cardiac dysfunction variables was smaller (n=676) than the overall sample with available echocardiographic measures of LV systolic and diastolic function (n=1038; Supplementary Figure 1).

Outcome

Participants have been under ongoing, routine surveillance for incident HF since the baseline examination in 1987-1990. HF was defined as satisfying the previously published Framingham criteria¹⁹ (presence of 2 major, or of one major plus two minor criteria) and adjudicated by a panel of three experienced investigators. The ejection fraction closest to the date of the HF event was used to categorize HF into HFPEF (EF>45%) or HFREF (EF≤45%).²⁰ Measurements of ejection fraction performed at HF onset, during HF hospitalization, or within 1 year of HF onset in the absence of intervening myocardial infarction were eligible.²⁰

Statistical analyses

We used Cox proportional hazards regression models to assess the relationship of cardiac and non-cardiac dysfunction variables to the incidence of HF after confirming that the assumption of proportionality of hazards was met. Covariates eligible for the multivariable model included LV systolic and diastolic dysfunction, continuous measures of non-cardiac function (defined above), as well as established HF risk factors (defined above).

For each non-cardiac function variable (including those found to be not significant), we initially examined generalized additive models with penalized splines to assess potential non-linearity of the association.²¹ None of the associations were found to be non-linear. Therefore, we proceeded to model linear associations in Cox models. In the absence of any non-linearity of the associations, we also used a priori cut-points based on the lower 25^th or upper 75^th percentile of each continuous variable in order to create binary variables defining organ “dysfunction” for incorporation into a risk score. The risk score for non-cardiac dysfunction was then calculated for each participant by allocating 1 point for each affected non-cardiac organ system that was significantly associated with the risk of HF. This scoring system approach assumes that the hazards posed by dysfunction in each of the non-cardiac systems are similar (weighted the same), and offers a simple, practical score that may be meaningfully applied in clinical settings. The association of the non-cardiac risk score with incident HF was then plotted using Kaplan-Meier curves, and assessed using Cox proportional hazards modeling adjusting for established HF risk factors (noted above) and cardiac dysfunction variables.

Finally, analyses were repeated separately for incident HFREF as the outcome (and censoring cases of HFPEF at the time of that event), or incident HFPEF as the outcome (and censoring cases of HFREF at the time of that event). Cox proportional hazards regression was used, in which variables entered into the model included LV systolic and diastolic dysfunction, measures of non-cardiac function, as well as established HF risk factors noted above. Valvular heart disease was excluded in the analysis for HFPEF consistent with current diagnostic criteria,¹⁷ and included as a covariate in the analysis for HFREF.

All analyses were performed using SAS and a two-sided P value of <0.05 was used to indicate statistical significance. All authors had full access to the data and take responsibility for the integrity of the data.

RESULTS

Baseline characteristics

The study sample consisted of 1038 elderly participants (Table 1). More than three-quarters of the sample was hypertensive and about half was on antihypertensive treatment. The prevalence of asymptomatic LV systolic and diastolic dysfunction were 5% and 36%, respectively, consistent with other community-based studies.³ The distributions of measures of non-cardiac function were within the ranges expected for elderly individuals in the general population.⁵

Table 1.

Baseline characteristics

Characteristics
N	1038
Age, yrs	76±5
Men, no.,(%)	409 (39)
Cardiovascular risk factors
Body mass index, kg/m2	26.6±4.5
Systolic blood pressure, mmHg	147±22
Hypertension, no., (%)	799 (77)
Hypertension treatment, no., (%)	551 (53)
Total cholesterol:HDL ratio	4.85±1.65
Diabetes mellitus, no., (%)	104 (10)
Myocardial infarction, no, (%)	96 (9)
Valve disease, no., (%)	41 (4)
Cardiac function
LV systolic dysfunction, no., (%)	57 (5)
LV diastolic dysfunction, *no., (%)	372 (36)
Non-cardiac function
Serum creatinine, mg/dl	0.93±0.23
Serum albumin, g/dl	4.23±0.36
FEV1:FVC ratio, % predicted	96.9±10.9
Hemoglobin concentration, g/dl	14.0±1.5
White blood cell count, ×109/l	6.8±2.3

Values are mean±SD unless otherwise stated. Serum creatinine concentration can be converted to μmol/l by multiplying the value in mg/dl by 88.4.

HDL, high-density lipoprotein; LV, left ventricular; EF, ejection fraction; FEV1:FVC, ratio of forced expiratory volume in 1 second to forced vital capacity expressed as % predicted for age and sex

^*Diastolic dysfunction included any abnormal relaxation (mitral E/A<0.5, deceleration time>280 ms); pseudonormal LV filling (distinguished from normal LV diastolic function by the presence of any of the following: left atrial size ≥ sex-specific 80^th percentile [4.8 cm in men, 4.4 cm in women], LV mass ≥ sex-specific 80^th percentile [158.4 g/m² in men, 141.9 g/m² in women], or any atrial fibrillation); or restrictive filling (mitral E/A>2.0, deceleration time<120 ms).

Cardiac dysfunction as a risk factor for incident HF

Over a mean follow-up of 11 years, there were 248 incident first HF events (146 in women; 119 HFREF, 101 HFPEF, EF was not available for 28 HF events). In multivariable models adjusting for established HF risk factors (age, sex, body mass index, systolic blood pressure, hypertension treatment, cholesterol, diabetes mellitus, prior myocardial infarction, and valvular heart disease), both LV systolic dysfunction and LV diastolic dysfunction were associated with increased risk of incident HF (Table 2).

Table 2.

Cardiac Dysfunction as a risk factor for incident heart failure

Characteristic	Hazards ratio (95%CI)	P value*
LV systolic dysfunction	2.33 (1.43–3.78)	<0.001
LV diastolic dysfunction	1.32 (1.01-1.71)	0.039

LV, left ventricular

^*Adjusted for age, sex, body mass index, systolic blood pressure, hypertension treatment, cholesterol, diabetes mellitus, prior myocardial infarction, and valvular heart disease in 1038 participants (248 HF events)

Non-cardiac risk factors and major organ system dysfunction risk score for incident HF

Participants (n=676; mean age 75±5 years; 42% men) without missing non-cardiac risk factor variables had similar baseline characteristics (body mass index, systolic blood pressure, diabetes) compared to those with missing variables (all P>0.05). Adjusting for established HF risk factors and presence of cardiac systolic and diastolic dysfunction, higher serum creatinine, lower FEV1:FVC ratios and lower hemoglobin concentrations were associated with increased risk of new-onset HF (Table 3). There was no association between serum albumin concentration (P=0.306) or white blood cell count (P=0.685) and incident HF. Individual Cox proportional hazards models with penalized splines for serum creatinine, FEV1:FVC ratio and hemoglobin did not reveal non-linearity for the association with HF risk (Figure 1). A risk score for non-cardiac dysfunction was calculated, therefore, using pre-determined cut-points (based on 25^th or 75^th percentiles of the variables in the sample, as defined above), and awarding 1 point for each affected organ system (range 0-3); regression coefficients for these variables were comparable in the multivariable models, further justifying their similar weighting in the score (Table 3). Of note, the cut-points used to define organ “dysfunction” were within the ranges frequently observed in ambulatory elderly individuals from the general population.⁵ Increasing non-cardiac risk score at baseline was positively associated with risk of HF (Figure 2).

Table 3.

Non-cardiac risk factors and risk score for incident heart failure

Characteristic	Hazards ratio (95% CI)*	P value*	Cut-off percentile	Cut-off value	Points awarded
Serum creatinine	1.21 (1.01–1.45)	0.036	>75th percentile	> 1.05 mg/dl (>92.8 μmol/l)	1
FEV1:FVC ratio	1.21 (1.02–1.43)	0.029	<25th percentile	< 91 % predicted	1
Hemoglobin concentration	1.24 (1.09–1.40)	<0.001	<25th percentile	< 13 g/dl	1

FEV1:FVC, ratio of forced expiratory volume in 1 second to forced vital capacity

^*Hazards ratio are for 1SD increase in serum creatinine, 1SD decrease in FEC1:FVC ratio and 1 unit decrease in hemoglobin concentration, adjusting for age, sex, body mass index, systolic blood pressure, hypertension treatment, cholesterol, diabetes mellitus, prior myocardial infarction, valvular heart disease, and left ventricular systolic and diastolic function in 676 participants without any missing variables (170 HF events)

Figure 1. Association of measures of major non-cardiac organ system function with risk of incident heart failure.

Generalized additive models with penalized splines were used to assess the association of multivariable-adjusted hazards ratio for heart failure with (A) serum creatinine concentration, (B) ratio of forced expiratory volume in 1 second to forced vital capacity (FEV1:FVC ratio) and (C) hemoglobin concentration. Lines indicate means (solid) and 95% confidence intervals (dotted). *Y-axes represent multivariable-adjusted Ln (hazards ratio). To obtain serum creatinine concentration in μmol/l, multiply values in mg/dl by 88.4.

Figure 2. Cumulative incidence of incident heart failure according to non-cardiac major organ system dysfunction risk score.

The non-cardiac organ system dysfunction risk score awarded 1 point for the presence of each of the following (range 0-3): serum creatinine >1.05 mg/dl (92.8 μmol/l), ratio of forced expiratory volume in 1 second to forced expiratory volume <91% predicted, hemoglobin concentration <13 g/dl. Increasing non-cardiac risk score at baseline was associated with increasing risk of incident heart failure (HF) in our community-based sample (log rank P value = 0.013).

In secondary analyses, results were similar when estimated glomerular filtration rate (eGFR by MDRD equation) was used instead of serum creatinine (hazards ratio for each 1SD decrease in eGFR was 1.24 (95% CI, 1.03-1.50; P=0.026) adjusting for established HF risk factors and presence of cardiac dysfunction). Other biomarkers of non-cardiac dysfunction (blood urea nitrogen, total bilirubin, transaminases, hematocrit, C-reactive protein measured by traditional assays [high sensitivity assays unavailable], uric acid) were also tested for their associations with incident HF in secondary analyses, and results are shown in Supplementary Table 1. These secondary analyses supported the original selection of creatinine, albumin, FEV1:FVC ratio and hemoglobin as simple, convenient and widely available biomarkers to include in the final risk score. We also performed sensitivity analyses using difference percentile cutpoints to define the risk score (tertiles instead of quartiles) and found similar results (data not shown).

Multivariable risk factors for all incident HF, HFREF and HFPEF

In multivariable modeling for all incident HF, LV systolic dysfunction, LV diastolic dysfunction and the non-cardiac risk score were each associated with incident HF, adjusting for established HF risk factors (Table 4). There was no significant interaction between non-cardiac risk score and the presence of LV systolic or diastolic dysfunction (P=0.78 and 0.84 respectively). In multivariable modeling for HFREF and HFPEF separately, antecedent LV systolic dysfunction, greater serum creatinine and lower hemoglobin concentration were associated with incident HFREF; whereas antecedent LV diastolic dysfunction and lower FEV1:FVC ratio were associated with incident HFPEF, adjusting for established HF risk factors (Table 4).

Table 4.

Association of cardiac and non-cardiac dysfunction with incident heart failure

Characteristics	Hazards ratio (95%CI)*	P value*
All incident HF
LV systolic dysfunction	1.97 (1.05–3.68)	0.034
LV diastolic dysfunction	1.40 (1.02-1.93)	0.039
Non-cardiac risk score† per 1 unit increase	1.30 (1.06-1.60)	0.013
Incident HFREF
LV systolic dysfunction	3.93 (1.86 – 8.30)	<0.001
Serum creatinine per 1SD increase	1.32 (1.04 – 1.69)	0.025
Hemoglobin concentration per 1unit decrease	1.31 (1.10 – 1.55)	0.002
Incident HFPEF
LV diastolic dysfunction	1.88 (1.13 – 3.13)	0.016
FEV1:FVC ratio per 1SD decrease	1.38 (1.04 – 1.83)	0.024

HF, heart failure; LV, left ventricular; HFREF, heart failure with reduced ejection fraction; HFPEF, heart failure with preserved ejection fraction; FEV1:FVC, ratio of forced expiratory volume in 1 second to forced vital capacity

^*Adjusted for age, sex, body mass index, systolic blood pressure, hypertension treatment, cholesterol, diabetes mellitus, prior myocardial infarction, and valvular heart disease (valvular heart disease excluded for HFPEF) in 676 participants without any missing variables (170 total HF events: 82 HFREF, 66 HFPEF, 22 EF unavailable).

^†Components of the non-cardiac risk score are described and their individual hazards ratios are shown in Table 3.

In further analyses adjusting for smoking and excluding participants with overt pulmonary dysfunction (FEV1/FVC < 60%; N=3), hypoalbuminemia (albumin < 2.5 g/dL; N=0, no participant had hypoalbuminemia) or anemia (hemoglobin < 10.5g/dL; N=7), results were essentially unchanged (data not shown). Using an EF cutpoint of 50% (instead of 45%) to distinguish HFPEF from HFREF revealed similar results (Supplementary Table 2).

DISCUSSION

Principal findings

In our prospective study of a large, community-based sample, antecedent subclinical cardiac and non-cardiac major organ system dysfunction were associated with risk of future HF. The presence of asymptomatic LV systolic and diastolic dysfunction preceded and increased the risk of incident HF by more than two-fold and more than 30%, respectively. These findings support the emphasis in current HF guidelines regarding the progressive nature of HF and importance of recognizing preceding asymptomatic cardiac dysfunction. Our data also extend the previous HF staging system by providing evidence for the association of non-cardiac dysfunction with progression to clinical HF. Adjusting for cardiac dysfunction, the presence of subclinical renal impairment, airflow limitation or anemia was each associated with 30% increased risk of incident HF. Finally, antecedent LV systolic dysfunction was associated with future HFREF, whereas antecedent LV diastolic dysfunction was associated with future HFPEF. Further, subclinical renal impairment and lower hemoglobin concentrations were associated with a higher incidence of HFREF, whereas baseline airflow obstruction was related positively to the risk of future HFPEF. The implications of these findings for the early identification of individuals at risk of HF, and potential strategies to prevent progression to overt HF, deserve further study.

Left ventricular systolic and diastolic dysfunction and risk of HF

Previous cross-sectional studies have provided evidence for the existence of asymptomatic LV dysfunction in the general community (Stage B HF in the American College of Cardiology/American Heart Association classification system),^{3, 22} as well as increased prevalence and severity of LV dysfunction in patients with clinical HF (Stage C HF).^{3, 4} While these cross-sectional data supported the proposed stages of the HF, prior studies were limited by potential reverse causality, since assessment of LV function was performed at the same point in time as the diagnosis of clinical HF. Further, cross-sectional studies may be criticized for scientific circularity of reasoning in that the presence of LV systolic or diastolic dysfunction is used to make the diagnosis HFREF or HFPEF, respectively. Prospective data are needed to resolve these issues. In the Cardiovascular Health Study,²³ LV systolic and diastolic dysfunction predicted incident HF over a mean follow-up of 5.2 years. However, the relations between the type of LV dysfunction (systolic versus diastolic) and the type of HF (HFREF versus HFPEF) were not assessed. More recently, researchers from the Mayo Clinic reported a 2-year HF incidence rate of 1.9% in a selected sample of 82 patients with preclinical diastolic dysfunction.²⁴ Patients with systolic dysfunction were not studied. Our current data are consistent with these prior data and extend previous knowledge by demonstrating that LV systolic dysfunction predicts future HFREF, whereas LV diastolic dysfunction portends HFPEF. The current data, therefore, help to fill the knowledge gap linking Stage B to Stage C in the American College of Cardiology/American Heart Association classification scheme, whether referring to HFREF or HFPEF.

Non-cardiac dysfunction and risk of HF

HF is a clinical syndrome characterized by a constellation of signs and symptoms involving multiple organ systems besides the heart. Thus, even mild functional derangement of a non-cardiac organ system, which in itself is not severe enough to produce symptoms, may accelerate the manifestation of overt HF particularly when other organ systems are also involved. A decline in renal function affects sodium handling and fluid homeostasis, thus increasing the propensity to manifest fluid overload.⁵ Pulmonary function has a direct impact on the manifestation of dyspnea. Subclinical chronic pulmonary disease is characterized by low-grade inflammation and may contribute to progression of atherosclerosis and myocardial dysfunction,⁸ while even mild airflow obstruction is associated with abnormal LV diastolic filling.⁹ Anemia affects the oxygen-carrying capacity of the blood and is an adverse marker in overt HF.¹⁰ The availability of systematic, multi-system measurements during routine surveillance in the Framingham Heart Study enabled comprehensive assessment of these varied non-cardiac organ systems in relation to incident HF in the community. Our findings regarding the association with renal impairment are consistent with the Health ABC Heart Failure Model for incident HF in the elderly.²⁵ In contrast, we did not find a significant association with hypoalbuminemia, and this may be due to differences in study samples (larger proportion of blacks and lower baseline serum albumin in the Health ABC Study).

In aggregate, these results suggest that the manifestation of clinically overt HF may be hastened by subclinical dysfunction in multiple organ systems. This is likely to particularly affect elderly individuals who have age-related decline in multi-organ function or multiple non-cardiac co-morbidities. Recognizing the contribution of non-cardiac dysfunction to HF progression may carry important clinical implications for preventing and managing heart failure. Further studies are warranted to validate these findings in other populations, evaluate for potential effect modification by covariates such as sex, and assess the potential impact of treatment of these risk factors on the risk of future HF.

Association of non-cardiac dysfunction with HFREF versus HFPEF

The distinction between factors associated with incident HFREF versus HFPEF deserves comment. The association of renal dysfunction and anemia with the risk of HFREF are consistent with classic studies of the cardiorenal syndrome and the known prognostic impact of anemia in overt HFREF.¹⁰ Interestingly, the most prominent non-cardiac predictor of incident HFPEF was airflow obstruction. This observation is supported by large epidemiologic studies showing a high prevalence of pulmonary disease in patients with HFPEF,^{26, 27} the frequent co-existence of HF in patients with chronic obstructive lung disease,²⁸ as well as a recent study in the Multi-Ethnic Study of Atherosclerosis (MESA) cohort demonstrating an association between airflow obstruction and abnormal LV filling.⁹ While this also raises the question of potential misdiagnosis of HFPEF, this is unlikely given the high specificity of the Framingham criteria for HF,²⁹ the reliability of the diagnosis as demonstrated by consistent application of the same criteria over decades and stringent adjudication of endpoints in the Heart Study, and the lack of an alternative explanation for the clinical presentation (since the extent of pulmonary impairment was not severe enough to explain symptoms). Overall, the different predictors of HFREF versus HFPEF are consistent with prior epidemiologic,²⁰ pathophysiologic,³⁰ molecular,³¹ and outcome³² data supporting the notion that HFREF and HFPEF may represent separate entities. These observations may guide future clinical trial design, particularly in HFPEF where trials have, so far, been disappointing.

Strengths and limitations

The strengths of our study include the large community-based sample, uniform measurements of function of multiple organ systems, and longitudinal follow-up with continuous surveillance for, and careful validation of, HF outcomes. Further, the use of purely clinical criteria for the diagnosis of HF¹⁹ independent of LVEF is particularly advantageous in this setting. Limitations include the lack of tissue Doppler characterization of diastolic dysfunction, and the inherent pitfalls in using echocardiographic indices of diastolic filling as indicators of diastolic dysfunction. Nonetheless, mitral Doppler indices are widely available; and our results are consistent with previous studies using more comprehensive assessment of diastolic dysfunction.^{3, 24} We acknowledge the observational nature of our study which precludes conclusions regarding causality, as well as the modest effect sizes for non-cardiac organ system variables and potential for residual confounding or effect modification by other factors. We did not use time-varying covariates, but speculate that organ dysfunction would worsen with aging in most individuals, leading to even stronger associations with incident HF. Our uniformly white sample limits the generalizability of our findings to other ethnicities, and independent validation in other cohorts is warranted.

Conclusions

Our prospective observations in a large, community-based sample demonstrate that antecedent subclinical cardiac and non-cardiac major organ system dysfunction are associated with increased risk of clinical HF. These findings contribute to the understanding of HF as a progressive disease syndrome, and underscore the potential importance of non-cardiac risk factors in predisposing to the manifestation of overt HF.

CLINICAL COMMENTARY.

In our prospective study of a large, community-based sample, antecedent subclinical cardiac and non-cardiac major organ system dysfunction were associated with risk of future heart failure. The presence of asymptomatic left ventricular systolic and diastolic dysfunction preceded and increased the risk of incident heart failure by more than two-fold and more than 30%, respectively. Adjusting for cardiac dysfunction, the presence of subclinical renal impairment, airflow limitation or anemia was each associated with 30% increased risk of incident heart failure. Notably, the significant risk factors differed according to the type of incident heart failure (preserved versus reduced ejection fraction). Antecedent left ventricular systolic dysfunction, subclinical renal impairment and lower hemoglobin concentrations were associated with a higher incidence of heart failure with reduced ejection fraction, whereas antecedent diastolic dysfunction and baseline airflow obstruction were related positively to the risk of future heart failure with preserved ejection fraction. This study provides longitudinal evidence for the progressive nature of heart failure as emphasized in current heart failure guidelines, and underscores the potential importance of non-cardiac risk factors in predisposing to the manifestation of overt heart failure. The implications for the early identification of individuals at risk of heart failure and potential strategies to prevent progression to overt heart failure deserve further study.