Pulmonary Capillary Wedge Pressure Patterns During Exercise Predict Exercise Capacity and Incident Heart Failure

Abstract

Background

Single measurements of left ventricular filling pressure at rest lack sensitivity for identifying heart failure with preserved ejection fraction (HFpEF) in patients with dyspnea on exertion. We hypothesized that exercise hemodynamic measurements (i.e. changes in pulmonary capillary wedge pressure, PCWP, indexed to cardiac output, CO) may more sensitively differentiate HFpEF and non-HFpEF disease states, reflect aerobic capacity, and forecast HF outcomes in individuals with normal PCWP at rest.

Methods and Results

We studied 175 patients referred for cardiopulmonary exercise testing (CPET) with hemodynamic monitoring: controls (N=33), HFpEF with resting PCWP≥15 mmHg (N=32), and patients with dyspnea on exertion with normal resting PCWP and LVEF [DOE-nlrW] (N=110). Across 1,835 paired PCWP-CO measurements throughout exercise, we used regression techniques to define normative bounds of “PCWP/CO slope” in controls, and tested the association of PCWP/CO slope with exercise capacity and composite cardiac outcomes (defined as cardiac death, incident resting PCWP elevation or HF hospitalization) in the DOE-nlrW group. Relative to controls (PCWP/CO slope 1.2±0.4 mmHg/L/min), patients with HFpEF had a PCWP/CO slope 3.6±1.9 mmHg/L/min. We used a threshold (2 standard deviations above the mean in controls) of 2 mmHg/L/min to define “abnormal.” PCWP/CO slope >2 in DOE-nlrW patients was common (N= 45/110) and was associated with reduced peak VO₂ (p<0.001) and adverse cardiac outcomes after adjustment for age, sex, and BMI (hazard ratio HR 3.47, p = 0.03) at a median 5.3 year follow-up.

Conclusions

Elevated PCWP/CO slope during exercise (>2mmHg/L/min) is common in DOE-nlrW and predicts exercise capacity and HF outcomes. These findings suggest that current definitions of HFpEF based on single measures during rest are insufficient and that assessment of exercise PCWP/CO slope may refine early HFpEF diagnosis.

Resting cardiac filling pressures are often normal in patients with heart failure with preserved ejection fraction (HFpEF)¹, despite exertional limitations that are a cardinal manifestation of this condition.² Both filling pressures and cardiac output (CO) are dynamic in patients with suspected and established HFpEF,³ suggesting that single time point hemodynamic measurements may not fully capture the underlying physiology of this disease.

Exercise has been promoted as an attractive physiologic probe to unmask early disease phenotypes in HFpEF. In HFpEF, exercise pulmonary arterial pressures (PAP) and pulmonary capillary wedge pressures (PCWP; a surrogate for left atrial pressure) exceed those in hypertensive controls.¹ Consequently, PCWP during exercise can aid in HFpEF diagnosis⁴ and has been used to define inclusion criteria for clinical trials in HFpEF.^{5,6, 7} However, published reports have advocated different single-value thresholds for abnormal elevation in exercise PCWP (i.e. 15mmHg⁸, 20mmHg⁹, or 25mmHg¹), and precise data on normative¹⁰ and abnormal PCWP responses to exercise are limited. Single exercise hemodynamic measures to define other cardiovascular disease states (e.g., mean PAP ≥ 30 mmHg for pulmonary hypertension) have been appropriately abandoned¹¹ based on growing recognition of the importance of accounting for the amount of exercise and defining the physiologic and clinical significance of exercise measurements.¹² Moreover, optimal ways to assess PCWP during exercise (i.e. frequency of PCWP measurements, approach to zero leveling during supine vs. semi-supine vs. upright exercise) and to index PCWP to the “dose” of exercise remain undefined.

To address these knowledge gaps, we sought to define PCWP responses to exercise and the association of exercise PCWP patterns with peak aerobic capacity and progression to HF. We performed comprehensive cardiopulmonary exercise testing (CPET) with simultaneous quantification of gas exchange and hemodynamic measurements during maximal upright cycle ergometry in 175 subjects with subsequent follow up to ascertain future HF status. By analogy to ongoing efforts with PAP response to exercise in pulmonary vascular disease states,¹² we hypothesized that PCWP changes in relation to CO (the “PCWP/CO slope”) in dyspneic individuals reflect abnormal fitness and higher rates of progression to overt HF, independent of resting hemodynamics.

METHODS

The analytic methods will be made available to other researchers via email to the corresponding author for purposes of reproducing the results but patient data will not be made available because of ongoing separate analyses.

Study sample and study design

We defined three separate samples in order to compare PCWP responses to exercise among groups: (1) a “control” group: 33 individuals without any abnormalities on CPET (defined by >85% age/sex-predicted peak VO₂, with normal PAP, PCWP, and CO at rest, and normal biventricular function by radionuclide ventriculography); (2) an “overt” HFpEF group: 32 individuals with hemodynamic criteria for HFpEF (symptoms of HF adjudicated by a physician, with left ventricular [LV] ejection fraction >0.50 by echocardiography or radionuclide ventriculography and a mean PCWP ≥ 15 mmHg at rest); (3) a DOE-nlrW sample (dyspnea on exertion with normal resting PCWP): 110 consecutive patients clinically referred to our center for CPET with hemodynamic monitoring for evaluation of exertional dyspnea. From this group, we excluded patients with elevated resting PCWP (resting supine PCWP ≥15 mmHg), those with LV systolic dysfunction (LV ejection fraction <0.45 on echocardiography or radionuclide ventriculography), or elevated resting PAP (≥25mmHg). This study was approved by the Partners Institutional Review Board and all subjects provided informed consent.

Cardiopulmonary exercise testing (CPET)

All subjects underwent placement of a pulmonary arterial and a radial arterial catheter to monitor right heart hemodynamics (Witt Biomedical, Melbourne, FL), systemic blood pressure, and arterial blood gas sampling during incremental exercise. First-pass radionuclide imaging was performed at rest and peak exercise to quantify left ventricular ejection fraction. CPET was performed on an upright cycle ergometer, and consisted of a ≥3-minute period of resting measurements followed by 3-minutes of unloaded exercise. Subsequently, patients engaged in a continuous incremental ramp cycle ergometry protocol (5–30 watts/min, based on estimated exercise capacity) designed to yield 8–12 minutes of total exercise duration at a constant cadence (60 revolutions per minute). Gas exchange was assessed via breath-by-breath measurements (Medgraphics, St. Paul, MN). A respiratory exchange ratio (VCO₂/VO₂) greater than 1.0 or a maximum heart rate greater than 85% age-predicted maximal heart rate was used to define an adequate effort. Peak VO₂ was defined as the highest median (over 30 second intervals) during the last minute of exercise.

In addition to gas exchange indices, right atrial pressure (RAP), pulmonary arterial systolic, diastolic and mean PAP and mean systemic arterial pressure (MAP) were measured at rest and continuously during exercise, with values recorded each minute. PCWP was obtained by PA catheter distal balloon inflation. CO was calculated at 1-minute intervals using the Fick principle with a measured VO₂, hemoglobin, and simultaneous radial and pulmonary arterial (mixed venous) oxygen content, as previously described.¹³ To ensure a standardized workload to compare across patients, we determined a “30-watt PCWP” using linear interpolation between two points if a patient completed an incremental ramp that did not include exactly 30 Watts (i.e. a 20W/min ramp, we averaged values from 20W and 40W).

Statistical analysis

Baseline demographic, clinical, and echocardiographic parameters were compared between individuals across all cohorts using Fisher’s exact test (for categorical) or ANOVA with Bonferroni-adjusted post hoc test (for continuous covariates). Tests of difference between non-normally distributed data were performed using a Wilcoxon rank-sum test. To examine the degree of elevation in LV filling pressures relative to increments in CO during exercise, we used least-squares regression to calculate the slope of the relationship between PCWP and CO (PCWP/CO slope) from rest to peak exercise within a given individual. To define a “normative” value for PCWP/CO slope, we used our control population, defining a PCWP/CO slope two standard deviations above the mean value as “abnormal.” Multiple-subject PCWP/CO data were pooled for analysis according to the method of Poon¹⁴, as has been employed by other studies in this area¹⁵.

Outcomes were adjudicated by 3 physicians (JEH, RS, GDL) based on review of the electronic medical record at our center from the time of CPET examination to 9/1/2017 (median follow-up 5.3 years). Our primary endpoint was a composite of cardiovascular death, abnormal resting PCWP on future right heart catheterization (performed after initial CPET, defined as PCWP≥15 mmHg), or HF hospitalization. The rationale for inclusion of future right heart catheterization with PCWP≥15 mmHg was to capture those individuals with incident HF suspected clinically and diagnosed on hemodynamic criteria. Cardiovascular death was defined as any death due to ischemic heart disease, arrhythmia, or progressive heart failure. HF hospitalization was adjudicated as any hospitalization prompted by clinical HF (as defined by Framingham Criteria) or the need for escalation of diuretic or vasoactive therapy. Kaplan-Meier survival analysis with the log-rank test was used to determine the ability of PCWP/CO (as a dichotomous variable) to predict outcomes. Cox proportional hazards models (adjusted for age, sex, and body mass index) were used to evaluate the association between PCWP/CO slope and clinical outcomes. All analyses were performed in R (version 3.1.0) or Stata (version 12). A two-sided P value of <0.05 was considered statistically significant.

RESULTS

Baseline characteristics of the study population

Characteristics of our study population (N=175) are reported in Table 1, stratified by the three clinical groups (controls, HFpEF, or undifferentiated dyspnea on exertion with normal resting PCWP, hereafter labeled “DOE-nlrW”). The DOE-nlrW cohort was intermediate to controls and HFpEF for age (57±17 vs. 49±18 vs. 67±13, respectively) and BMI (29.0±6.7 vs. 25.9±4.3 vs. 32.2±7.1, respectively). Frequency of comorbid risk factors for HFpEF (lipid abnormalities, diabetes, and hypertension) as well as medication use (renin-angiotensin-aldosterone system blockade and beta-blockers) was also intermediate in the DOE-nlrW cohort relative to other groups. By definition, supine resting PCWP was higher in HFpEF (20 ± 3 mmHg) compared to controls or individuals in the DOE-nlrW group (9±3 and 10±3 mmHg, respectively).

Table 1.

Baseline characteristics of the study cohorts.

Characteristic	All (N=175)	Controls (N=33)	HFpEF (N=32)	DOE-nlrW (N=110)
Age (years)	57±17	49±18	67±13*	56±17‡
Women (%)	63	64	59	64
Body Weight	80.3±19.2	73.4±16.5	88.9±16.7	79.9±19.8
Body mass index (kg/m2)	28.7±6.6	25.9±4.3	32.2±7.1*	29.0±6.7‡
Comorbidities (%)
Hypertension	46	21	66*	47‡
Hyperlipidemia	37	15	53*	40‡
Diabetes	14	3	34*	11‡
Pharmacotherapy (%)
ACE Inhibitor or ARB	27	15	41*	27
Beta blocker	33	9	56*	34†‡
Diuretic	30	3	72*	27†‡
Exercise Capacity and Hemodynamics
PCWP/CO slope	1.7 (1.2–2.8)	1.2 (0.9–1.5)	3.1 (2.3–4.6)*	1.7 (1.2–2.6) †‡
HR (rest)	76±15	77±16	73±16	76±15
HR (peak)	137±26	157±19	119±26*	137±24†‡
HR (% pred)	84±13	92±7	78±17*	84±12†‡
VCO2/VO2 (RER, peak)	1.14±0.12	1.18±0.09	1.07±0.11*	1.15±0.12‡
Lactate (rest)	0.7±0.4	0.7±0.4	0.8±0.3	0.7±0.4
Lactate (peak)	5.8±2.2	7.5±2.2	4.6±1.9*	5.7±2.0†‡
Peak VO2 (ml)	1398±553	1872±584	1152±388*	1327±449†
Peak VO2/kg (ml/kg)	18.0±7.3	26.1±8.1	12.9±3.5*	17.0±5.8†‡
CO (rest, L/min)	5.2±1.6	5.9±1.3	4.5±1.2*	5.2±1.7
CO (peak, L/min)	12.3±3.6	15.1±3.0	10.1±2.8*	12.1±3.4†
VE/VCO2 slope	34.0±7.9	33.5±7.5	36.1±2.4	33.6±9.0
Supine PCWP (mmHg)	12±5	9±3	20±3*	10±3‡
RAP (rest)	2±2	2±2	4±2*	1±2†‡
RAP (peak)	9±5	8±3	14±6*	8±5‡
PAP (rest)	17±5	14±3	22±5*	17±5†‡
PAP (peak)	36±10	30±5	46±8*	35±10†‡

Abbreviations: BMI, body mass index; HTN, hypertension; ACE, angiotensin converting enzyme; ARB, angiotensin II receptor blocker.

ANOVA with Bonferroni-adjusted post-hoc test was used to determine between-group differences.

^*P<0.05 between Controls and HFpEF;

^†P<0.05 between Controls and DOE-nlrW;

^‡P<0.05 between HFpEF and DOE-nlrW.

PCWP/CO slope is reported as median (IQR).

Gas exchange and hemodynamic response to exercise

Gas exchange and hemodynamic measurements at rest are shown in Table 1. Mean exercise capacity of the DOE-nlrW (peak VO₂/kg: 17.0±5.8 ml/min/kg) was intermediate to that of control (26.1±8.1ml/min/kg) and HFpEF (12.9±3.5 ml/min/kg) groups. We examined the minute-by-minute PCWP responses during exercise, relative to CO augmentation. We observed a linear relationship between PCWP and CO during exercise (Figure 1 and Supplemental Figure 1). For within-individual linear regressions of minute by minute PCWP and CO associations, the mean r² for all subjects was 0.8±0.1 with a mean number of contributing data points of 10±2 per participant, which was not different among the three groups. When all 3 groups were combined, we found a PCWP/CO mean slope of 2.2±1.5 mmHg/L/min. Control participants demonstrated the lowest PCWP/CO slope (1.2±0.4 mmHg/L/min) relative to HFpEF (3.6±1.9 mmHg/L/min) and DOE-nlrW groups (2.1±1.2 mmHg/L/min).

Figure 1.

Pulmonary capillary wedge pressure versus cardiac output measurements across two cohorts: HFpEF and healthy controls. The data has been pooled using the method of Poon to correct for differing numbers of data points (minutes of exercise) for each patient when calculating mean slope and standard deviation. The hashed line represents a PCWP/CO slope of 2.0, which nearly perfectly discriminates the two groups.

Based on our a priori definition of upper-range of normal being two standard deviations above the mean among controls (see Statistical analysis, above), the upper-limit of normal derived from our control population for PCWP/CO slope was 2.0 mmHg/L/min (Figure 2). Of note, we found that right atrial pressure was similar at rest and differed only modestly at peak exercise between individuals with steep versus normal PCWP/CO slope (PCWP/CO≤2 vs. >2mmHg/L/min, RAP: 8mmHg vs 10mmHg, respectively p=0.03), suggesting similar contributions of pericardial constraint in both groups.

Figure 2.

Box plots showing the difference between patients with PCWP/CO slope ≤ 2 vs. >2mmHg/L/min in peak VO2 (Panel A) and VE/VCO2 slope (Panel B). Tests of difference were performed using a Wilcoxon rank-sum test.

Association of PCWP/CO responses with peak aerobic capacity, ventilatory efficiency, and NT-proBNP

Based on our normative values determined above, we stratified the DOE-nlrW group into normal (≤ 2 mmHg/L/min) and abnormal (> 2 mmHg/L/min) PCWP/CO slope, to reflect PCWP responses to exercise (Table 2). Among the 110 patients in this group, we found 45 patients with an abnormally elevated PCWP/CO slope. Compared to those with normal PCWP/CO responses, patients with abnormal PCWP/CO responses were older without appreciable differences in HFpEF risk factors or medication use, and of note, no difference in resting PCWP (supine: 10±3 vs 10±3 mmHg, p = 0.54; upright: 6±3 vs 6±2 mmHg, p=0.89, respectively). With respect to other exercise parameters, those with abnormal PCWP/CO slope achieved a lower peak VO₂ (14.3±3.5 vs. 19.0±6.3 ml/kg/min, p<0.0001, Figure 2) and had worse ventilatory efficiency when compared with participants with normal PCWP/CO slope (V_E/VCO₂ slope: 37.4±11.1 vs 31.5±6.2, p=0.002, Figure 2). When analyzed as continuous variable, PCWP/CO was related to peak VO₂ (ρ=0.37, P<0.001) and VE/VCO2 slope (ρ=0.29, p=0.003), whereas peak PCWP was not significantly related to either exercise parameter (p>0.50 for both).

Table 2.

Characteristics of DOE-nlrW cohort, stratified by PCWP<2 vs. ≥2mmHg/L/min.

Characteristic	PCWP/CO ≤ 2 (N=65)	PCWP/CO > 2 (N=45)	P Value
Age (years)	52±16	63±16	0.0004
Female (%)	58	71	0.18
Body mass index (kg/m2)	28.3±6.8	28.8±6.6	0.68
Comorbidities (%)
Hypertension	46	49	0.78
Hyperlipidemia	37	44	0.43
Diabetes	8	16	0.19
Pharmacotherapy (%)
ACE-I or ARB	28	27	0.91
Beta blocker	30	40	0.26
Diuretic	22	33	0.18
Exercise Capacity
Peak VO2/kg (ml/kg)	19.0±6.3	14.3±3.5	<0.0001
VE/VCO2 slope	31.1±6.2	37.1±11.1	0.002
Peak VCO2/VO2 (RER)	1.17±0.11	1.13±0.14	0.12
Vital Signs
HR (rest)	76±15	76±15	0.87
SBP (rest)	139±22	151±22	0.004
DBP (rest)	74±11	76±9	0.29
HR (peak)	143±23	128±24	0.0009
SBP (peak)	181±29	187±28	0.25
DBP (peak)	88±16	88±14	0.95
Fick Cardiac Output (CO)
Fick CO (rest, L/min)	5.4±1.8	4.8±1.3	0.06
Fick CO (peak, L/min)	13.3±3.4	10.4±2.7	<0.0001
CaVO2 (rest, ml/dL)	5.9±1.2	5.7±1.1	0.39
CaVO2 (peak, ml/dL)	11.3±1.9	10.4±2.1	0.03
Hemodynamics (mmHg, end-expiratory)
Supine PCWP (rest)	10±3	10±3	0.54
Seated PCWP (rest)	6±3	6±2	0.89
PCWP (30 Watts)	10±5	16±7	<0.0001
PCWP (peak)	18±6	25±8	<0.0001
RAP (rest)	1±2	1±2	0.64
RAP (peak)	8±5	10±4	0.03
PAP (rest)	16±5	18±6	0.09
PAP (peak)	33±9	40±11	0.0004
LVEF (rest)	66±7	63±9	0.07
LVEF (peak)	69±7	65±8	0.01

Abbreviations: ACE-I, angiotensin converting enzyme inhibitor; ARB, angiotensin II receptor blocker; RAP, right atrial pressure; PAP, pulmonary artery pressure (mean).

Finally, the median NT-proBNP level across all 110 participants in the dyspnea cohort was 82 pg/ml (interquartile range 33–195 pg/ml), with 13 (12%) over 400 pg/ml and 6 (5%) above age-adjusted normal value cut-points. NT-proBNP was significantly related to PCWP/CO slope in the 110 patients with undifferentiated dyspnea (Spearman ρ=0.37, P<0.0001) but was not significantly related to peak PCWP (p=0.17).

Association of PCWP/CO slope with progression to heart failure

We followed the DOE-nlrW group for a primary outcome of cardiovascular death, incident HF hospitalization, or subsequent referral for right heart catheterization with abnormal resting PCWP (≥ 15 mmHg). Over a median follow-up time of 5.3 years, we observed 20 primary events (11 HF hospitalizations, 4 cardiovascular deaths, and 5 right heart catheterizations with elevated resting PCWP) among the 45 patients with PCWP/CO > 2mmHg/L/min compared to 4 primary events (2 HF hospitalizations and 2 RHCs with elevated PCWP) among the 65 individuals with normal PCWP/CO slope (Figure 3). The unadjusted hazard ratio for the primary outcome was 4.5 (95% confidence interval (1.6–12.4, p=0.004). Adjusted for age, sex, and BMI, individuals with an abnormal PCWP/CO slope had a 3.4-fold higher hazard of our primary outcome (95% CI 1.2–10.4, P=0.03) relative to patients with a normal slope. Further adjustment for resting PCWP (HR 3.7, 95% CI 1.2–11, p=0.02) did not attenuate this effect, nor did exclusion of PCWP ≥ 15 mmHg in subsequent right heart catheterization from our endpoint (i.e. HR 3.5 for the primary outcome, 95% CI 1.2–10.4, P=0.03).

Figure 3.

Event-free survival in the absence of the composite endpoint of HF hospitalization, death, or elevated pulmonary capillary wedge pressure (defined in text). Individuals with a lower PCWP/CO slope have an improved event-free survival (log-rank P=0.004).

DISCUSSION

Based on more than 1,800 paired PCWP and CO measurements during incremental exercise in 175 patients, our principal original findings were threefold: (1) PCWP increases linearly relative to CO; (2) the normative PCWP/CO slope is <2 mmHg/L/min; and (3) an “abnormally steep” PCWP/CO slope during exercise is associated with worse aerobic fitness and adverse clinical outcomes in a referral population with normal resting PAP and PCWP and without known HFpEF. Importantly, along with modest NT-proBNP levels, over 40% of participants with dyspnea and normal resting LV filling pressures, displayed an abnormal PCWP response to exercise with correspondingly worse long-term outcomes. Collectively, these findings suggest that the assessment of resting LV filling pressures in evaluating exertional dyspnea and early HFpEF may miss prognostically and mechanistically distinct subtypes. Our findings highlight the potential of exercise hemodynamics in providing greater resolution to disease phenotypes and subsets for clinical surveillance and application of therapies.

HFpEF is a highly heterogeneous and complex entity, affecting many organ systems from the heart to lungs to vasculature, all of which may result in specific disease subphenotypes. Specific subphenotypes not only identify distinct pathophysiology, but also define HFpEF groups with distinct clinical course and prognosis.¹⁶ Moreover, several physiologic observations not classically a part of the HFpEF definition (e.g., exercise left atrial dysfunction¹⁷ and impaired peripheral skeletal muscle oxygen extraction¹⁸) impact exercise capacity¹⁹ and cardiovascular outcome. Therefore, a more precise definition of a cardiac contribution to dyspnea in individuals with suspected HFpEF is warranted.

Exercise is a potent physiologic probe that may uncover occult hemodynamic dysfunction, specifically in those individuals with effort intolerance. The growing recognition that responses in LV filling pressures with exercise may be a defining hallmark of HFpEF has been captured by several clinical studies.^{5, 6} Nevertheless, there remains significant heterogeneity in what constitutes a normal response in PCWP during exercise. Reeves et al. found that in normal individuals, PCWP values may exceed 30 mmHg during intense cycle ergometry exercise,¹⁰ suggesting the need to exercise caution in use of solitary cut-points for PCWP that do not account for exercise intensity. Furthermore, a detailed approach to zero-leveling during supine versus upright exercise—which can significantly impact the measured threshold values—remains a technical challenge.

In our study, we sought to address the aforementioned limitations using serial assessment of exercise hemodynamics. Analogous to prior efforts to index filling pressures to flow (cardiac output), we calculated a PCWP-to-CO slope to “correct” the augmentation in PCWP for an increase in cardiac performance as measured by CO. This definition quantifies the “cost” of rise in CO with exercise in terms of LV filling pressure. In addition, given concerns raised by Poole et al. of accurate estimation of VO₂max with a single exercise test^{20, 21} and the challenge of measuring PCWP exactly at peak exercise, our analysis of paired PCWP/CO slope throughout exercise provides data that is not subject to imprecise definition of “peak” exercise. The use of a control population in this study with exercise capacity above 85% predicted and hemodynamic measurements facilitated calculation of a normative PCWP/CO slope and its upper bound (>2 mmHg/L/min). This knowledge of normal PCWP/CO slope complements previous findings from our group and others indicating that the upper bound of PAP/CO slope is 3 mmHg/L/min.^{22, 23} Of note, the 67% contribution of PCWP to PAP during exercise mirrors the relationship of normal resting PCWP and mean PAP values of 10 mmHg and 15 mmHg, respectively. In addition, the use of a composite index like PCWP/CO slope may shed light on peripheral mechanisms of exercise intolerance: for example, the observation that the A-V oxygen content differences were lower in individuals with PCWP/CO slope >2 suggest co-existing impairment in central hemodynamics and oxygen diffusive conductance in the periphery, potentially referable to abnormalities in the microcirculation.

Leveraging serial hemodynamic data during exercise, we found that over 40% of individuals with undifferentiated dyspnea and normal PCWP at rest had an abnormal PCWP/CO response that was unmasked with exercise, which in turn was associated with worse functional capacity and clinical outcomes. We observed these results in the face of resting PCWP and CO that were not different among patients with abnormal versus normal PCWP/CO slope. This suggests that resting hemodynamic information may not be sufficient in defining contributions of exercise LV dysfunction to high filling pressure and dyspnea. Our results not only uniquely develop PCWP/CO as a normative hemodynamic index for further mechanistic study (especially with emerging therapies targeting LV diastolic function), but also provide evidence for its importance in defining disease heterogeneity and its prognostic relevance.

There is mounting evidence emerging in HFpEF that perturbations in intracardiac hemodynamics with exercise may be useful: exercise PAP and PCWP values in HFpEF exceed those in hypertensive controls without HF, exercise also more accurately distinguishes individuals with HFpEF from healthy volunteers relative to intravenous saline loading.²⁴ Further, the limitations of single point-values of exercise hemodynamics to define pathophysiology are increasingly recognized in the characterization of pulmonary hypertension.¹² To date there has been one study of outcomes based on resting PCWP versus a single exercise PCWP measurement indexed to workload, which showed superior prognostication with exercise PCWP/Watt measures.²⁵ Our study extends the findings of Dorfs et al. by providing normative value data set, associating exercise PCWP patterns with exercise performance, and specifically delineating the future burden of progression to HF as opposed to all cause mortality.

While trials evaluating efficacy of medical therapies for HFpEF have been disappointing to date, newer efforts to treat HFpEF have targeted specific phenotypic subgroups, including abnormal right ventricular-pulmonary arterial coupling^{26, 27}, venous capacitance and LV-aortic coupling⁶, heart rate modulation,²⁸ pulmonary and systemic vasodilator therapies (e.g., nitric oxide modulation^{7, 29–32}), and other comorbid illnesses contributing to LV stiffness (e.g., obesity, atrial fibrillation, and myocardial fibrosis). A positive impact of each of these distinct therapeutic approaches relies on a relatively homogenous underlying physiology. Given the lack of success of HFpEF intervention studies in broadly defined HFpEF, more precise phenotyping through PCWP/CO slope characterization may aid in selecting patients for targeted therapies to improve cardiac filling pressures.

The strengths of our study include detailed and standardized minute-by-minute gas exchange and hemodynamic responses to exercise, attention to exercise intensity normalization, and careful adjudication of clinical outcomes. Several limitations deserve mention. Ours was an observational study of clinically referred patient samples, and subject to referral and selection bias on that basis. While our control population for deriving a normative bound was limited (N=33), serial measures of hemodynamics within individuals were used to derive normative PCWP/CO, greatly increasing precision. It is important to note that while there was a nearly 20-year difference between the three subgroups in our study (Table 1; due to a referral sample used to establish normative values), downstream survival analyses were adjusted for age. Nevertheless, future studies using non-invasive metrics of filling pressure (e.g., echocardiography) in an age/sex-matched fashion will be required to confirm our findings. In addition, our study was not designed to identify what aspects of LV physiology lead to abnormal PCWP/CO slope (e.g., degree of LV fibrosis) or what role pericardial constraint may play in the observed relationship. Nevertheless, we found that right atrial pressure was similar at rest and differed only modestly at peak exercise (Table 2) in individuals with steep and normal PCWP/CO slope, suggesting limited contribution of pericardial constraint. Further studies to understand underlying mechanisms are warranted. Our HF-event free survival analysis should be viewed as exploratory, given potential ascertainment bias (which individuals had a follow-up right heart catheterization) and the limited number of events. Further investigations in a larger population at higher cardiac risk are warranted to validate and extend our findings.

In conclusion, the PCWP/CO slope integrates serial hemodynamic measures in response to exercise, and provides additive information above and beyond resting hemodynamics to identify individuals with subclinical HFpEF and normal PCWP at rest. Over 40% of individuals with dyspnea and normal LV filling pressure at rest without overt HFpEF had an abnormally steep PCWP/CO slope. Importantly, a steep PCWP/CO slope in these patients was associated with worse cardiorespiratory fitness and worse clinical outcomes. These results suggest that current definitions of HFpEF based on single hemodynamic measures during rest or exercise may miss a significant proportion of individuals with latent HFpEF, unique physiology, and potential targets for improved therapy and surveillance. Further investigation to characterize individuals with this “exercise induced” phenotype is warranted.

What is new

• Exercise may unmask heart failure with preserved ejection fraction (HFpEF) when resting hemodynamic measurements are normal. However, the upper limit of normal left heart filling pressures during exercise is not well-established and should account for degree of exercise exposure.

• We defined the distribution of pulmonary capillary wedge pressure (PCWP) relative to cardiac output (CO) as a marker of dynamic left-sided filling pressures during incremental exercise, demonstrating that an abnormally steep “PCWP/CO slope” > 2.0 mmHg/L/min was common in patients with dyspnea and normal resting hemodynamics.

What are the clinical implications

• A steep PCWP/CO slope was closely related to impaired exercise capacity and predicted worse HF-free survival.