Proteomic Correlates of the Urinary Protein/Creatinine Ratio in Heart Failure With Preserved Ejection Fraction

Abstract

Proteinuria is common in heart failure with preserved ejection fraction (HFpEF), but its biologic correlates are poorly understood. We assessed the relation between 49 plasma proteins and the urinary protein/creatinine ratio (UPCR) in 365 participants in the Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist Trial. Linear regression and network analysis were used to represent relations between protein biomarkers and UPCR. Higher UPCR was associated with older age, a greater proportion of female gender, smaller prevalence of previous myocardial infarction, and greater prevalence of diabetes, insulin use, smoking, and statin use, in addition to a lower estimated glomerular filtration rate, hematocrit, and diastolic blood pressure. Growth differentiation factor 15 (GDF-15; β = 0.15, p <0.0001), followed by N-terminal proatrial natriuretic peptide (NT-proANP; β = 0.774, p <0.0001), adiponectin (β = 0.0005, p <0.0001), fibroblast growth factor 23 (FGF-23, β = 0.177; p <0.0001), and soluble tumor necrosis factor receptors I (β = 0.002, p <0.0001) and II (β = 0.093, p <0.0001) revealed the strongest associations with UPCR. Network analysis showed that UPCR is linked to various proteins primarily through FGF-23, which, along with GDF-15, indicated node characteristics with strong connectivity, whereas UPCR did not. In a model that included FGF-23 and UPCR, the former was predictive of the risk of death or heart-failure hospital admission (standardized hazard ratio 1.83, 95% confidence interval 1.49 to 2.26, p <0.0001) and/or all-cause death (standardized hazard ratio 1.59, 95% confidence interval 1.22 to 2.07, p = 0.0005), whereas UPCR was not prognostic. Proteinuria in HFpEF exhibits distinct proteomic correlates, primarily through its association with FGF-23, a well-known prognostic marker in HFpEF. However, in contrast to FGF-23, UPCR does not hold independent prognostic value.

Heart failure with preserved ejection fraction (HFpEF) is extremely prevalent and is associated with significant morbidity and mortality. Its complex pathophysiology is incompletely understood but involves multiple organs, including but not limited to the heart, blood vessels, and the kidney.^1,2 Previous studies have shown that albuminuria is associated with worse clinical outcomes independently of common co-morbidities like diabetes, hypertension, and renal disease in cohorts comprising predominantly heart failure with reduced ejection fraction (HFrEF).^3–5 Similarly, proteinuria is present in nearly 1/2 of patients with HFpEF and is associated with a worse prognosis.^4,6–8 Proteinuria may be a biomarker of various pathophysiologic processes, including systemic inflammation, venous congestion, endothelial dysfunction, and microvascular dysfunction,^9–11 all of which are known or suspected to be involved in the pathophysiology of HFpEF. Although previous studies have assessed the prognostic value of proteinuria in HFpEF, the underlying biologic pathways associated with proteinuria in this population are poorly understood.¹¹ Using plasma proteomics approaches, it is possible to uncover new risk biomarkers and/or biologic insights in patients with HFpEF. Several previous studies have focused on the role of plasma proteomics in HFpEF to characterize phenogroups, or as prognostic indicators^12–21 in this population, but there is limited knowledge about the biologic or proteomic correlates of proteinuria in HFpEF. The purpose of this study was to assess plasma proteomic correlates of proteinuria, as measured by urine protein/creatinine ratio (UPCR), in HFpEF in participants in the Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist Trial (TOPCAT).

Methods

We used data and available baseline urine and plasma samples from participants in TOPCAT, obtained through the National Institutes of Health BioLINCC repository. The general characteristics, study design, and primary results have been previously published.^22–24 Briefly, TOPCAT enrolled 3,445 adults with HFpEF (left ventricular ejection fraction >45%) and was performed in 266 centers across 6 countries (including the Americas and Eastern Europe) between 2006 and 2012. It was an international, double-blind, multicenter, parallel group, placebo-controlled trial that evaluated whether spironolactone leads to improved cardiovascular outcomes in patients with HFpEF with cumulative incidence of the composite primary outcome of cardiovascular death, heart-failure hospitalization, or aborted cardiac arrest, as opposed to placebo. Written informed consent was obtained from all study participants.

Of the 3,445 participants, 365 patients provided baseline (prerandomization) plasma and urine samples for analysis for de novo UPCR and plasma protein quantification, along with clinical data, which were obtained through the National Institutes of Health BioLINCC repository.

The raw data and analytical methods of this study are not publicly available for purposes of reproducing the results or replicating the procedures. These data might be available subject to the establishment of appropriate data-sharing agreements and regulatory approvals. The parent TOPCAT trial data are available through the US National Institutes of Health BioLINCC repository.

We used a Luminex bead-based multiplexed assay (Bristol-Myers Squibb, Ewing Township, New Jersey) to measure 49 protein analytes. Analytes were chosen to represent a diverse number of physiologic processes that are related to cardiovascular disease, downstream effects including angiogenesis, atherothrombosis, cardiomyocyte injury, extracellular matrix turnover, cell matrix interactions, tissue remodeling, inflammation, adipocyte signaling, intermediary metabolism, kidney function/injury, calcification/mineral metabolism, neurohormonal regulation, and myocyte stretch (Supplementary Table 1, Figure 1).

Figure 1.

Pathophysiologic domains related to cardiovascular disease for the biomarkers included in the Luminex panel and general analytical design.

Participant characteristics were summarized using mean ± SD for continuous variables with a normal distribution and median (interquartile range) for continuous variables with a skewed distribution. Categorical variables were expressed as counts (percentages). We compared continuous variables between tertiles of UPCR using analysis of variance for normally distributed variables, the Kruskal –Wallis test for skewed variables, and the chi-square or Fisher’s exact test, as appropriate, for categorical data.

We performed linear regression to assess proteomic correlates of UPCR, in unadjusted analyses, and analyses that adjusted for key clinical correlates of UPCR. Volcano plots were plotted to visualize the relation between each plasma protein and UPCR. We corrected the alpha level for multiple comparisons using the number of principal components underlying >95% of the variability of all measured proteins, as previously described.^16,20,25,26 All regression coefficients (β) are standardized (expressed per unit increase in z score) to facilitate comparisons between different plasma proteins.

The relation between UPCR and the risk of death or heart-failure–related hospitalization (DHFA) and the risk of all-cause death was assessed using Cox regression. To provide an intuitive unit-independent comparison between biomarkers, hazard ratios (HRs) for all biomarkers are standardized. We also performed survival analyses using Kaplan-Meier plots for tertiles of UPCR.

To visualize the complex relation between multiple plasma proteins and UPCR, we applied network analysis, with the nodes representing individual biomarkers and the edges (connections) between nodes representing the correlation coefficients in a given biomarker (node) pair. To better visualize structural patterns within this correlation matrix, the network connectivity backbone and the minimum spanning network were extracted, revealing dominant connections and clusters of dense connectivity.²⁷

Statistical significance was defined as a 2-tailed p <0.05. Statistical analyses were performed using the Matlab statistics and SPSS for Mac version 22 (Version 22.0., IBM Corp., Armonk, New York).

Results

Table 1 presents a comparison of participants in the trial who had available urine samples for measurement of UPCR with those who did not have available urine samples. Study participants with available UPCR included a greater prevalence of male gender (53.99%, p = 0.015), smoking (47.77%, p = 0.005), hypertension (94.84%, p = 0.007), atrial fibrillation (42.25%, p = 0.001), previous myocardial infarction (31.92%, p = 0.003), and history of coronary artery bypass graft (17.84%, p = 0.001) and exhibited lower blood pressure, greater body mass index, and higher Kansas City Cardiomyopathy Questionnaire score. They also exhibited less use of angiotensin-converting enzyme inhibitor/angiotensin receptor blocker and greater use of statins than did participants without available urine samples.

Table 1.

General characteristics of study participants with versus without available UPCR measurements (the study was conducted in participants with available UPCR measurements)

	Participants Without Available UPCR (n= 3016)	Participants With Available UPCR (n= 426)	P value
Demographic Characteristics
Age, yrs	69 (61,76)	69 (61,77)	0.2647
Male Sex	1438 (47.68%)	230 (53.99%)	0.0147
Race			0.1231
White	2668 (88.46%)	391 (91.78%)
Black	274 (9.08%)	28 (6.57%)
Asian and other	74 (2.45%)	7 (1.64%)
BMI, kg/m 2	30.8 (27.1,35.6)	32 (27.8,36.4)	0.0150
Systolic BP, mmHg	130 (120,140)	128 (120,134)	<0.0001
Diastolic BP, mmHg	80 (70,81)	75 (68,80)	<0.0001
Pulse pressure, mmHg	50 (45,60)	50 (44,60)	0.0244
History of Smoking	1085 (40.20%)	182 (47.77%)	0.0049
Medical History
Myocardial infarction	757 (25.11%)	136 (31.92%)	0.0027
Stroke	234 (7.76%)	31 (7.28%)	0.7257
PCI	429 (14.23%)	71 (16.67%)	0.1814
CABG	367 (12.17%)	76 (17.84%)	0.0011
COPD	349 (11.58%)	54 (12.68%)	0.5084
Hypertension	2742 (90.95%)	404 (94.84%)	0.0073
Atrial Fibrillation	1033 (34.26%)	180 (42.25%)	0.0012
Diabetes Mellitus	980 (32.50%)	138 (32.39%)	0.9639
Insulin Use	375 (12.43%)	52 (12.21%)	0.8941
eGFR, mL min −1 1.78m −2	65.4 (53.7, 79.3)	65.7 (53.7,77.9)	0.6747
Hematocrit, %	40 (37,43)	40.4 (36.9,44)	0.1658
Medication use
Beta-blockers	2334 (77.41%)	342 (80.28%)	0.1826
Calcium channel blockers	1141 (37.84%)	152 (35.68%)	0.3881
ACE inhibitors or ARBs	2558 (84.84%)	341 (80.05%)	0.0110
Aspirin	1972 (65.41%)	278 (65.26%)	0.9520
Statins	1528 (50.68%)	277 (65.02%)	<0.0001
KCCQ Score Overall	53.1 (38.8,69)	59.4 (43.5,75.5)	<0.0001
KCCQ clinical sum score	56.3 (41.9,71.4)	61.5 (46.4,76.6)	<0.0001
MAGGIC score	14 (10,18)	15 (11,19)	0.0940

Values represent median (interquartile range) or n (%)

ACE = angiotensin-converting enzyme; ARB = angiotensin receptor blocker; BMI = body mass index; BP = blood pressure; CABG = coronary artery bypass graft; COPD = chronic obstructive pulmonary disease; GFR = glomerular filtration rate.

Table 2 lists the characteristics of the study participants stratified by tertiles of UPCR. Participants in higher tertiles were older (p = 0.003), with a greater proportion of women (p = 0.046) and greater prevalence of smoking (p = 0.038). They also exhibited less previous myocardial infarction (25.41%, p = 0.01), greater prevalence of diabetes (40.38%, p = 0.013), greater insulin use (18.85%, p = 0.004) and statin use (72.95%, p = 0.033), and lower estimated glomerular filtration rate (eGFR), hematocrit, and diastolic blood pressure, and higher pulse pressure. Participants in higher tertials also exhibited higher Meta-Analysis Global Group in Chronic Heart Failure scores.

Table 2.

Demographic and clinical characteristics of the 365 study participants included in this study, stratified by tertile of UPCR

	Lowest Tertile	Mid Tertile	Highest Tertile	P value
	(UPCR= 7.42 to 27.08)	(UPCR= 27.08 to 54.6)	(UPCR= 54.6 to 1140.81)
Demographic Characteristics
Age, yrs	67 (61,74)	70 (61.8,76.3)	72 (65,79)	0.0030
Male Sex	77 (63.11%)	60 (49.59%)	60 (49.18%)	0.0457
Race				0.6751
White	112 (91.80%)	109 (90.08%)	112 (91.90%)
Black	8 (6.56%)	10 (8.26%)	10 (8.20%)
Asian and other	2 (1.64%)	2 (1.65%)	0 (0.00%)
BMI, kg/m 2	30.9 (27.7,35.3)	31.7 (27.4,36)	32.6 (28.7,36.6)	0.3471
Systolic BP, mmHg	125 (120,130)	130 (115,135)	127 (118,135)	0.9514
Diastolic BP, mmHg	80 (70,80)	74 (64,80)	70 (64,80)	0.0009
Pulse pressure, mmHg	50 (43,55)	50 (43.8,60.5)	54 (44,62)	0.0121
History ofSmoking	45 (45.00%)	48 (42.48%)	66 (58.41%)	0.0376
Medical History
Myocardial infarction	50 (40.98%)	31 (25.62%)	31 (25.41%)	0.0104
Stroke	6 (4.92%)	12 (9.92%)	9 (7.38%)	0.3301
PCI	20 (16.39%)	13 (10.74%)	27 (22.13%)	0.0568
CABG	18 (14.75%)	19 (15.70%)	31 (25.41%)	0.0610
COPD	14 (11.48%)	11 (9.09%)	20 (16.39%)	0.2101
Hypertension	118 (96.72%)	115 (95.04%)	113 (92.62%)	0.3502
Atrial Fibrillation	47 (38.52%)	51 (42.15%)	59 (48.36%)	0.2919
Diabetes Mellitus	29 (23.77%)	44 (36.36%)	50 (40.38%)	0.0131
Insulin Use	6 (4.92%)	16 (13.22%)	23 (18.85%)	0.0039
eGFR, mL min −1 .78m −2	68.5 (58.1,80)	65.4 (53,75)	60.4 (48.9,72.6)	0.0026
Hematocrit, %	41.4±5.7	39.7±5.1	39.1±5	0.0024
Medication use
Beta-blockers	96 (78.69%)	97 (80.1%)	97 (79.51%)	0.9601
Calcium channel blockers	40 (32.79%)	41 (33.88%)	47 (38.52%)	0.6085
ACE inhibitors or ARBs	98 (80.33%)	100 (82.64%)	97 (79.51%)	0.8129
Aspirin	81 (66.39%)	82 (67.77%)	77 (63.11%)	0.7343
Statins	70 (57.38%)	82 (67.77%)	89 (72.95%)	0.0327
KCCQ Score Overall	60.3 ±20	61.3±21.6	60.3±21.1	0.9113
KCCQ clinical sum score	62.8±19.9	63.1±21.2	60.7±21.3	0.6202
MAGGIC score	14.1±5.5	15±5.5	16.8±5.6	0.0007

Values represent median (interquartile range) or n (%).

ACE = angiotensin-converting enzyme; ARB = angiotensin receptor blocker; BMI = body mass index; BP = blood pressure; CABG = coronary artery bypass graft; COPD = chronic obstructive pulmonary disease; GFR = glomerular filtration rate.

Figure 2 shows a volcano plot indicating the relation of various plasma proteins and UPCR. In nonadjusted models, UPCR was found to be significantly associated with 10 proteins. These included biomarkers related to intermediary metabolism (growth differentiation factor 15 [GDF-15]), adipocyte biology (adiponectin), 2 biomarkers related to mineral metabolism/calcification (fibroblast growth factor 23 [FGF-23] and osteoprotegerin [OPG]), 3 biomarkers related to inflammation (soluble tumor necrosis factor receptor [sTNFR 1, sTNFR2], interleukin 6 [IL-6]), 2 biomarkers related to neurohormonal regulation and myocyte stretch (N-terminal proatrial natriuretic peptide [NT-proANP], N-terminal pro-B-type natriuretic peptide [NT-proBNP]), and 1 biomarker related to extracellular matrix turnover (matrix metallopeptidase 2 [MMP-2]). GDF-15 (β = 0.15, p <0.0001) followed by NT-proANP (b = 0.774, p <0.0001), adiponectin (β = 0.0005, p <0.0001), FGF-23 (β = 0.17, p <0.0001), and sTNF1 (β = 0.002, p <0.0001) and sTNF2 (b = 0.093, p <0.0001) showed the strongest associations with UPCR.

Figure 2.

Volcano plot of protein correlates of UPCR through linear regression. Dashed lines represent the nominal (bottom dashed line) and alpha-error corrected (top dashed line) levels of significance.

In analyses that adjusted for age, gender, history of myocardial infarction, diabetes mellitus, insulin, smoking, eGFR, hematocrit, diastolic blood pressure, and statin use, no significant associations were found between any protein and UPCR, after alpha correction for multiple comparisons.

Figure 3 represents a network connectivity backbone, showing the relations between the multiple plasma proteins and UPCR. Network analysis showed that UPCR was linked to various proteins primarily through FGF-23, which, along with GDF-15, revealed node characteristics with strong connectivity. The UPCR did not exhibit node characteristics in either the minimum spanning network (Figure 3) or the network connectivity backbone and is peripheral to the network.

Figure 3.

Network analysis of plasma proteins and UPCR. Panel (A) shows the network connectivity backbone obtained through network analysis. The nodes represent individual biomarkers, and the connections between nodes represent the correlation coefficient in a given biomarker pair. Panel (B) shows the minimum spanning network. In both panels, the node size represents betweenness centrality, which quantifies the number of times a node acts as a bridge along the shortest path between 2 nodes. The node color represents the eigenvector centrality (which depends both on the number of neighbors and the strength of its connections). Eigenvector centrality measures a node’s importance while considering the importance (number of connections) of its neighbors.

In a model including FGF-23 and UPCR, FGF-23 was an independent predictor of the risk of DHFA (standardized HR 1.83, 95% confidence interval [CI] 1.49 to 2.26, p <0.0001), whereas UPCR was not (standardized HR 1.00, 95% CI = 0.82 to 1.24, p = 0.96). Kaplan–Meier plots showing the incidence of DHFA for tertiles of UPCR and FGF-23 are presented in Supplementary Figure 1. Similarly, FGF-23 was an independent predictor of the risk of all-cause death (standardized HR 1.59; 95% CI 1.22 to 2.07, p = 0.0005), whereas UPCR was not (standardized HR 1.07; 95% CI = 0.83 to 1.38, p = 0.59). Kaplan–Meier plots illustrating the incidence of death for tertiles of UPCR and FGF-23 are shown in Supplementary Figure 2.

Discussion

In the present study, we measured the UPCR and 49 selected proteins using a multiplex assay in baseline urine and plasma samples, respectively, to assess the plasma proteomic correlates of UPCR in participants in the TOPCAT study. Higher levels of UPCR were associated with multiple clinical factors, including older age, female gender, smoking, previous myocardial infarction, diabetes, insulin and statin use, lower eGFR, lower hematocrit, lower diastolic blood pressure, and higher pulse pressure. We found that UPCR is associated with various well-known plasma proteins related to intermediary metabolism (GDF-15), adipocyte biology (adiponectin), mineral metabolism/calcification (FGF-23 and OPG), inflammation (sTNFRI, sTNFRII, IL-6), myocyte stretch (NT-proANP, NT-proBNP), and extracellular matrix turnover (MMP-2). Using network analysis, we showed that UPCR is linked to various examined proteins primarily through its association with FGF-23, which, along with GDF-15, revealed node characteristics with pronounced connectivity in the network of proteins associated with UPCR. Moreover, UPCR was not associated with incident outcomes in this sample, in contrast to FGF-23, which was associated with incident death and DHFA, independently of UPCR.

We report various novel positive associations between plasma proteins and UPCR in HFpEF. Among the examined plasma proteins, GDF-15 exhibited the strongest positive associations with UPCR. GDF-15 is a stress-responsive cytokine and a member of the transforming growth factor/bone morphogenetic protein superfamily, which has been involved in multiple cardiovascular conditions, including heart failure, coronary artery disease, atrial fibrillation, and risk factors such as diabetes and hypertension.^28,29 It has been recently shown that GDF-15 secretion by the heart can increase in pathologic conditions and mediate physiologic effects on peripheral tissues (including regulation of liver growth hormone signaling, therefore acting as a cardiac hormone), in addition to autocrine effects on the myocardium.³⁰ GDF-15 is secreted by senescent cells. Senescence is a process by which cells exit the cell cycle permanently, entering a quiescent state characterized by secretion of a characteristic senescence-associate secretory phenotype, including GDF-15.³¹ GDF-15 has been previously shown by us and others to be associated with adverse outcomes in HFpEF.^21,32

We also found FGF-23 to be strongly associated with UPCR and to be closely linked to UPCR in the protein network. FGF-23, along with GDF-15, also exhibited node characteristics in network analyses, indicating its relation to multiple other plasma proteins. FGF-23 has been extensively studied and has been previously shown to be increased in chronic kidney disease and HFpEF. It has been implicated in the development of left ventricular hypertrophy and has also been associated with abnormal cardiac mechanics in the general population.³³ In a study in 6,542 participants who were free from cardiovascular diseases at baseline, it was found that greater serum FGF-23 concentrations were associated with incident HFpEF, but not with incident HFrEF, in the Multiethnic Study of Atherosclerosis.³⁴ FGF-23 has also been associated with fibrosis in HFpEF.³⁵

We found associations of UPCR with the natriuretic peptides NT-proANP and NT-proBNP. A particularly strong relation was found between UPCR and NT-proANP (Figure 2). Our findings are consistent with a recent study by Boorsma et al,³ who reported an association between plasma NT-proBNP and the urinary albumin/creatinine ratio (UACR) in both HFpEF and HFrEF. Interestingly, the authors observed that UACR clustered with clinical and biomarker indicators of congestion, rather than with glomerular filtration and tubular injury biomarkers.³ This led the authors to postulate that in heart failure, UPCR is a marker of venous congestion rather than kidney injury per se. Our findings of a particularly strong association between UPCR and NT-proANP support this notion, although in our cohort, GDF-15 appears to mediate the association of UPCR with other plasma proteins, as previously mentioned. Interestingly, in the study by Boorsma et al,³ the UACR also clusters closely with GDF-15 and FGF-23 in hierarchical cluster analyses, adding confidence to our observations regarding the relation between UPCR and these 2 biomarkers in HFpEF.

We also found an association between UPCR and adiponectin, which is a protein secreted by adipose cells and involved in fat metabolism. In a mouse model of HFpEF, adiponectin overexpression ameliorated left ventricular hypertrophy, diastolic dysfunction, lung congestion, and myocardial oxidative stress, without affecting blood pressure and the left ventricular ejection fraction.³⁶ A small study in 25 patients with mild HFpEF found an association between diastolic dysfunction and lower adiponectin concentrations.³⁷ Our study revealed a significant positive correlation between adiponectin and UPCR. Interestingly, adiponectin was also linked to inflammatory biomarkers, GDF-15, MMP-2, and NT-proBNP in network analyses, which suggests that despite its potentially favorable biologic effects, it may represent a marker of inflammation, venous congestion, or both, likely though noncausal mechanisms.

Inflammation has been proposed to play a key role in the pathophysiology of HFpEF.³⁸ The UPCR was also associated with inflammatory proteins, including sTNFRI, sTNFRII, and IL-6. All 3 of these biomarkers have been previously shown to be increased in participants with HFpEF relative to controls without heart failure, and sTNFRII was also reported to be increased in HFpEF compared with HFrEF.³⁹ A case-cohort study comprising 200 participants who developed HF and a random group of 761 controls showed that IL-6 is associated with new-onset HFpEF, independently of potential confounders.⁴⁰ Kalogeropoulos et al⁴¹ have shown in the Health, Aging, and Body Composition Study, a cohort involving older patients, that IL−6 was associated with incident HF and that this association was especially pronounced in patients with HFpEF. However, in a meta-analysis of various large cohorts, in the Multiethnic Study of Atherosclerosis cohort, de Boer et al⁴² showed IL-6 to be associated with new–onset HF, without apparent difference between HFpEF and HFrEF. We have previously shown that sTNFRI and IL-6 were associated with the risk of DHFA in participants with established HFpEF, and both sTNFRI and IL-6 were prioritized by the multimarker machine learning model for the prediction of DHFA in this population.²¹

Finally, UPCR was associated with OPG, a protein that belongs to the tumor necrosis factor superfamily implicated in bone metabolism.⁴³ OPG has also been shown to be associated with the risk of DHFA in HFpEF, although not independently of the Meta-Analysis Global Group in Chronic Heart Failure risk score.²¹ Higher OPG levels have been associated with left ventricular hypertrophy and increased left ventricular myocardial stiffness in a previous cohort comprising African-Americans with hypertension and their siblings.⁴⁴ Further research is required to elucidate the role of OPG in HFpEF and its link to UPCR in this population.

Study strengths and limitations

Our study should be interpreted in the context of its strengths and limitations. Strengths of our study include the well-characterized HFpEF cohort, the use of multiple plasma protein biomarkers preselected, given their representation of important biologic processes, and the use of both linear regression analyses and network analysis to better assess the structure of interrelations between UPCR and associated plasma proteins. Our study also has several limitations. We did not have available plasma samples and UPCR measurements from all participants in TOPCAT, and our study was therefore restricted to a subpopulation with available samples, which exhibited some differences from the remaining participants. Our study sample size was moderate, which limited the power to detect weaker associations. On the basis of recent literature, UPCR is not as precise an estimator per se as UACR, and its wide variance could contribute to less certainty than one might observe in assays of UACR.⁴⁵ Finally, the organ source of various plasma proteins could not be ascertained, limiting the interpretation of our findings. Importantly, our data are observational and cannot assess causality. However, our study indicates multiple relevant proteomic associations of UPCR in HFpEF and suggests that these proteins, and not UPCR per se, are associated with prognosis in this population.

In conclusion, proteinuria in HFpEF correlates with increased levels of various relevant plasma proteins in HFpEF, primarily through its association with FGF-23, a well-known prognostic marker in this population. However, in contrast to FGF-23, UPCR does not hold independent prognostic value.