Fibroblast Growth Factor 23 and Outcomes in Acute Heart Failure Authors: Arancha Martí-Martínez, Julio Núñez, Herminio López-Escribano, Rafael de la Espriella, Enrique Santas, Enrique Rodriguez-Borja, Òscar Miró, Pere Llorens, Aitor Alquézar-Arbé, Pablo Herrero-Puente Published: 2026 ABSTRACT Worsening heart failure (WHF) is a complex clinical syndrome associated with a high risk of mortality and further decompensations, characterized by fluid overload and upregulated inflammatory and fibrotic processes linked to organ dysfunction. Fibroblast growth factor 23 (FGF23) is an emerging proxy of fibrosis and may interact to influence long-term outcomes in WHF patients. This study evaluated the association between FGF23 levels and long-term outcomes in patients with WHF evaluated in the emergency department (ED). A total of 635 patients from the EAHFE registry, enrolled during 2007, 2009, and 2011, were prospectively included. FGF23 was measured in the first blood sample obtained at ED presentation. During a median follow-up of 380 days (interquartile range: 126 to 456), all-cause mortality occurred in 216 patients (34.0%), while 351 patients (55.3%) experienced the composite endpoint of death/WHF ED reconsultation and 295 (46.5%) experienced death/HF hospitalization. Event rates increased progressively across FGF23 levels. Patients with FGF23 above the median (48.4 pg/mL) displayed higher rates (per 100 patient-years) of all-cause mortality (49 vs. 31, p = 0.003), death/WHF ED reconsultation (93 vs. 65, p = 0.003), and death/HF hospitalization (77 vs. 56, p = 0.043). In continuous analyses, each doubling of FGF23 remained significantly associated with an increased risk of all three outcomes after multivariable adjustment. These associations were consistent across clinically relevant subgroups, including those defined by age, sex, and kidney function, supporting the potential prognostic value of FGF23 within this cohort, although external validation is warranted. 1. Introduction Worsening heart failure (WHF) is a complex and heterogeneous clinical syndrome characterized by high mortality rates and frequent decompensation episodes, imposing a substantial clinical and socioeconomic burden worldwide [1,2]. Identifying patients at higher risk remains challenging. A defining feature of WHF is the coexistence of volume expansion and inflammatory activation [3], processes that are closely intertwined and contribute to progressive organ dysfunction [4]. These mechanisms are further linked to myocardial remodeling and organ fibrosis, which play a central role in disease progression [4]. Increasing evidence supports the concept that biomarkers integrating these pathways may provide incremental prognostic information beyond traditional measures [5]. In this context, circulating biomarkers reflecting key pathophysiological pathways, such as volume expansion, inflammation and fibrosis, have emerged as potentially valuable tools for risk stratification, often preceding overt clinical deterioration [5,6]. Fibroblast growth factor 23 (FGF23), a member of the fibroblast growth factor family, is a bone-derived hormone that regulates phosphate homeostasis by promoting phosphate excretion, suppressing calcitriol synthesis, and modulating parathyroid hormone levels [7]. Beyond its canonical role in mineral metabolism, FGF23 has emerged as a pleiotropic molecule with significant cardiovascular (CV) implications [7]. Elevated FGF23 levels rise early in chronic kidney disease (CKD) and have been consistently associated with increased CV morbidity and mortality, as well as adverse cardiac remodeling [8]. Moreover, accumulating experimental and clinical data suggest that FGF23 may actively contribute to myocardial hypertrophy and fibrosis, rather than merely reflecting disease severity [7]. In patients with heart failure (HF), circulating FGF23 concentrations are markedly elevated, especially during episodes of acute decompensation and have been linked to greater disease severity, volume expansion and clinical outcomes [8]. These findings position FGF23 as a potential integrative biomarker bridging mineral metabolism, inflammation, and adverse disease course. Against this background, the present study aimed to evaluate the prognostic significance of FGF23 in a real-world cohort of patients with WHF evaluated in the emergency department (ED). Specifically, we examined the association between FGF23 levels and long-term adverse clinical outcomes, including all-cause mortality and composite endpoints of death/WHF ED reconsultation, and death/HF hospitalization. 2. Materials and Methods 2.1. Study Sample Of 764 consecutive patients with acute heart failure (AHF) enrolled in this EAHFE biobank substudy, 635 patients (83.1%) had adequate serum aliquot availability for simultaneous determination of circulating intact FGF23. The remaining 129 patients (16.9%) lacked biomarker measurements due to pre-analytical technical constraints (insufficient aliquot volume or hemolyzed specimens). As detailed in Table A1, excluded patients were clinically and demographically comparable to included patients, with similar event rates across all endpoints (all p > 0.25). This cohort has previously been used to investigate other circulating biomarkers in AHF [9]. The EAHFE registry is an ongoing Spanish multicenter observational registry enrolling consecutive patients evaluated for AHF in emergency departments from university, referral, and community hospitals [10]. To date, the EAHFE registry has completed eight phases of patient inclusion. For the present study, cases were randomly selected. AHF was diagnosed on the basis of compatible symptoms (e.g., dyspnea, orthopnea, paroxysmal nocturnal dyspnea) and signs (e.g., third heart sound, pulmonary crackles, jugular venous distension > 4 cm, resting sinus tachycardia, peripheral edema, hepatomegaly, or hepatojugular reflux) together with objective evidence of congestion; more than 95% of participants had additional diagnostic support from natriuretic peptide testing and/or echocardiography. Patients presenting with ST-elevation myocardial infarction as the primary diagnosis accompanied by AHF were excluded. Enrollment procedures and diagnostic criteria were standardized across participating centers. In this study, patients were categorized as having worsening chronic HF (acutely decompensated HF in patients with a confirmed prior history of HF; n = 401, 63.1%) or new-onset (de novo) AHF (first-time presentation without prior HF history; n = 234, 36.9%). The EAHFE registry adheres to the ethical principles outlined in the Declaration of Helsinki for medical research involving human subjects. Patient recruitment periods and sample collection were approved by the Institutional Review Board of Principado de Asturias (protocol codes 160/15, 205/17, and 2022.015.) All patients provided written informed consent for participation in the registry. The study was also approved by the Institutional Review Board of Hospital Clínico Universitario de Valencia (protocol code 2024/371), approved on 3 March 2025. 2.2. Biomarker Measurements Blood for biomarker assessment was obtained at the time of the initial ED evaluation. Plasma was prepared from lithium-heparin specimens and stored at −20 °C until analysis. Before FGF23 determination, samples underwent one freeze–thaw cycle. The FGF23 biomarker was subsequently analyzed at the Clinical Biochemistry and Molecular Pathology Laboratory of the University Clinical Hospital of Valencia. Physicians in charge of patient management were unaware of FGF23 levels. The commercially available assay used was a chemiluminescence immunoassay LIAISON® FGF23, for the in vitro quantitative determination of intact FGF23 polypeptide, in the LIAISON®-XL Analyzer (DiaSorin S.p.A., Saluggia, Italy). The assay has an analytical measurement range of 5–5000 pg/mL, a limit of detection of 5 pg/mL, and a limit of quantification of 6.5 pg/mL. Calibration was performed according to the manufacturer’s instructions and samples with concentrations exceeding the analytical measurement range were diluted 1:10 and reanalyzed according to the assay procedure. All samples were analyzed in a single batch. The intact FGF23 assay, which exclusively recognizes the complete and active FGF23 molecule, better reflects its biological activity compared to the carboxy-terminal form, which detects both the active form and C-terminal fragments and reflects the bone transcription of FGF23 [11]. The mean values of the quality control for FGF23 were as follows: Level 1 = 149.8 pg/mL and Level 2 = 570.5 pg/mL. The total variation coefficient for the assay, as indicated by the manufacturer, was 2.8% for both control levels. FGF23 was categorized based on the median of the overall sample (48.4 pg/mL) and into quartiles (Q1 ≤ 27.3 pg/mL; Q2 = 27.4–48.4, Q3 = 48.5–88.2; Q4 > 88.2). Two patients exhibited values above 1000 pg/mL. In both cases patients showed estimated glomerular filtration rate (eGFR) lower than 20 mL/min/1.73 m2. 2.3. Endpoints Three time-to-event outcomes were prespecified for the present analysis: all-cause mortality, the composite of all-cause death or WHF ED reconsultation, and the composite of all-cause death or HF hospitalization. WHF events comprised an ED reconsultation or an unplanned HF hospitalization for clinical deterioration requiring intravenous diuretics or inotropic therapy. Outcomes were ascertained from electronic medical records and outpatient follow-up. Investigators assessing endpoints were blinded to FGF23 concentrations. 2.4. Statistical Analysis Continuous data were summarized using mean ± SD for approximately normally distributed variables and median (25th–75th percentile) otherwise; categorical data are reported as counts and percentages. Because no clinically established FGF23 cutoff exists in AHF, the biomarker was examined primarily on a continuous scale and, for descriptive purposes, according to its median (48.4 pg/mL) and quartiles. Between-group comparisons used ANOVA or the Kruskal–Wallis test for continuous variables, as appropriate, and the χ2 test for categorical variables. Incidence rates (IR) for clinical outcomes were calculated as the number of first events divided by the total accrued person-time at risk in years and expressed as the number of events per 100 person-years (P-Y), with 95% confidence intervals based on the Poisson distribution Associations between the exposure (FGF23) and time-to-event outcomes were estimated using multivariable Cox proportional hazards regression. Potential confounders were identified a priori on clinical grounds and from previous evidence rather than according to their univariable associations with the outcomes. Model reduction was subsequently performed using backward elimination with p < 0.20 as the retention criterion. Variables reflecting clinical congestion were kept in the models irrespective of their statistical contribution to ensure adjustment for the burden of volume overload. The resulting adjustment set comprised age, sex, previous HF, ischemic heart disease, atrial fibrillation at presentation, systolic blood pressure, heart rate, Barthel index, orthopnea, paroxysmal nocturnal dyspnea, third heart sound, jugular venous distension, pleural effusion, peripheral edema, hemoglobin, NT-proBNP, and creatinine. Collinearity by Variance Inflation Factors (VIF) was verified for all predictors (mean VIF = 1.15; maximum = 1.34), ruling out multicollinearity. In complementary analyses, multivariable fractional polynomials (MFP) were used to test for non-linear functional forms. A retention threshold of p < 0.20 was used during MFP variable selection, which asymptotically approximates the Akaike Information Criterion (AIC) and prevents underfitting and coefficient bias; FGF23 was forced into all models. The proportional hazards assumption was verified via Schoenfeld residual tests (global p > 0.05 across all endpoints). Incremental predictive performance of FGF23 beyond the base multivariable model was assessed using Harrell’s C-concordance statistic and likelihood-ratio tests. In addition, the Integrated Discrimination Improvement (IDI) and category-free (continuous) Net Reclassification Improvement (NRI) were calculated with corresponding 95% confidence intervals (95% CI) to assess discrimination slopes and reclassification performance [12]. Estimates of risk were presented as a continuous variable after logarithmic transformation with base 2 (log2 FGF23). In this formulation, hazard ratios (HRs) directly reflect the relative risk associated with a doubling of circulating FGF23 levels. For standardized comparison across biomarkers, effect estimates were also expressed per 1-SD increase in log2 (“FGF23”). To assess potential selection bias arising from unavailable biomarker samples (n = 129, 16.9% of the 764 cohort participants, due to insufficient plasma volume or hemolysis), a secondary sensitivity analysis was conducted using Multiple Imputation by Chained Equations (MICE). Under the assumption of missing at random (MAR), log2-transformed FGF23 was imputed across 10 datasets using linear regression conditional on all baseline covariates as well as the survival time and event indicators [13], Cox proportional hazards models were fitted across imputed datasets and combined using Rubin’s rules. Complete-case analysis (N = 635) was retained as the primary analytical framework. To address the competing risk of all-cause death for nonfatal outcomes (WHF ED reconsultation and HF hospitalization), both cause-specific Cox models (censoring at competing death to evaluate etiological HRs) and Fine-Gray subdistribution hazard models (accounting for competing death to evaluate associations with the cumulative incidence function) were fitted. In the analysis stratified by subgroups, renal function was evaluated as continuous eGFR using the CKD-EPI formula and classified into three clinically established strata: preserved/mild impairment (eGFR ≥ 60 mL/min/1.73 m2), moderate impairment (eGFR 30–59.9 mL/min/1.73 m2), and severe impairment (eGFR < 30 mL/min/1.73 m2). Potential effect modification between FGF23 and renal function was formally tested by incorporating multiplicative cross-product terms in multivariable flexible parametric models, evaluated via 2-degree-of-freedom Wald tests. Stratum-specific adjusted HRs and 95% CI were obtained through linear combinations. A two-sided p-value of <0.05 was considered the threshold for statistical significance. All analyses were conducted using STATA version 16.1 [Stata Statistical Software, Release 16 (2019); StataCorp LP, College Station, TX, USA]. 3. Results 3.1. Baseline Characteristics of the Whole Sample Stratified by the FGF23 Median The mean ± SD age of the population was 82.4 ± 10 years, and 339 (53.4%) participants were women. A prior history of hypertension was highly prevalent (84.1%), and 401 patients (63.1%) had a previous diagnosis of HF. The median (p25% to p75%) creatinine, hemoglobin and NT-proBNP were 1.2 mg/dL (0.9–1.6), 12 g/dL (10.6–13.4), and 4207 pg/mL (2280–8421), respectively. Regarding exposure, the median (p25% to p75%) of FGF23 was 48.4 pg/mL (27.3–88.2). The baseline clinical characteristics of the study cohort stratified according to median FGF23 levels are summarized in Table 1. Patients with FGF23 concentrations above the median showed higher rates of CKD and ischemic heart disease. These individuals also showed a greater burden of fluid overload, with more frequent pleural effusion and jugular venous distension, and exhibited lower left ventricular ejection fraction (LVEF). Also, those above the median showed higher NT-proBNP and lower eGFR and hemoglobin. Overall, these patients had a worse baseline risk profile, reflected by a higher MEESSI risk score. Similar differences were also found when assessing FGF23 quartiles (Table A2). 3.2. Adverse Clinical Events At a median follow-up of 380 days (25th–75th percentile, 126–456), 216 patients (34.0%) died from any cause, 233 (36.7%) experienced a non-fatal WHF ED reconsultation (median time to event: 29.0 days, IQR: 16.0–47.0), and 152 (23.9%) had a non-fatal HF hospitalization (median time to event: 27.5 days, IQR: 15.0–42.5). At median follow-ups of 362 days (44–442) and 354 days (40–435), 351 patients (55.3%) experienced the composite endpoint of death/WHF ED reconsultation, and 295 patients (46.5%) experienced death/HF hospitalization, respectively. The corresponding incidence rates per 100 person-years were 39 (95% CI, 35–45) for all-cause mortality, 78 (95% CI, 70–87) for death/WHF ED reconsultation, and 66 (95% CI, 59–74) for death/HF hospitalization. 3.3. Relationship Between FGF23 and Adverse Clinical Events Patients with FGF23 concentrations above the median (>48.4 pg/mL) exhibited higher incidence rates across all three adverse outcomes (Table 1). When stratified by quartiles, individuals in the highest quartile (≥88.2 pg/mL) had nearly a two-fold increase in event rates compared with those in the lowest quartile (≤27.2 pg/mL) (Table A3). Kaplan–Meier analyses showed a consistent pattern across all three endpoints. Patients with FGF23 concentrations above the median had a significantly higher cumulative risk of all-cause mortality, death/WHF ED reconsultation, and death/HF hospitalization, with curves diverging after 2–3 months (Figure 1; all p-values < 0.05). Likewise, Kaplan–Meier curves stratified by FGF23 quartiles displayed a clear, graded association, with progressively worse outcomes across increasing quartiles (Figure 2; all p-values < 0.05). In the overall cohort, multivariable analyses showed a consistent and independent association between higher FGF23 concentrations and increased risk of all three endpoints (all p-values < 0.05). When modeled as a continuous variable using fractional polynomials, FGF23 exhibited a non-linear association with all clinical outcomes. For all-cause mortality (Figure 3a), we found a progressive and nearly linear increase in hazard above 40–50 pg/mL. At higher concentrations, the risk increased steadily, reaching HRs approaching ~1.7–1.8 at the upper range of the distribution (p = 0.001). Each doubling of FGF23 was independently associated with a 17% higher risk of all-cause mortality (adjusted HR: 1.17, 95% CI: 1.05–1.31; p = 0.004; equivalent to adjusted HR: 1.24, 95% CI: 1.07–1.43 per 1-SD increase). For death/WHF ED reconsultation (Figure 3b) the adjusted HRs rose sharply from lower until values about 50–60 pg/mL with a more gradual increase at higher concentrations (p = 0.001). A similar pattern was observed for death/HF hospitalization (Figure 3c), where the association followed a comparable trajectory, with an early steep rise and subsequent attenuation of the slope, although risk continued to increase across the full range of FGF23 values (p = 0.007). Similarly, each doubling of FGF23 independently predicted an increased hazard of death/WHF ED reconsultation (adjusted HR: 1.14, 95% CI: 1.05–1.24; p = 0.002; adjusted HR: 1.20, 95% CI: 1.07–1.34 per 1-SD increase) and death/HF hospitalization (adjusted HR: 1.12, 95% CI: 1.02–1.22; p = 0.013; adjusted HR: 1.17, 95% CI: 1.03–1.31 per 1-SD increase). Risk estimates across median, quartiles, per doubling (log2 scale), and per 1-SD increase in log2 are presented in Table 2. As a sensitivity analysis, we additionally fitted a clinically defined fully saturated model including all candidate covariates simultaneously, without variable elimination. The adjusted effect estimates for FGF23 remained virtually unchanged compared with those obtained using the MFP-selected model, supporting the robustness of the association irrespective of the variable-selection strategy (Table A4). In a sensitivity analysis evaluating the full registry cohort (N = 764) using multiple imputations by chained equations (10 imputed datasets) to account for the 129 patients with unmeasured biomarker aliquots, the independent prognostic association of circulating FGF23 remained fully consistent with primary complete-case models. After adjusting for all covariates, each doubling of FGF23 was independently associated with an increased hazard of all-cause mortality (adjusted HR: 1.18, 95% CI: 1.06–1.31; p = 0.003), death/WHF ED reconsultation (adjusted HR: 1.12, 95% CI: 1.03–1.22; p = 0.006), and death/HF hospitalization (adjusted HR: 1.11, 95% CI: 1.02–1.21; p = 0.014). When evaluating the incremental value of FGF23, our findings show that incorporating circulating FGF23 results in statistically significant improvements in overall model fit (by likelihood-ratio tests) and IDI across all three clinical endpoints, alongside significant net reclassification for all-cause mortality and ED reconsultations (Table 3). 3.4. Association of FGF23 with Adverse Clinical Events Stratified by Subgroups As exploratory analyses, we assessed the association between FGF23 and age, sex and kidney function subgroups. Risk estimates among subgroups are presented in Table 4. 3.4.1. Age When stratified by age, the association between FGF23 and adverse outcomes was qualitatively similar in patients aged <80 and ≥80 years across all endpoints (all-cause mortality, death/HF hospitalization and death/WHF ED reconsultation). In both age groups, higher FGF23 concentrations were associated with a progressive increase in risk, following a comparable non-linear pattern characterized by an early rise and subsequent attenuation at higher levels (Figure 4a). Formal interaction testing showed no evidence of effect modification by age for any outcome (all-cause mortality: p for interaction = 0.274; death/WHF ED reconsultation: p for interaction = 0.801; death/HF hospitalization: p for interaction = 0.984), indicating no significant heterogeneity across age categories. 3.4.2. Sex When stratified by sex, the association between FGF23 and adverse outcomes was broadly similar in women and men across all endpoints (Figure 4b). Formal interaction testing showed no evidence of effect modification by sex for any outcome (all-cause mortality: p for interaction = 0.424; death/WHF ED reconsultation: p for interaction = 0.745; death/HF hospitalization: p for interaction = 0.613), indicating no significant heterogeneity between women and men. 3.4.3. Kidney Function When stratified by baseline kidney function, the association between FGF23 and adverse outcomes was qualitatively consistent across eGFR categories for all endpoints (all-cause mortality, death/WHF ED reconsultation, and death/HF hospitalization; Figure 5a–c). Higher FGF23 concentrations were associated with a progressive increase in risk across all strata, with a similar non-linear pattern characterized by an early rise followed by a more gradual increase at higher levels. Although the slope of the association appeared numerically steeper among patients with more advanced renal dysfunction (eGFR < 30 mL/min/1.73 m2), formal interaction testing showed no evidence of statistically significant interaction by renal function for any of the outcomes (all-cause mortality: p for interaction = 0.314; death/WHF ED reconsultation: p for interaction = 0.142; death/HF hospitalization: p for interaction = 0.463), indicating no significant heterogeneity across eGFR categories. 3.5. FGF23 and Risk of Isolated Endpoints When evaluated as isolated endpoints, each doubling of circulating FGF23 was independently associated with all-cause mortality (adjusted HR 1.17, 95% CI: 1.05–1.31; p = 0.004). When nonfatal outcomes were evaluated separately accounting for the competing risk of death, the multivariable-adjusted associations were modest and did not reach statistical significance: the Fine–Gray subdistribution hazard ratio (sHR) was 1.08 (95% CI: 0.97–1.20; p = 0.141) for WHF ED reconsultation and 1.07 (95% CI: 0.95–1.21; p = 0.285) for HF hospitalization. Cause-specific Cox models yielded concordant estimates (csHR 1.09; p = 0.123 and csHR 1.07; p = 0.298, respectively). 4. Discussion In this post hoc biomarker analysis of the EAHFE registry, higher circulating FGF23 levels were independently associated with an increased risk of long-term adverse outcomes in patients with WHF attended in the ED. Interestingly, higher FGF23 was associated not only with mortality but also with the composite outcome of death or WHF ED reconsultation during follow-up. These associations remained consistent after comprehensive adjustment for key confounders, including age, sex, and kidney function, and across clinically relevant subgroups, including kidney dysfunction. These findings, consistent with previous evidence [14,15], further support an association between FGF23 and adverse clinical outcomes in HF, especially mortality, and suggest its potential value as a prognostic biomarker in this high-risk setting. Although the absolute increase in the C-statistic was small, this is expected when evaluating a novel biomarker against a well-performing multivariable base model. While the C-statistic is often relatively insensitive to the addition of new predictors in multivariable models [16], the observed improvements in likelihood-ratio tests and IDI across the three endpoints, together with significant net reclassification for all-cause mortality and death/WHF ED reconsultation, provide statistical evidence of incremental prognostic performance of FGF23 beyond established cardiorenal risk factors. However, these findings should not be interpreted as evidence of clinical utility, which requires independent external validation and assessment of its impact on clinical decision-making. Importantly, when nonfatal outcomes were considered separately, the associations with FGF23 were modest and did not reach statistical significance. Fine–Gray subdistribution hazard models, which characterize the association with the cumulative incidence of nonfatal events in the presence of competing death, and cause-specific Cox models, which estimate the instantaneous event rate among individuals remaining at risk, yielded concordant results. These findings suggest that the associations observed for the composite endpoints may be driven predominantly by mortality and should not be interpreted as evidence of a significant independent association between FGF23 and isolated nonfatal events. 4.1. FGF23 and WHF: Prior Evidence Our findings extend previous evidence linking FGF23 to adverse outcomes in cardiovascular and cardiorenal populations [17]. Beyond its role as a biomarker of mineral metabolism, FGF23 has been increasingly recognized as a key player in several pathophysiological mechanisms involved in HF progression [14]. Experimental studies have demonstrated that FGF23 directly induces left ventricular hypertrophy [18], and promotes sodium retention and plasma volume expansion through enhanced distal tubular sodium reabsorption [19], thereby providing a plausible mechanistic link with congestion and WHF. In chronic HF, a substantial body of evidence has consistently shown that higher FGF23 concentrations are associated with greater disease severity and worse prognosis. Elevated levels correlate with poorer functional status, impaired renal function, and increased mortality risk [20], and have been identified as an independent predictor of death in patients with HFrEF [21]. This relationship has also been described in HFpEF [22,23], where FGF23 is associated with reduced functional capacity and adverse outcomes, supporting its role across the full spectrum of HF phenotypes. Evidence is also available in patients with AHF and WHF. In patients with new-onset or WHF, elevated FGF23 levels have been associated with greater congestion, poorer optimization of guideline-directed medical therapy, and an increased risk of mortality and HF-related hospitalizations [8]. Similarly, in hospitalized acute HF cohorts, early measurement of intact FGF23 has been associated with disease severity and increased one-year mortality, with prognostic performance comparable to established risk prediction models [24]. In more severe forms of decompensation, including cardiogenic shock, higher FGF23 levels have also been linked to short-term mortality [25]. In this context, our study provides additional insights. First, we found that FGF23 measured in patients evaluated in the emergency department, rather than exclusively in selected hospitalized cohorts, offers prognostic information from the earliest stages of acute decompensation. This finding is particularly important, as early risk stratification in the emergency setting remains challenging and may help guide decisions regarding patient disposition and intensity of follow-up [26]. Second, our population was very elderly, a group frequently underrepresented in biomarker studies [27] despite representing a large proportion of patients presenting with acute HF in routine clinical practice. Third, our cohort included a broad spectrum of HF phenotypes, reflecting the heterogeneity encountered in routine clinical practice. Finally, the observed association with WHF and HF-related admissions is especially relevant, as it suggests that FGF23 may identify patients at risk of subsequent clinical instability beyond its previously established relationship with mortality. Altogether, our findings support the concept that FGF23 reflects not only cardiorenal dysfunction but also congestion-related pathophysiological pathways driving recurrent decompensation. From a mechanistic perspective, the clinical associations observed in our study are also supported by a solid body of experimental evidence. FGF23 is a phosphaturic hormone central to mineral metabolism, regulating phosphate homeostasis and suppressing active vitamin D synthesis [14,28], by inhibiting renal phosphate reabsorption and downregulating 1α-hydroxylase while upregulating 24-hydroxylase [29]. Its canonical actions occur in the kidney and parathyroid glands via FGFR–klotho signaling [29]. Beyond its role in mineral metabolism, FGF23 has emerged as a direct modulator of CV damage [14]. At the myocardial level, klotho-independent activation of FGFR4 triggers the PLCγ–calcineurin–NFAT pathway [30], increasing intracellular calcium and promoting hypertrophic gene expression [31]. Concurrently, it activates pro-fibrotic (β-catenin, TGF-β/Smad) and pro-inflammatory pathways, enhancing extracellular matrix production and cytokine release, thereby contributing to myocardial fibrosis, inflammation, and adverse remodeling [7,32]. In parallel, FGF23 may be integrated within the congestion-inflammation axis that characterizes most episodes of WHF [33]. Volume expansion promotes systemic inflammatory activation [34], with cytokines such as TNF-α, IL-1β, and IL-6, which have been shown to induce FGF23 expression [7]. Thus, elevated FGF23 levels may further amplify myocardial remodeling, perpetuating cardiac dysfunction and congestion [29], suggesting an active role within this vicious cycle. Furthermore, beyond the myocardial effects, FGF23 may exert indirect cardiovascular actions through neurohormonal activation [7]. By reducing calcitriol levels, FGF23 disinhibits renin expression and activates the renin–angiotensin–aldosterone system, resulting in increased angiotensin II and aldosterone activity [7,30]. This neurohormonal cascade further amplifies key pathological processes, including hypertrophy, fibrosis, inflammation, and hypertension [30]. Collectively, these mechanisms support the concept that FGF23 is not merely a biomarker of disease severity but may also function as an active mediator of CV damage [34]. In summary, our results support the potential role of FGF23 as an early prognostic biomarker in acute and WHF, particularly in elderly patients evaluated in the ER. Future studies should determine whether incorporating FGF23 into multimarker strategies improves risk stratification and whether serial measurements may better capture the dynamic changes associated with decompensation and treatment. 4.2. Potential Clinical Implications From a clinical perspective, these results suggest the potential utility of FGF23 as a biomarker for risk stratification in patients with WHF. Its association with adverse outcomes across multiple subgroups supports its potential value for prognostic assessment, although further studies and external validation are needed before its incorporation into clinical risk models. Moreover, previously proposed links between FGF23 and cardiac remodeling raise the hypothesis that FGF23 or its downstream signaling pathways could represent potential therapeutic targets. However, our observational findings do not establish a causal or mechanistic role for FGF23, and this possibility requires further investigation in experimental and interventional studies. 4.3. Limitations There are several limitations worthy of mention. First, the possibility of residual confounding cannot be excluded, as some relevant variables were not measured. In addition, echocardiographic parameters other than LVEF, as well as klotho and other mineral-metabolism variables were not collected. Therefore, despite adjustment for measured markers of renal function, residual confounding related to renal and mineral metabolism cannot be excluded, and our findings should not be interpreted as establishing a renal-function-independent or mechanistic role of FGF23. Furthermore, given the limited number of participants with eGFR < 30 mL/min/1.73 m2, findings in this subgroup should be considered exploratory and interpreted with caution. Second, the observed improvements in discrimination and reclassification indices provide statistical evidence of incremental prognostic performance within this derivation cohort. However, these findings do not establish clinical utility, which requires independent external validation and assessment of their impact on clinical decision-making. Third, the available data do not allow us to clarify the underlying biological mechanisms that may explain the observed associations. Fourth, the median concentration of 48.4 pg/mL used in descriptive summaries was cohort-derived and does not represent an established biological or clinical decision threshold. Fifth, although our primary findings rely on continuous modeling across the full biomarker spectrum, defining and externally validating clinically actionable thresholds will be required for practical bedside implementation. Sixth, this study included an older Spanish cohort, which may limit the generalizability of our findings to other HF populations. Lastly, future studies, including more detailed echocardiographic evaluation and external validation in independent cohorts, are needed to validate these findings and better understand their potential clinical implications. 5. Conclusions Elevated circulating FGF23 levels are independently associated with an increased risk of long-term adverse outcomes in patients with WHF. These findings support the prognostic relevance of FGF23 in patients with WHF. External validation in independent populations is warranted, and whether FGF23 has a causal pathophysiological role or represents a potential therapeutic target requires further investigation in experimental and interventional studies.