Diagnostic and prognostic implications of growth differentiation factor 15 in heart failure with preserved ejection fraction: a systematic review and meta-analysis

Introduction

Heart failure with preserved ejection fraction (HFpEF) has become a significant public health concern, affecting approximately 50% of the global heart failure (HF) population (Borlaug & Paulus, 2011; Borlaug et al., 2023). The 5-year mortality and rehospitalisation rates for this condition are comparable to those observed in HF patients with reduced ejection fraction (HFrEF) (Borlaug et al., 2023). However, the underlying cause of HFpEF is not fully understood, leading to challenges in accurately diagnosing the condition and selecting a suitable treatment (Borlaug, 2020; Nagueh, 2021; Mishra & Kass, 2021). HFpEF remains a challenging condition to diagnose, with conventional biomarkers such as natriuretic peptide often failing to meet diagnostic thresholds, particularly in those with obesity-related HFpEF, despite elevated left ventricular (LV) filling pressures at rest or during exercise (Santhanakrishnan et al., 2012; Borlaug et al., 2023). This limitation highlights the need for novel biomarkers that better reflect the pathophysiology of HFpEF.

Growth differentiation factor-15 (GDF-15), a novel biomarker that has the potential to be a valuable diagnostic and prognostic tool in HF, has been associated with chronic inflammation and oxidative stress (di Candia et al., 2021; Morfino et al., 2022). GDF-15 appears to more effectively represent this ongoing low-grade inflammatory condition than other conventional biomarkers (Chan et al., 2016), as HFpEF is a phenotype that is primarily associated with a chronic inflammatory state (Shah et al., 2016). Prior meta-analysis demonstrated that GDF-15 is a promising biomarker for diagnosing HFpEF, with a pooled area under the curve (AUC) of 0.88, indicating strong diagnostic accuracy (Wang, Gu & Guo, 2023). However, the study was limited by a small sample size (1,550 patients), and it did not assess the prognostic value of GDF-15 in HFpEF.

Examining a biomarker comprehensively provides a strong means of contextualising the biomarker and enhancing the knowledge of disease-related pathways, as the utilisation of this approach can optimise clinical decision-making, thereby improving overall patient outcomes. Therefore, this systematic review and meta-analysis aimed to systematically analyse, from published literature, the diagnostic and prognostic value of GDF-15 in HFpEF.

Materials and Methods

This systematic review and meta-analysis were conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline (Page et al., 2021). The present study was registered in the international prospective register of systematic reviews (PROSPERO) under the protocol number CRD42024569609.

Results

Literature search results and quality assessment

The initial search identified 328 records, with 280 remaining after duplicate removal. A total of 17 reports underwent full-text review, but one was not retrieved, and four were excluded due to irrelevant outcomes or duplicate populations. Ultimately, 12 studies were included in the systematic review and meta-analysis (Fig. 1). All but one of the studies were assessed as being of good quality, as summarised in Table S1.

Diagnostic performance of GDF-15 in distinguishing HFpEF from controls with conventional biomarkers

GDF-15 levels were elevated in HFpEF patients compared to control groups across all studies analysed, with the exception of a study by Oyama et al. (2023) that compared the HFpEF group with a control group of individuals with atrial fibrillation (AF) and no HF (median GDF-15: 1,599.0 vs. 1,674.0 pg/mL). Within the obese population with HFpEF, the levels of GDF-15 were markedly higher than in the obese population with normal LV function (median GDF-15 662 vs. 451 pg/mL) (Baessler et al., 2012) (Tables S2, S3).

Figure 2A depicted that GDF-15 levels were significantly elevated in HFpEF compared to a control group of the non-HF population (MD 647.60 pg/mL; 95% CI [148.43–1,146.77]; p = 0.01; I² 98%). This result was deemed robust with a leave-one-out sensitivity analysis (Fig. 2B), as there was no change in the p-value when omitting any study. However, the MD was slightly higher when studies by Oyama et al. (2023) or Aulin et al. (2022) were excluded, indicating that these two studies contributed to narrowing the overall difference in the meta-analysis. Additionally, the heterogeneity remained unchanged, indicating that the observed variability across studies was not driven by the exclusion of any single study.

Four studies estimated the diagnostic performance of GDF-15 in identifying HFpEF from controls (non-HF population (Sinning et al., 2017), obese with normal LV function (Baessler et al., 2012), non-CAD and non-HF population (Santhanakrishnan et al., 2012), and population without HFpEF, dilated cardiomyopathy, and ischaemic cardiomyopathy (Jirak et al., 2020)). From the qualitative analysis, the overall diagnostic performance of GDF-15 in HFpEF appears to be slightly superior compared to NT-proBNP, as a conventional biomarker for HFpEF (Table S4). While the combination of GDF-15 and NT-proBNP resulted in a higher AUC (0.956; 95% CI [0.919–0.994]), this combined biomarker strategy was not statistically different from using GDF-15 or NT-proBNP alone (p = 0.31; p = 0.33, respectively) (Santhanakrishnan et al., 2012). Figure 3A depicted the pooled AUC of GDF-15 for HFpEF (0.82; 95% CI [0.72–0.91]; I² 84%), indicating an overall good diagnostic accuracy. Sensitivity analysis reveals a robustness in the result with unchanged statistical significance and direction of effect. However, there was a reduction in heterogeneity when omitting the study by Santhanakrishnan et al. (2012), indicating that this study was a major contributor to the observed heterogeneity (Fig. 3B).

GDF-15 and prognostic outcomes in patients with HFpEF

A total of six studies evaluated the prognostic implications of serum GDF-15 in relation to all-cause mortality, adverse cardiovascular events, and HF hospitalisation with a mean follow-up duration ranging from 12 to 112 months (Tables S5–S7) (Izumiya et al., 2014; Sinning et al., 2017; Mendez Fernandez et al., 2020; Oyama et al., 2023; Yin et al., 2023; Lyu et al., 2024). A prospective cohort reported that GDF-15 concentrations independently predict all-cause mortality in the Spanish population based on unadjusted multivariate analysis (Mendez Fernandez et al., 2020). Additionally, elevated serum GDF-15 levels were significantly associated with an increased risk of adverse cardiovascular events in HFpEF patients, underscoring its prognostic value in this population (Izumiya et al., 2014; Oyama et al., 2023).

From three studies included in the meta-analysis, elevated GDF-15 was significantly increased the hazard of all-cause mortality with a pooled HR of 1.46 (95% CI [1.30–1.62], p < 0.01; I² 0%) (Fig. 4A). The resulted did not differ when sensitivity analysis was performed (Fig. 4B). Two studies also evaluated the effect size of GDF-15 and HF hospitalisation (Yin et al., 2023; Lyu et al., 2024), with a pooled HR of 1.76 (95% CI [1.30–2.38]; p < 0.01; I² 0%) (Fig. 4C), indicating a significant association between elevated GDF-15 levels and increased risk of HF hospitalisation. The findings indicate that GDF-15 may serve as a valuable prognostic biomarker, especially for assessing all-cause mortality and HF hospitalisation risk in clinical settings.

Discussion

In this systematic review and meta-analysis, a total of 5,696 HFpEF patients were included from 28,193 participants. GDF-15 levels were significantly higher in the HFpEF patients compared to the control group of the non-HF population. GDF-15 has a good diagnostic accuracy in diagnosing HFpEF and appears to be slightly superior compared to conventional biomarkers. GDF-15 also demonstrates significant prognostic potential in HFpEF, as higher levels of GDF-15 are associated with an increased risk of all-cause mortality and HF hospitalisation.

HFpEF is a complex disease with interrelated pathophysiological mechanisms (Logeart, 2024). The condition’s prevalence has surged to the point of being considered an epidemic, with recent research highlighting that HFpEF encompasses a range of distinct clinical phenotypes. These phenotypes, although grouped under the same definition, exhibit diverse pathophysiological characteristics and mechanisms (Stoicescu et al., 2024). Over the last decade, several groups of investigators were able to obtain myocardial tissue from HFpEF patients, revealing specific alterations in myocardial structure, function, and intramyocardial signaling, which were relevant to the concentric LV remodeling and diastolic LV dysfunction characteristically observed in patients with HFpEF (van Heerebeek & Paulus, 2016). These alterations include increased myocardial fibrosis, heightened oxidative stress, endothelial dysfunction, and chronic low-grade inflammation—all of which are associated with disease progression in HFpEF. HFpEF is additionally marked by a sustained elevation in inflammatory biomarkers, including GDF-15. This cytokine is elevated due to systemic inflammation and oxidative stress, both of which drive myocardial remodeling and endothelial dysfunction in HFpEF. Inflammation may be a primary factor in the development and advancement of HFpEF and its multiple related comorbidities, and GDF-15 levels often reflect this ongoing cellular stress and inflammatory burden of HFpEF (Lewis et al., 2022; Pugliese et al., 2023).

Although NT-proBNP is a valuable biomarker in HF, it may have limitations in HFpEF due to confounding effects from prevalent comorbidities such as AF, obesity, and renal impairment. NT-proBNP levels in HFpEF patients can be deceptive. For example, AF may elevate NT-proBNP in a manner that is disproportionate to the severity of HF, whereas obesity has a tendency to lower NT-proBNP levels in patients with significant HFpEF-related pathology. In addition, renal impairment complicates the interpretation of NT-proBNP, as reduced clearance can result in artificially elevated levels, further complicating the clinical picture (Januzzi & Myhre, 2020). Our study showed that GDF-15 appears to be slightly superior compared to NT-proBNP in HFpEF based on the qualitative analysis of AUC.

On the quantitative assessment, our study also confirmed the overall good diagnostic performance of GDF-15 based on pooled AUC. However, population-based disparities may have influenced our quantitative findings, particularly in the sensitivity analysis, in which diminished heterogeneity was observed when we omitted Santhanakrishnan et al. (2012), who conducted a study in Singapore. Importantly, the remaining three studies were conducted in Germany, where European cohorts may exhibit more distinct HFpEF vs. non-HFpEF characteristics (Ingimarsdóttir et al., 2025), making GDF-15 a stronger discriminator with a higher AUC. These findings suggest that geographic and racial differences may influence the overall diagnostic accuracy of GDF-15 in HFpEF, emphasising the need for a population-based cut-off, further validating its reliability in diverse cohorts.

AF is one of the main comorbidities found in HFpEF, with a prevalence ranging from 40% to 60% (Fauchier, Bisson & Bodin, 2023). Our study showed that GDF-15 was higher in the population with AF without HF compared to the HFpEF population. AF is linked to systemic inflammation, oxidative stress, and structural alterations in the heart, perhaps resulting in increased levels of GDF-15 even in populations without HF. Moreover, AF frequently coexists with comorbidities such as hypertension, diabetes, and obesity, which exacerbate inflammatory and oxidative stress pathways, elevating GDF-15 levels (Sartipy et al., 2017; Zafrir et al., 2018). This particular mechanism may influence our result, as the comparator group with AF comorbidities may have had inherently high GDF-15 levels. This effect was observed in our sensitivity analysis, where omitting the studies by Oyama et al. (2023), Aulin et al. (2022) resulted in a narrowed difference in GDF-15 levels between HFpEF and non-HFpEF groups. Therefore, further studies are needed to determine the optimal cut-off to use GDF-15 in diagnosing HFpEF in the setting of the AF population.

GDF-15 is an attractive diagnostic marker for HF patients across their ejection fraction and could play an important role in diagnosis in specific HF patients with attenuated levels of BNP or NT-pro BNP. One such important group are those with accompanying comorbidities of obesity and metabolic syndrome (Madamanchi et al., 2014; Kozhuharov et al., 2019). Recent studies have shown that BNP and NT-pro BNP levels are compromised in obese patients. In the GUIDE-IT trial involving 873 HFrEF patients, NT pro-BNP levels were 59.0% (95% CI [39.5–83.5%]) lower in obese patients (≤1,000 pg/mL). Yet despite this, those with higher BMI were associated with a greater risk of adverse cardiovascular events, regardless of low or high NT pro-BNP levels (Felker et al., 2017). This finding was further confirmed in the Asian population through the Suita study, showing its cross-cultural validity (Sugisawa et al., 2010). The strongest predictors of low BNP levels found were high BMI, younger age, higher LVEF, and lower creatinine (Bachmann et al., 2021). Hence, in obese patients with HFpEF, diagnosis through the use of conventional cardiac biomarkers would be a unique challenge. Our study demonstrated that GDF-15 could be a potential biomarker for HFpEF diagnosis in this subgroup of patients, as GDF-15 were significantly higher in the obese HFpEF population than in the obese population with normal LV function. However, further studies are needed to verify this result, as only one of the studies included in our analysis contains this subgroup of population.

The incidence of HFpEF, in particular is still rising even while the overall prevalence of HF seems to be constant or even declining (Tsao et al., 2018). According to four community-based cohorts, the incidence rate of HFpEF is estimated to be 27 cases per 10,000 person-years, and it is anticipated that its prevalence is going to surpass that of HFrEF. HFpEF also has a comparably poor clinical outcome with HFrEF (Bhambhani et al., 2018). Our study has shown that elevated GDF-15 is significantly associated with all-cause mortality and HF hospitalisation, making it a valuable prognostic marker.

This meta-analysis evaluates both the diagnostic and prognostic value of GDF-15 in the HFpEF population. It involves a wide range of studies conducted in different countries and includes HFpEF patients in both the acute and chronic clinical settings. The majority of studies included, except one, were all of good quality and provided consistent results. However, several limitations should also be noted. There was apparent heterogeneity across all studies, which may have impacted the study results. One of the major sources of heterogeneity arises from the utilisation of various GDF-15 assay kits across studies, including research-grade Enzyme-linked immunosorbent assay (ELISA) kits (e.g., Biovendor, R&D Systems, Millipore) and the in-vitro diagnostic grade Roche Diagnostics Elecsys electrochemiluminescence immunoassay. Research-grade ELISA kits exhibit a coefficient of variation of roughly 10% to 20%, while the Roche Elecsys assay demonstrates a coefficient of variation of less than 4% (Karusheva et al., 2022). This disparity may result in variations in assay precision and measurement accuracy. Moreover, differences in HFpEF comorbidities and race among the study populations may have further contributed to study heterogeneity. However, due to limited data, we were unable to perform subgroup analysis or meta-regression. Future studies should aim to standardise GDF-15 measurement techniques to improve data consistency and reliability in meta-analysis. The second limitation is that only limited studies were included in the prognostic meta-analysis; therefore, the interpretation of these results requires caution. Additional research involving larger sample sizes is required to enhance the evidence in this domain.

Although our study has demonstrated the significance of GDF-15 in HFpEF diagnosis and prognosis, several challenges still need to be addressed before it can be widely implemented in clinical practice. First, an optimal cut-off value of GDF-15 for risk stratification still needs to be elucidated. Second, the effects of comorbidity, such as AF, on the diagnostic value of GDF-15 need to be considered. Third, our study was unable to assess the relationship between GDF-15 and specific hemodynamic phenotypes of HFpEF, as distinct profiles phenotype may present with differing levels of GDF-15, potentially reflecting varying underlying pathophysiology and severity. Therefore, large multi-centred prospective studies are still required for these questions. In the future, based on its extensive pathophysiological mechanism, GDF-15 also carries potential for therapeutic applications, and studies are currently ongoing on its effectiveness against cardiometabolic disease, non-alcoholic fatty liver disease, and obesity (Wang et al., 2021).

Conclusions

GDF-15 could serve as a promising biomarker in HFpEF, with the overall diagnostic performance appearing to be slightly superior compared to conventional biomarkers and also demonstrating significant prognostic potential in HFpEF. Higher levels of GDF-15 are linked to increased risk of all-cause mortality and HF hospitalisation.

Supplemental Information