C-reactive protein-to-albumin ratio and the mortality of patients with heart failure: a meta-analysis

Abstract

Background

The C-reactive protein-to-albumin ratio (CAR) reflects the interplay between systemic inflammation and nutritional status, and has recently been proposed as a prognostic biomarker in heart failure (HF). However, its association with mortality risk across acute decompensated HF (ADHF) and chronic HF (CHF) populations remains uncertain. This meta-analysis evaluated the relationship between baseline CAR and all-cause mortality in HF.

Methods

PubMed, Embase, Web of Science, CNKI, and Wanfang were searched for relevant longitudinal studies. Risk ratios (RRs) were pooled using a random-effects model accounting for heterogeneity. Prespecified subgroup and meta-regression analyses were performed to evaluate the influence of study characteristics.

Results

Twelve cohort studies involving 6,377 patients were included. High CAR was associated with a significantly increased risk of mortality (RR = 2.34, 95% CI 1.86–2.93), with moderate heterogeneity (I² = 66%). Notably, the association was weaker in prospective studies (RR = 1.45) compared with retrospective studies (RR = 2.50). Subgroup findings were consistent across regions (Asian: RR = 2.62; Western: RR = 1.87), HF phenotype (ADHF: RR = 2.15; CHF: RR = 2.47), age (< 65 years: RR = 2.36; ≥65 years: RR = 2.33), CAR cutoff (< 0.5 mg/g: RR = 2.60; ≥0.5 mg/g: RR = 2.16), follow-up duration (< 30 months: RR = 2.20; ≥30 months: RR = 2.45), and analytic model (univariate: RR = 2.71; multivariate: RR = 1.99). Meta-regression identified no significant moderators.

Conclusions

Elevated baseline CAR is consistently associated with higher mortality risk in patients with HF. However, the strength of this association appears lower in prospective studies, suggesting that the overall pooled estimate may be influenced by biases inherent to retrospective designs. In addition, the clinical utility of CAR for risk stratification of patients with HF requires further validation in well-designed prospective studies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-026-05983-7.

Introduction

Heart failure (HF) is a major global health burden affecting more than 64 million individuals worldwide, with rising prevalence driven by population aging, improved survival from acute cardiac events, and increasing rates of hypertension, diabetes, and obesity [1, 2]. Despite advances in pharmacologic and device-based therapies, the prognosis of HF remains poor, with annual mortality rates ranging from 10% to 30% depending on disease severity and clinical setting [3]. Identifying simple, accessible biomarkers that can improve early risk stratification is therefore of substantial clinical importance [4]. Among emerging predictors, inflammatory and nutritional markers have received increasing attention, as systemic inflammation, malnutrition, and catabolic imbalance contribute to HF progression and adverse outcomes [5, 6].

The C-reactive protein-to-albumin ratio (CAR) is an inflammation–nutrition composite index calculated as C-reactive protein (CRP) divided by serum albumin [7]. Clinically, elevated CAR reflects the coexistence of heightened inflammatory activity and impaired nutritional or hepatic synthetic function [8]—two processes strongly associated with HF severity, congestion, frailty, and mortality risk [9, 10]. Mechanistically, higher CAR may mirror cytokine-driven myocardial injury, endothelial dysfunction, and protein-energy wasting, thereby integrating multiple prognostic pathways into a single measurable parameter [7, 11]. Although several observational studies have examined the prognostic significance of CAR in HF, findings remain heterogeneous [12–23]. Therefore, this meta-analysis aimed to quantitatively assess the association between baseline CAR and all-cause mortality in HF and to explore study-level factors that may influence this relationship.

Methods

The conduct and reporting of this meta-analysis adhered to the PRISMA 2020 recommendations [24] and relevant guidance from the Cochrane Handbook [25], encompassing protocol development, data collection, statistical synthesis, and presentation of findings. The protocol was preregistered with PROSPERO (CRD420251247153). Clinical trial number: not applicable.

Search strategy

Eligible studies were located through an extensive literature search of PubMed, Embase, Web of Science, Wanfang, and the China National Knowledge Infrastructure (CNKI), employing a broad set of predefined search terms, which included: (1) “C-reactive protein” OR “C reactive protein” OR “CRP”; (2) “albumin”; (3) “rate” OR “ratio”; and (4) “heart failure” OR “cardiac failure” OR “cardiac dysfunction” OR “cardiac insufficiency”. The search was restricted to human research and full-text articles published in English or Chinese in peer-reviewed journals. To supplement the electronic search, reference lists of pertinent original studies and reviews were manually examined to identify additional eligible publications. All databases were searched from their inception through November 10, 2025. Detailed search strategies for each database are provided in Supplemental File 1.

Eligibility criteria

Study eligibility was defined according to the PICOS framework:

• P (Population): Adult patients (≥ 18 years) diagnosed with acute decompensated HF (ADHF), including de novo ADHF or acute decompensation of chronic HF (CHF); or CHF in the stable or ambulatory setting. HF diagnosis based on clinical, laboratory, imaging, or guideline-defined criteria as reported by the original studies.

• I (Exposure): CAR measured at baseline, such as at hospital admission or at study inclusion. CAR was calculated as: CRP (mg/dL) / albumin (g/dL). Patients with high CAR were considered as exposure, with the methods for determining the cutoffs of CAR consistent with those in the original studies.

• Comparator (C): Patients with low CAR values, serving as the reference group in categorical comparisons.

• Outcomes (O): All-cause mortality during follow-up, compared between HF patients with high vs. low CAR at baseline.

• S (Study Design): Longitudinal observational studies, including cohort studies, nested case–control studies, and post-hoc analyses of randomized or prospective cohorts.

Studies were excluded if they did not enroll patients with HF or did not evaluate CAR as a categorized variable. Studies reporting CAR only as a continuous measure or lacking comparisons between high and low CAR groups were excluded. We also excluded studies that did not report mortality or survival outcomes, or did not provide sufficient data to extract effect estimates. Cross-sectional studies, case reports, case series, editorials, reviews, conference abstracts, and non–peer-reviewed publications were not eligible. Studies focusing primarily on valvular heart disease, surgical populations, or other non-HF cohorts were excluded unless HF was the defined underlying diagnosis. Non-human studies and studies involving pediatric populations were also excluded. If multiple articles were derived from the same underlying cohort, we included only the version with the most extensive dataset or the largest cohort.

Quality assessment

Two reviewers independently performed the literature search, study screening, quality appraisal, and data extraction, with any disagreements resolved through discussion and consensus between with the authors. Study quality was assessed using the Newcastle–Ottawa Scale (NOS) [26], which evaluates selection, adjustment for confounding, and outcome assessment, yielding scores from 1 to 9, with higher scores indicating better methodological rigor. Studies scoring ≥ 7 were classified as high quality.

Data extraction

Data extraction was conducted independently by two reviewers using a standardized, predesigned data extraction form in Microsoft Excel. Any discrepancies between the two reviewers were resolved through discussion and consensus between the two authors. Extracted data included study-level information (author, publication year, design, and country), patient characteristics (diagnosis, sample size, mean age, sex distribution, proportion of patients with HF with reduced ejection fraction [HFrEF], and proportion of patients with ischemic HF), exposure characteristics (timing of CAR evaluation, methods to determine the cutoff of CAR, and cutoff values for defining a high CAR), follow-up durations, number of patients who died during follow-up, and covariates adjusted for in the estimation of the association between CAR at baseline and mortality risk during follow-up. To ensure consistency in CAR calculation, CRP values reported in different units were standardized prior to analysis. Specifically, CRP values reported in mg/L were converted to mg/dL (1 mg/dL = 10 mg/L), while albumin values were consistently expressed in g/dL. Accordingly, CAR was uniformly calculated or interpreted as mg/g across all studies.

Statistical analysis

The influence of baseline CAR on mortality risk of patients with HF was summarized as risk ratio (RR) and corresponding 95% confidence intervals (CIs), compared between patients with high vs. low CAR at baseline. When multiple models were reported, the most fully adjusted effect estimates were extracted; otherwise, unadjusted estimates were used. When necessary, RRs and their standard errors were derived from reported CIs or p-values and subsequently log-transformed to stabilize variance and approximate normality [25]. Between-study heterogeneity was examined using the Cochrane Q statistic and the I² metric [27], with thresholds of < 25%, 25–75%, and > 75% interpreted as low, moderate, and high heterogeneity, respectively. Pooled estimates were generated using a random-effects model to account for underlying variability across studies [25]. Robustness of the overall effect was evaluated through leave-one-out sensitivity analyses [28]. In addition, sensitivity analysis limited to studies with the adjustment of B-type natriuretic peptide (BNP) or N-terminal pro-BNP (NT-pro-BNP) was also performed. Prespecified subgroup analyses were performed to explore whether study characteristics influenced the observed associations. Stratifications included study country (Asian vs. Western countries), study design (prospective vs. retrospective), status of HF (ADHF vs. CHF), mean ages of the patients (< 65 years vs. ≥ 65 years), cutoff values of CAR, follow-up durations, and NOS scores. For continuous moderators, subgroups were formed using median values as cutoff points to balance the numbers of studies available in each subgroup. Additionally, univariate meta-regression was performed to assess the relationship between study-level variables (e.g., sample size, mean age, proportion of men, proportion of HFrEF, proportion of ischemic HF, cutoff of CAR, follow-up duration, and NOS score) and the effect estimates [25]. Potential publication bias was evaluated using funnel plot symmetry, visual inspection, and Egger’s regression test [29]. A two-sided p-value < 0.05 was considered statistically significant. All analyses were performed using RevMan (version 5.3; Cochrane Collaboration, Oxford, UK) and Stata (version 17.0; StataCorp, College Station, TX, USA).

Results

Study selection

Figure 1 depicts the study selection workflow. A total of 132 records were retrieved from the five databases, of which 29 duplicates were removed. Screening of titles and abstracts resulted in the exclusion of 77 records that did not meet the eligibility criteria. The full texts of the remaining 26 articles were evaluated independently by two reviewers, and 14 were excluded for reasons shown in Fig. 1. Ultimately, 12 studies met all criteria and were included in the quantitative synthesis [12–23].

Fig. 1.

Flowchart of database search and study inclusion

Study characteristics

The key characteristics of the included studies are summarized in Table 1. A total of 10 retrospective [12, 13, 15–20, 22, 23] and 2 prospective cohort [14, 21] studies published between 2020 and 2025 were included, representing diverse geographical regions including China, Turkey, Brazil, Poland, and Japan. The studies encompassed both ADHF and CHF populations: four studies exclusively enrolled hospitalized patients with ADHF [18, 20, 21, 23], while the remaining eight studies included either hospitalized or outpatient cohorts with CHF [12–17, 19, 22]. Sample sizes varied substantially, ranging from 69 to 1,272, yielding a combined study population of 6,377 patients. The mean age of participants ranged from 53.0 to 81.0 years, and the proportion of men varied from 44.8% to 82.0% across studies. The reported prevalence of HFrEF varied widely (34.8% to 100%), reflecting differences in cohort composition, while the proportion of ischemic etiology ranged from 0% to 100% among studies reporting this information. All studies measured the CAR at baseline—either at admission for hospitalized ADHF or CHF patients, or at study inclusion for outpatient CHF cohorts. Cutoff values for CAR varied considerably, ranging from 0.12 to 2.78 mg/g. Most cutoffs were determined using ROC curve analysis [12, 15, 17–19, 22, 23], while a few used tertile (T3:T1) [13, 16] or quartile (Q4:Q1) [20, 21] of CAR, or cutoff value defined by previous study [14]. Follow-up durations ranged widely from within hospitalization to 92 months, and a total of 1,886 (29.6%) patients died during follow-up. Multivariate analysis was performed in six studies [13, 17–21], with the adjustment of clinical, laboratory, and echocardiographic variables to a varying degree, while the other six studies reported results of univariate analysis only [12, 14–16, 22, 23].

Table 1.

Characteristics of the included studies

Study	Country	Design	Diagnosis	Sample size	Mean age (years)	Men (%)	HFrEF (%)	Ischemic (%)	Timing of CAR measurements	Methods to determine cutoff of CAR	Cutoff value of CAR (mg/g)	Follow-up duration (months)	No. of patients died	Variables adjusted or matched
Su 2020 [12]	China	RC	Hospitalized CHF patients	125	67.1	44.8	NR	NR	At admission	ROC curve analysis	0.38	18	50	None
Çinier 2021 [13]	Turkey	RC	Hospitalized CHF patients for ICD implantation	1011	63	80.4	100	76.5	At admission	T3:T1	0.38	38	147	Age, sex, indication for ICD implantation (primary/secondary prevention), device types, WBC, lymphocytes, urea, FBG, and LVEF
Lima 2022 [14]	Brazil	PC	Outpatients with CHF	77	59.9	59.7	100	NR	At study inclusion	Previous study derived	1.2	12	5	None
Ren 2022 [15]	China	RC	Hospitalized CHF patients	69	71.3	56.5	34.8	31.9	At admission	ROC curve analysis	0.34	24	53	None
Feng 2023 [16]	China	RC	Hospitalized non-ischemic CHF patients	1250	53	68.1	NR	0	At admission	T3:T1	0.17	33.6	360	None
Kerkutluoglu 2023 [17]	Turkey	RC	Outpatients with CHF	218	56	82	100	100	At study inclusion	ROC curve analysis	0.12	26	42	Age, SCr, sodium, BNP, TG, HDL, AST, ALT, diabetes, and beta-blocker use
Sonsöz 2023 [18]	Turkey	RC	Hospitalized patients with ADHF	374	69	52.9	62.8	56.7	At admission	ROC curve analysis	0.78	Within hospitalization	84	Age, functional class ≥ 2, presence of S3, fingertip oxygen saturation, LVEF, NLR, SCr, and LDH
Tanık 2024 [19]	Turkey	RC	Outpatients with CHF	404	61	69.3	100	49	At study inclusion	ROC curve analysis	2.78	30	162	Age, CAD, hypertension, hyperlipidemia, CRF, AF, LVEF, absence of cardiac device, SCr, NT-pro-BNP, TnI, and Hb
Xu 2024 [20]	China	RC	Hospitalized patients with ADHF	1196	66.4	62	NR	51.8	At admission	Q4:Q1	0.63	32.9	568	Age, NYHA class, HR, BNP, UA, chlorine, Hb, TC, and SCr
Szymczak 2025 [22]	Poland	RC	Outpatients with CHF	154	72.1	57.8	50	NR	At study inclusion	ROC curve analysis	1.7	92	85	None
Zuo 2025 [23]	China	RC	Hospitalized patients with ADHF	227	74.1	56.6	NR	55.9	At admission	ROC curve analysis	0.86	46	64	None
Matsuo 2025 [21]	Japan	PC	Hospitalized patients with ADHF	1272	81	57.7	NR	35.7	At admission	Q4:Q1	0.24	24	266	Age, MAGGIC risk score and log-transformed BNP

RC retrospective cohort, PC prospective cohort, CHF chronic heart failure, ADHF acute decompensated heart failure, ICD implantable cardioverter-defibrillator, HFrEF heart failure with reduced ejection fraction, WBC white blood cell, FBG fasting blood glucose, LVEF left ventricular ejection fraction, SCr serum creatinine, BNP B-type natriuretic peptide, NT-pro-BNP N-terminal pro-B-type natriuretic peptide, TG triglyceride, HDL high-density lipoprotein, AST aspartate aminotransferase, ALT alanine aminotransferase, LDH lactate dehydrogenase, CAD coronary artery disease, CRF chronic renal failure, AF atrial fibrillation, TnI troponin I, Hb hemoglobin, NYHA New York Heart Association, HR heart rate, UA uric acid, TC total cholesterol, MAGGIC Meta-Analysis Global Group in Chronic Heart Failure, NR not reported, CAR C-reactive protein-to-albumin ratio

Quality assessment

Study quality was assessed using the NOS (Table 2). Total NOS scores ranged from 6 to 9, indicating that the overall quality of the included evidence base was moderate to high. Four studies [13, 17, 19, 21] achieved the highest score of 9, reflecting strong cohort representativeness, clearly defined exposure measurement, adequate control for confounding (including age and additional variables), standardized outcome assessment, and sufficiently long and complete follow-up. Two studies scored 8, generally performing well in cohort selection, exposure ascertainment, and outcome assessment but with minor limitations such as in adequate representativeness of the exposed cohort [20] or inadequate follow-up duration [18]. The remaining six studies scored 6 to 7 [12, 14–16, 22, 23], most commonly due to limited cohort representativeness, lack of adjustment for key confounders, or shorter duration of follow-up. Collectively, the NOS results support an overall reliable evidence base, with no studies rated as poor quality.

Table 2.

Study quality evaluation via the Newcastle-Ottawa Scale

Study	Representativeness of the exposed cohort	Selection of the non-exposed cohort	Ascertainment of exposure	Outcome not present at baseline	Control for age	Control for other confounding factors	Assessment of outcome	Enough long follow-up duration	Adequacy of follow-up of cohorts	Total
Su 2020 [12]	0	1	1	1	0	0	1	1	1	6
Çinier 2021 [13]	1	1	1	1	1	1	1	1	1	9
Lima 2022 [14]	1	1	1	1	0	0	1	1	1	7
Ren 2022 [15]	0	1	1	1	0	0	1	1	1	6
Feng 2023 [16]	1	1	1	1	0	0	1	1	1	7
Kerkutluoglu 2023 [17]	1	1	1	1	1	1	1	1	1	9
Sonsöz 2023 [18]	1	1	1	1	1	1	1	0	1	8
Tanık 2024 [19]	1	1	1	1	1	1	1	1	1	9
Xu 2024 [20]	0	1	1	1	1	1	1	1	1	8
Szymczak 2025 [22]	0	1	1	1	0	0	1	1	1	6
Zuo 2025 [23]	1	1	1	1	0	0	1	1	1	7
Matsuo 2025 [21]	1	1	1	1	1	1	1	1	1	9

Overall analysis

Pooled results of the 12 studies [12–23] showed that compared to HF patients with a low CAR at baseline, those with a high CAR were associated with an increased risk of mortality during follow-up (RR: 2.34, 95% CI: 1.86 to 2.93, p < 0.001; Fig. 2A) with significant heterogeneity (p for Cochrane Q test < 0.001; I² = 66%). Sensitivity analysis by excluding one study at a time showed consistent results (RR: 2.17 to 2.48, p all < 0.05). In addition, the sensitivity analysis limited to the four studies [17, 19–21] with the adjustment of BNP or NT-proBNP showed consistent results (RR: 1.88, 95% CI: 1.40 to 2.53, p < 0.001; I² = 54%). Moreover, because four studies [13, 14, 17, 19] included only patients with HFrEF, post-hoc sensitivity analysis limited to these studies also showed similar result (RR: 2.26, 95% CI: 1.46 to 3.49, p < 0.001) with mild heterogeneity (I² = 22%).

Fig. 2.

Forest plots for the meta-analysis of the association between baseline CAR and mortality risk in patients with HF. A, overall meta-analysis; and B, subgroup analysis according to the study region

Further subgroup analyses showed similar results between studies from Asian and Western countries (RR: 2.62 vs. 1.87, p for subgroup difference = 0.10; Fig. 2B). The association between CAR and mortality risk in patients with HF was weaker in prospective cohorts as compared to retrospective cohorts (RR: 1.45 vs. 2.50, p for subgroup difference = 0.001; Fig. 3A). Subsequent subgroup analyses showed similar results in patients with ADHF and CHF (RR: 2.15 vs. 2.47, p for subgroup difference = 0.60; Fig. 3B), in patients with mean ages < 65 or ≥ 65 years (RR: 2.36 vs. 2.33, p for subgroup difference = 0.95; Fig. 4A), in studies with cutoff of CAR < 0.5 or ≥ 0.5 mg/g (RR: 2.60 vs. 2.16, p for subgroup difference = 0.44; Fig. 4B), in studies with follow-up duration < 30 or ≥ 30 months (RR: 2.20 vs. 2.45, p for subgroup difference = 0.66; Fig. 5A), and between studies with univariate and multivariate analyses (RR: 2.71 vs. 1.99, p for subgroup difference = 0.15; Fig. 5B).

Fig. 3.

Forest plots for the subgroup analysis of the association between baseline CAR and mortality risk in patients with HF. A, subgroup analysis according to the study design; and B, subgroup analysis according to the clinical status of HF

Fig. 4.

Forest plots for the subgroup analysis of the association between baseline CAR and mortality risk in patients with HF. A, subgroup analysis according to the mean ages of the patients; and B, subgroup analysis according to the cutoff values of CAR

Fig. 5.

Forest plots for the subgroup analysis of the association between baseline CAR and mortality risk in patients with HF. A, subgroup analysis according to the mean follow-up duration; and B, subgroup analysis according to the analytic models

Based on the univariate meta-regression results (Table 3), none of the examined study-level variables—including sample size, mean age, sex distribution, proportion of HFrEF or ischemic HF, CAR cutoff values, follow-up duration, or NOS score—significantly explained the between-study heterogeneity, with all p-values > 0.05. All covariates except NOS showed an adjusted R² of 0%, indicating no meaningful moderating effect. The NOS score explained a small proportion of variance (adjusted R² = 19.8%), but its association with effect size remained statistically non-significant.

Table 3.

Results of univariate meta-regression analysis

Variables	RR for the association between CAR and mortality of patients with HF
	Coefficient	95% CI	p values	Adjusted R2
Sample size	-0.00012	-0.00067 to 0.00043	0.64	0%
Mean age (years)	-0.011	-0.043 to 0.021	0.46	0%
Men (%)	-0.0026	-0.0323 to 0.0270	0.85	0%
HFrEF (%)	0.00091	-0.01050 to 0.01232	0.85	0%
Ischemic HF (%)	0.0016	-0.0102 to 0.0134	0.76	0%
Cutoff of CAR	-0.14	-0.48 to 0.20	0.38	0%
Follow-up duration (months)	-0.0014	-0.0139 to 0.0112	0.81	0%
NOS	-0.12	-0.33 to 0.10	0.26	19.8%

CAR C-reactive protein-to-albumin ratio, HF heart failure, CI confidence interval, RR risk ratio, HFrEF heart failure with reduced ejection fraction, NOS Newcastle–Ottawa Scale

Publication bias

Figure 6 displays the funnel plots evaluating potential publication bias for the meta-analyses of the association of CAR and mortality risk in patients with HF. No obvious asymmetry was observed on inspection, and Egger’s regression tests did not suggest significant small-study effects (p = 0.48).

Fig. 6.

Funnel plots estimating the potential publication bias underlying the meta-analysis of the association between baseline CAR and mortality risk in patients with HF

Discussion

This meta-analysis of 12 cohort studies involving 6,377 patients demonstrated that elevated baseline CAR was associated with a significantly increased risk of all-cause mortality in patients with HF, with a pooled risk ratio of 2.34. The association remained consistent across multiple subgroup analyses, including study region, HF phenotype, age group, CAR cutoff values, follow-up duration, and analytic models. Notably, the magnitude of the association was lower in prospective studies compared with retrospective studies. Sensitivity analyses, including restriction to studies adjusting for natriuretic peptides and those including patients with reduced ejection fraction, yielded similar results, supporting the robustness of the findings.

The association between elevated CAR and mortality in HF is likely related to cardiometabolic dysregulation and cachexia-related pathways that are central to disease progression. Elevated CRP reflects activation of systemic inflammatory signaling, which contributes to endothelial dysfunction, myocardial remodeling, impaired contractility, and neurohormonal activation [30–32]. Persistent inflammation also promotes anabolic resistance, skeletal muscle wasting, and the development of cardiac cachexia—conditions strongly associated with adverse outcomes in HF [33]. In parallel, hypoalbuminemia reflects not only poor nutritional status but also hepatic congestion, reduced protein synthesis, and increased capillary permeability, all of which are common in advanced HF and contribute to fluid imbalance and reduced physiological reserve [34, 35]. Importantly, albumin is closely linked to metabolic homeostasis and antioxidant capacity; reduced levels may therefore indicate a state of heightened oxidative stress and impaired compensatory mechanisms [36]. As a composite index, CAR does not represent a specific biological pathway but rather integrates inflammatory burden and nutritional/metabolic impairment. Therefore, it may serve as a surrogate marker of the complex cardiometabolic and catabolic milieu characteristic of HF, rather than a direct mediator of adverse outcomes [37]. Further mechanistic and longitudinal studies are needed to clarify whether CAR reflects disease severity alone or also participates in pathways that contribute to HF progression.

Interpretation of subgroup findings offers additional insight. The weaker association observed in prospective studies compared with retrospective cohorts deserves careful consideration. Prospective studies are generally characterized by more rigorous study design, standardized data collection, and more comprehensive adjustment for confounding variables, which may attenuate effect estimates by reducing bias and residual confounding [38]. In contrast, retrospective studies often include broader and more heterogeneous patient populations and may be more susceptible to selection bias, incomplete adjustment, and unmeasured confounders, potentially leading to inflated effect sizes [38]. Additionally, prospective cohorts may implement more standardized follow-up and outcome assessment, thereby improving data accuracy but also reducing variability that could exaggerate associations. Differences in clinical management over time and more contemporary treatment strategies in prospective studies may further mitigate the prognostic impact of CAR. Therefore, the observed discrepancy likely reflects methodological differences rather than a true inconsistency in the underlying association, and the directionally consistent findings across study designs support the robustness of CAR as a prognostic marker.

Although similar effect magnitudes were observed in ADHF and CHF subgroups, this finding should not be interpreted as indicating identical underlying pathophysiology. ADHF and CHF represent distinct clinical states with differing inflammatory profiles and hemodynamic characteristics, and the lack of a statistically significant subgroup difference may reflect limited statistical power or unmeasured confounding. The consistent associations across CAR cutoffs indicate that the prognostic value of CAR is not dependent on a single threshold, and that even relatively modest elevations may reflect physiologically relevant derangements. Likewise, stability of the association across short-term and long-term follow-up suggests that CAR reflects both immediate decompensation risk and long-term prognosis. Meta-regression demonstrated no significant moderators among available study-level variables, implying that the prognostic value of CAR is largely independent of age, HF subtype distribution, ischemic etiology, or study quality, though the lack of patient-level data limits definitive conclusions.

Strengths and limitations

This study possesses several methodological strengths. First, the search strategy was comprehensive and up to date, covering Western and Chinese databases to maximize global representativeness. Second, all included studies were longitudinal cohort designs, reducing the possibility of reverse causation compared with cross-sectional analyses. Third, we performed multiple prespecified subgroup analyses and sensitivity analyses to evaluate the robustness of the association. Fourth, meta-regression was used to explore potential sources of heterogeneity, and the lack of significant moderators supports the consistency of the findings. Fifth, the included studies were generally of moderate to high quality based on NOS assessments, further strengthening confidence in the evidence base.

Nonetheless, several limitations merit careful consideration. Most included studies were retrospective cohorts, making them susceptible to recall bias, residual confounding, and selective reporting, which may overestimate associations [39]. Considerable heterogeneity existed across studies with respect to HF etiology, ejection fraction phenotype, HF severity, comorbidities, concomitant medications, and management strategies—factors that could not be fully examined due to lack of individual patient data (IPD). Moreover, the heterogeneity may be partly attributable to unmeasured differences in disease severity and comorbidity burden across studies, including congestion status, frailty, renal and hepatic dysfunction, and systemic inflammatory activity, which were not consistently adjusted for. Although moderate heterogeneity was observed, exploration of additional clinically relevant factors such as HF phenotype (e.g., HFrEF vs. HFpEF) and natriuretic peptide levels was limited by incomplete reporting across studies. Nevertheless, additional sensitivity analyses restricted to studies adjusting for BNP/NT-proBNP and to those including only HFrEF populations yielded consistent results with reduced heterogeneity, supporting the robustness of the findings. Importantly, only four studies adjusted for natriuretic peptides (BNP/NT-proBNP), which are established prognostic markers in HF. Although sensitivity analysis restricted to these studies showed a consistent association, the limited number of such studies prevents firm conclusions regarding the incremental prognostic value of CAR beyond natriuretic peptides. Moreover, the observed heterogeneity may also be partly explained by unmeasured differences in HF severity (e.g., NYHA functional class), variations in treatment strategies including guideline-directed medical therapy, and temporal or regional differences in clinical practice across studies published between 2020 and 2025, which were not consistently reported and therefore could not be formally examined. In addition, differences in laboratory methods, timing of CAR measurement, and cutoff selection may also have introduced variability. Indeed, another important limitation is the substantial variability in CAR cutoff values across studies, which ranged widely and may limit the direct clinical applicability of the findings. The lack of a standardized threshold makes it difficult to define a clinically meaningful risk stratification strategy based on CAR. The use of different methods to define CAR thresholds, including ROC curve–derived cutoffs and quantile-based approaches (e.g., tertiles or quartiles), may further contribute to variability in effect estimates and limit comparability across studies. Moreover, six of the included studies reported only unadjusted effect estimates [12, 14–16, 22, 23], which substantially increases the risk of residual confounding and may have led to overestimation of the association between CAR and mortality. Therefore, the pooled results should be interpreted with caution, as unmeasured or incompletely adjusted clinical factors may partially account for the observed relationship. Given the observational nature of the included studies, the findings should be interpreted as associations rather than evidence of causality. Besides, the modest number of studies in several subgroups limited statistical power to detect nuanced interactions. Finally, although no significant publication bias was detected, the reliability of funnel plot symmetry and Egger’s regression test may be limited due to the relatively small number of included studies, and thus the possibility of publication bias cannot be fully excluded.

Clinical perspectives and future directions

Despite these limitations, the findings carry meaningful clinical implications. CAR is inexpensive, widely available, and easily calculated using routine laboratory parameters, making it an attractive candidate for incorporation into initial risk stratification upon hospital admission or outpatient evaluation. Elevated CAR may help identify patients at higher risk who may warrant closer monitoring and comprehensive clinical assessment. CAR may provide additional prognostic information. However, its incremental value beyond established biomarkers such as natriuretic peptides, renal function markers, and validated risk scores remains uncertain and has not been demonstrated in the current analysis. Future research should aim to establish standardized and clinically relevant CAR cutoff values, ideally through large prospective studies with predefined thresholds or through individual patient data meta-analyses, to facilitate its integration into routine clinical risk assessment. In addition, studies are also needed to evaluate the incremental prognostic value of CAR beyond existing HF risk scores and biomarkers and assess its dynamic changes over time. Well-designed prospective multicenter studies with standardized CAR measurement, uniform adjustment for key confounders, and prespecified thresholds are needed. IPD meta-analyses would further clarify how HF phenotype, comorbidities, and treatment patterns influence the CAR–mortality relationship.

Conclusions

In conclusion, this meta-analysis suggests that elevated baseline CAR is associated with an increased risk of mortality in patients with HF. Although the association was generally consistent across analyses, the findings should be interpreted with caution due to the observational design and heterogeneity of the included studies. While CAR is an easily obtainable biomarker reflecting inflammation and nutritional status, its incremental value for risk stratification remains uncertain. Further well-designed prospective studies are warranted to validate these findings and clarify its clinical utility.

Authors’ contributions

Yan Wang and Zhenfei Yuan designed the study, performed database search, literature review, study quality evaluation, data extraction, statistical analyses, and interpreted the results. Yan Wang drafted the manuscript. All authors revised the manuscript and approved the submission.

Funding

The study did not receive any specific funding from funding agencies in the public, commercial or non-profit sectors.

Data availability

The original contributions presented in the study are included in the article; further inquiries can be directed to the corresponding author.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.