Phenotype-specific associations of circulating adipokine levels with carotid atherosclerosis: a systematic review and meta-analysis

Abstract

Background

Atherosclerosis (AS) is driven by inflammatory and metabolic dysregulation. Carotid atherosclerosis (CA), assessable by ultrasonographic carotid intima–media thickness (cIMT) and plaque, provides a noninvasive window into AS. Adipose tissue–derived adipokines are involved in AS.

Aim

We conducted a phenotype-specific systematic review and meta-analysis of observational studies to quantify associations between circulating adipokines and CA.

Methods

Following a preregistered protocol, PubMed, Embase, and Web of Science were searched. Adults with ultrasound-defined CA (increased cIMT and/or carotid plaque) and controls were included. Mean and standard deviation (SD) of circulating adipokines were extracted and converted to standardized mean difference (SMD) for pooled analysis.

Results

Nineteen studies (n = 5860; Asia/Europe/Americas; 8 cross-sectional, 8 case-control, 3 cohort) were included for quantitative analysis. In the increased cIMT phenotype, adiponectin was lower in CA [SMD = −0.72 (−1.00, −0.44), P < 0.05], whereas in the plaque phenotype it was higher [SMD = 0.29 (0.11, 0.47), P < 0.05]. Leptin was higher in CA, reaching significance in the plaque phenotype [SMD = 0.70 (0.13, 1.28), P = 0.02]. Omentin was consistently lower in CA across phenotypes [SMD = −1.43 (−2.20, −0.66), P < 0.001]. Sensitivity analysis supported robustness for adiponectin (stronger effects after excluding healthy cohorts), and publication-bias assessment was feasible only for adiponectin and were negative.

Conclusions

This study indicated that circulating adipokine levels can serve as phenotype-specific biomarkers in CA.

1. Introduction

Atherosclerosis (AS) often starts in early life and advances without symptoms, so carotid artery are often used as a window into AS and cardiovascular disease (CVD) risk [1,2]. Carotid ultrasonography is a validated, noninvasive imaging modality widely used to assess subclinical atherosclerosis [3]. An increase in carotid intima–media thickness (cIMT) or the presence of plaque indicates development of carotid atherosclerosis (CA), and may help identify individuals at increased risk of future cardiovascular events [4]. Mechanistically, the pathogenesis of AS involves inflammation and disordered lipid metabolism, and perivascular adipose tissue (PVAT)—often regarded as the arterial wall's fourth layer—is thought to modulate these processes through adipokine-mediated paracrine signaling to vessel-wall cells [[5], [6], [7], [8]].

Adipose tissue—derived adipokines include adiponectin, CTRP9, leptin, chemerin, FABP4, resistin, RBP4, omentin, visfatin, adipsin, vaspin, among others [[11], [12], [13], [45]]. The effects of these adipokines on CA—particularly on cIMT and plaque—warrant further investigation. To our knowledge, only one meta-analysis has examined the association between adiponectin and atherosclerotic plaque development [14]. Given the limitations of prior work, we conducted a systematic review and meta-analysis of observational studies to comprehensively assess the associations between circulating adipokines and CA.

2. Materials and methods

This study was prospectively registered in the PROSPERO database (CRD420251120671). It was conducted following the Cochrane Handbook for Systematic Reviews of Interventions and reported in accordance with the PRISMA 2020 guidelines (Supplementary Appendix S1). Data retrieval, extraction, quality assessment and analysis were conducted independently by two researchers, with discrepancies adjudicated by a third researcher.

2.1. Search strategy

Studies were retrieved from PubMed, Web of Science, and Embase using a search strategy that combined subject headings and free-text terms. The search covered studies published up to July 24, 2025 and the detailed strategy is provided in Supplementary Appendix S2.

2.2. Eligibility criteria

The inclusion criteria were defined according to the PECOs framework.

• 1)Population (P): Adults (≥18 years) diagnosed with CA by ultrasound (equipment make/model unrestricted), defined as the presence of increased cIMT and/or carotid atherosclerosis plaques (CAP), as well as controls without the aforementioned findings. There were no restrictions on baseline health status, indicating that both healthy individuals and patients with underlying diseases were eligible;
• 2)Exposure (E): Circulating adipokine levels, including adiponectin, adipsin, chemerin, CTRP9, FABP4, leptin, omentin, RBP4, resistin, vaspin and visfatin. There were no restrictions on specimen type, indicating that adipokines could be measured in either plasma or serum samples;
• 3)Comparator (C): Populations stratified by CA phenotype, including lower versus higher cIMT and absence versus presence of CAP. These phenotypes could be applied individually or in combination as grouping criteria;
• 4)Outcome (O): Differences in circulating adipokine levels across groups;
• 5)Study design (S): Observational studies, including cross-sectional, case-control and cohort designs.

The exclusion criteria included.

• 1)Review, case reports, conference abstracts or randomized comparative trials;
• 2)Non-English publications;
• 3)Studies without accessible full text or duplicated publications.

2.3. Data extraction

Study details were systematically extracted using a structured Excel template. The collected items included: (i) basic information (first author, year of publication, region and study design); (ii)participant characteristics (sample size, baseline health status, age, sex, BMI and cIMT); (iii) outcome data (type of circulating adipokines, assay method, and expression levels—mean ± standard deviation or values converted from other continuous variables).

2.4. Quality assessment

The methodological quality of the included observational studies was assessed using the Newcastle–Ottawa Scale (NOS) [15]. This scale evaluates studies based on three domains: selection of participants, comparability of study groups, and ascertainment of outcomes/exposures. Studies scoring 7–9 points were considered high quality, 4–6 points moderate quality, and 0–3 points low quality.

2.5. Data preprocessing

The extracted circulating adipokines data were all continuous variables. In three-arm studies, the sample size of the control group was equally divided and compared separately with case group [16]. For studies reporting tertiles, the first and second tertiles were combined into a single group, and the third tertile was treated as a separate group. For studies reporting medians and interquartile ranges (IQRs), values were converted to mean ± standard deviation (SD) using established methods [[17], [18], [19]], which provide unbiased estimations suitable for meta-analytic calculations. Standardized mean difference (SMD) and their standard errors were calculated to harmonize results across studies with different units and measurement methods [20]. Data transformations were systematically performed in R 4.5.1 to ensure accuracy and consistency across studies.

2.6. Meta analysis

The meta-analysis was executed in Stata 18.0. The choice of meta-analytic model was determined based on study characteristics and heterogeneity: a common-effect model was applied when the studies were sufficiently homogeneous; a random-effects model (between-study variance τ² estimated by REML) was applied when substantial heterogeneity was present (I² > 50 % or Cochran's Q test P < 0.10). Effect sizes were interpreted according to Cohen's criteria (small: SMD = 0.2, medium: SMD = 0.5, and large: SMD = 0.8) [21]. Two-sided P < 0.05 was considered statistically significant.

2.7. Sensitivity analysis

Sensitivity analysis was conducted by performing leave-one-out within each subgroup to evaluate the influence of single study, and by assessing model robustness through comparisons of the common-effect model with multiple random-effects estimators, including REML, DerSimonian–Laird, Paule–Mandel, and Sidik–Jonkman. Two-sided P < 0.05 was considered statistically significant.

2.8. Publication bias

Publication bias was assessed using Egger's regression intercept test and Begg–Mazumdar's rank-correlation test [22,23]. A funnel plot of study effect size versus its standard error was produced with pseudo 95 % confidence limits, centered on the inverse-variance common-effect pooled estimate [24]. Two-sided P < 0.05 was considered statistically significant.

3. Result

3.1. Study selection

A total of 574 articles were initially identified. After removing duplicates, 404 articles remained. Following title and abstract screening, 168 articles were deemed eligible. Ultimately, 19 articles were included after full-text review (Fig. 1).

Fig. 1.

Flowchart of study inclusion and screening.

3.2. Participants’ characteristics

A total of 19 studies involving 5860 participants were included, comprising 8 cross-sectional study, 3 cohort studies study, and 8 case–control study. Studies were conducted in Asia, Europe, and the Americas. Study populations encompassed: (i) healthy participants or those without clinical CVD; (ii) participants with risk factors, including type 2 diabetes mellitus (T2DM), older age, obesity, metabolic syndrome (MS), or essential hypertension (EHP); (iii) participants with established organ disease, including chronic kidney disease (CKD) and AS; (iv) participants with immune-mediated or infectious conditions, including systemic lupus erythematosus, psoriasis, and HIV-1 infection. Based on the descriptions in the included articles, the studies were grouped into three categories: increased cIMT (n = 10), increased cIMT ± plaque (n = 3), plaque (n = 6), in which the presence of plaques category was considered CAP, whereas increased cIMT was defined as subclinical CA. Among participants with CA, age ranged 45–81 years (pooled weighted mean, 59.04 years), with a pooled weighted mean BMI of 25.83 kg/m² and cIMT of 1.09 mm; among those without CA, age ranged from 37 to 71 years (pooled weighted mean, 48.90 years), with a pooled weighted mean BMI of 25.12 kg/m² and cIMT of 0.72 mm. Circulating adipokines assessed included adiponectin (n = 12), leptin (n = 5), omentin (n = 5), resistin (n = 2), visfatin (n = 2), vaspin (n = 1), and adipsin (n = 1) (Table 1; Supplementary Appendix S3).

Table 1.

The basic characteristics of included studies.

First author year	Region 1 = Asia 2 = Europe 3 = America 4 = Oceania 5 = Africa	Design 1= Cross sectional study 2=Cohort study 3= Case control study	Total sample	Baseline	Adipokine 1 = Adiponectin 2 = Leptin 3 = Resistin 4 = Omentin 5 = visfatin 6 = adipsin 7 = vaspin	Category 1 = increased cIMT 2 = increased cIMT ± plaque 3 = plaque	Grouping criterion
Zhou, 2022 [25]	1	1	283	T2DM	1	1	Characteristics of patients with newly diagnosed type 2 diabetes mellitus stratified by tertiles of carotid intima-media thickness.
Chu, 2022 [26]	1	3	186	EHP	1	2	Subclinical carotid atherosclerosis, which was defined as having a carotid intima–media thickness (cIMT) ≥ 1.0 mm and/or plaque on the carotid artery without any clinical manifestations.
Bolignano, 2022 [27]	2	3	77	CKD	4	1	Pathological (high) cIMT was assumed for mean (right/left) values > 0.9 mm and/or a unilateral cIMT over the 75th percentile of the established age- and sex-dependent reference ranges.
Zhang, 2021 [28]	1	1	483	obesity	6	1	Increased CIMT was defined as the average CIMT ≥0.8 mm.
Bellinati, 2020 [29]	3	3	49	HIV-1	1	1	Increased cIMT was defined as ≥ 0.9 mm
Varona, 2019 [30]	2	2	78	participants without clinical CVD	1 2 3 4 5	2	Pathological CIMT was defined as 0.9 mm. We defined the presence of subclinical carotid atherosclerotic disease if a pathological CIMT was present and/or any carotid atherosclerotic plaque was found.
Liu, 2017 [31]	1	3	170	AS and control	1	1	85 patients with atherosclerosis (cIMT ≥1.2 mm) as our research group and 85 healthy volunteers with cIMT <1.2 mm as control group.
Kocijancic1, 2016 [32]	2	2	92	CKD	4	1	72 subjects without atherosclerosis and any symptoms of subclinical atherosclerosis (control group), 56 subjects as asymptomatic group (group with subclinical atherosclerosis; IMT ≥1.0 mm). Atherosclerosis group comprised patients with plaque (but the number of plaques was not counted and the diameter of the plaques was not determined), previous cardiovascular disease events or/and IMT ≥1.2 mm.
			100	CKD	4	3
Pititto, 2016 [33]	3	1	687	participants without clinical CVD and T2DM	1	1	Values above the 75th percentile for a given age, sex and race, according to previously published quantile-regression models, were considered ‘high’ CIMT.
Wang, 2014 [34]	1	1	235	elderly man	3	3	(1) plaque encroaching into the arterial lumen by at least 0.5 mm or 50 % of the surrounding IMT and (2) a thickness >1.5 mm as measured from the media adventia interface to the intima lumen interface.
McMahon, 2014 [35]	3	2	100	healthy	1 2	3	The following anatomic sites were examined for the presence of atherosclerotic plaque, defined as the presence of focal protrusion into the arterial lumen with a thickness exceeding that of the surrounding wall of at least 50 %
			210	SLE	1 2	3
Grönwall, 2014 [36]	3	1	105	SLE	1	3	The presence of plaque was defined as ≥50 % increase over background IMT in any arterial segment.
Esaki, 2014 [37]	1	1	201	healthy	7	1	Characteristics of study subjects stratified by tertiles of c-IMT levels.
Asha, 2014 [38]	1	3	80	psoriasis	2	1	All carotid ultrasound measurements were performed by an expert cardiologist. Values above 0.7 mm were defined as elevated CIMT.
Tang, 2013 [39]	1	3	180	CKD	5	3	The number of atherosderotic plaques (either as faint gray echoes or bright white echoes protruding into lumen) detected in the bulbar area (from 2 cm below to 2 cm above the bifurcation) of the carotid arteries was recorded on botll sides.
Kozakova, 2012 [40]	2	1	1012	healthy	1 2	1	Carotid plaque was defined as an IMT >1.5 mm in any carotid segment.
Liu, 2011 [41]	1	3	60	MS	4	2	All MetS patients were performed ultrasonography of carotid artery and classified into two subgroups: MetS + AS (CIMT >0.8 mm or with plaques) and MetS AS (CIMT 0.8 mm and without plaques)
Ahn, 2011 [42]	1	1	1353	healthy	1	1	Maximal CIMT value of <0.9 mm and ≥9 mm were used to define positive and negative CIMT test, respectively
Reynolds, 2010 [43]	3	3	119	SLE	1	3	Plaque was defined as ≥50 % increase over background IMT in any arterial segment.

Note: AS = atherosclerosis; cIMT = carotid intima–media thickness; CKD = chronic kidney disease; CVD = cardiovascular sidease; EHP = essential hypertension; HIV-1 = human immunodeficiency virus type 1; T2DM = type 2 diabetes mellitus; MS = metabolic syndrome; SLE = systemic lupus erythematosus.

3.3. Quality assessment

Quality was assessed using the NOS. Scores ranged from 8 to 9, indicating that all included studies were high quality. Detailed scoring criteria and item-level judgments are provided in Supplementary Appendix S4.

3.4. Pooled analysis

Adiponectin, leptin, and omentin were included in the primary quantitative synthesis; resistin and visfatin (k = 2 each) were pooled and are presented in Supplementary Appendix S4, while vaspin and adipsin (k = 1 each) were not pooled and are summarized narratively. (Supplementary Appendix S5).

3.4.1. The risk of developing CA and circulating adiponectin level

Eleven studies (twelve comparisons) reported differences in circulating adiponectin across carotid phenotypes, among which McMahon et al. analyzed the situation of plaque in both healthy and SLE populations. In the increased cIMT subgroup (k = 6), adiponectin level was significantly lower than in controls [SMD = −0.72 (−1.00, −0.44), P < 0.05], despite differences in baseline health status and in the criteria used to define increased cIMT, which is consistent with previous researches. In the increased cIMT ± plaque subgroup (k = 2), the SMD effect was directionally similar but not statistically significant. By contrast, in the plaque subgroup (k = 4), adiponectin level was higher among the CAP patients [SMD = 0.29 (0.11, 0.47), P < 0.05]. In addition, between-subgroup heterogeneity was significant (Qb(2) = 134.44, P < 0.001), which helps explain why the overall pooled effect was not significant. These divergent findings between cIMT and plaque likely reflect differences in phenotype definitions and study populations, underscoring the need to analyze these outcomes separately. (Fig. 2).

Fig. 2.

Subgroup analysis of the risk of developing CA and circulating adiponectin level.

3.4.2. The risk of developing CA and circulating leptin level

Four studies (five comparisons) evaluated circulating leptin across carotid phenotypes. In the increased cIMT subgroup (k = 2), leptin tended to be higher than in controls but was not statistically significant. In the increased cIMT ± plaque subgroup (k = 1), the effect was directionally lower and non-significant. By contrast, in the plaque subgroup (k = 2), leptin was significantly higher among CAP patients [SMD = 0.70, (0.13, 1.28), P = 0.02]. In addition, between-subgroup heterogeneity was significant (Qb(2) = 10.36, P = 0.01), underscoring the importance of analyzing cIMT and plaque separately. (Fig. 3).

Fig. 3.

Subgroup analysis of the risk of developing CA and circulating leptin level.

3.4.3. The risk of developing CA and circulating omentin level

Four studies (five comparisons) evaluated circulating omentin across carotid phenotypes. In the increased cIMT subgroup (k = 2), omentin was significantly lower in case than in controls [SMD = −1.33, (−1.84, −0.82), P < 0.001]. In the increased cIMT ± plaque subgroup (k = 2), the pooled estimate was also negative but not statistically significant. In the plaque subgroup (k = 1), omentin was lower among CAP patients too. In addition, between-subgroup heterogeneity was not significant and the overall effect indicated lower omentin in CA patients [SMD = −1.43, (−2.20, −0.66), P < 0.001], demonstrating the consistency of omentin in different subgroups. (Fig. 4).

Fig. 4.

Subgroup analysis of the risk of developing CA and circulating omentin level.

3.5. Sensitivity analysis

For the adiponectin analysis, removing studies that enrolled healthy participants (Kozakova et al.; McMahon et al.) yielded larger SMD and lower heterogeneity in both the increased cIMT and plaque subgroups, suggesting that adiponectin differences are smaller and less consistent among healthy individuals. In addition, the direction of effect was consistent across model specifications. Taken together, these findings indicate that the results are stable. (Supplementary Appendix S6). As for leptin and omentin, sensitivity analyses were not performed because each subgroup included two or fewer studies.

3.6. Publication bias

Funnel plots and Egger's regression were performed only when ≥10 comparisons were available (adiponectin). When fewer than 10 comparisons were available (leptin, omentin), formal tests were not conducted and funnel plots were inspected qualitatively. For adiponectin (k = 12), the funnel plot appeared approximately symmetric, Egger's and Begg's tests were non-significant. Using Duval and Tweedie's trim-and-fill, no studies were imputed (k = 0) and the pooled effect was unchanged. For leptin (k = 5) and omentin (k = 5), funnel plots are shown for qualitative inspection. (Supplementary Appendix S7).

4. Discussion

In recent decades, shifts in lifestyle and dietary patterns have accelerated the earlier onset and progression of various diseases, especially AS promoted by dyslipidemia and chronic inflammation [9,44]. Due to their superficial location, the carotid arteries can be assessed noninvasively for cIMT and plaque, providing a window into the development and risk of coronary heart disease and stroke. Traditionally, endothelial injury and dysfunction were regarded as the initiating factors in AS. More recently, PVAT has emerged as a focus of intensive investigation [9,45]. Beyond energy storage, AT functions as an active endocrine organ that secrets adipokines that regulate metabolism, inflammation, and other physiological processes [13]. Under normal metabolic conditions, AT secretes both pro-inflammatory and anti-inflammatory adipokines, maintaining homeostatic balance. However, when AT becomes dysfunctional, this secretory profile is altered, resulting in chronic low-grade systemic inflammation [46]. These changes—inflammation, oxidative stress, and altered adipokine production—can be detected by imaging modalities [47].

Against this background, we conducted a systematic review and meta-analysis of observational studies to quantify the associations between circulating adipokines and CA, stratified by carotid phenotype (cIMT and plaque). Nineteen observational studies were included, and adipokines such as adiponectin, leptin, omentin, visfatin and resistin were collected for meta-analyzed. Outcomes were classified into three phenotypes by plaque status: increased cIMT, increased cIMT ± plaque, and plaque. As cIMT thresholds varied by underlying disease, no further stratification was performed within the increased cIMT phenotype.

Adiponectin is a protective adipokine that attenuates oxidative stress and inflammation in endothelial cells via AMPK activation and inhibition of NF-κB pathway [10]. In SLE populations, circulating adiponectin level is abnormally elevated. Consistent with this, Carbone et al. reported that higher serum adiponectin was associated with accelerated CA in patients with SLE [48]. These findings suggest that autoimmune disease may influence circulating adiponectin level. Omentin is another protective adipokine that supports endothelial ffunction, regulates lipid metabolism, and suppresses macrophage inflammation [49]. The meta-analysis found lower circulating omentin in CA across phenotypes. Elevated leptin may promote AS by influencing established risk factors for AS—obesity, diabetes mellitus, hypertension, and sleep disorders [50]. In our analysis, leptin level is higher in both the increased cIMT and plaque phenotypes, with statistical significance observed only in the plaque subgroup. Given the small evidence base, no firm conclusion can be drawn about phenotype-specific differences between increased cIMT and plaque. Moreover, resistin and visfatin levels were higher in cases than in controls, and although not statistically significant, these trends may provide preliminary guidance for researchers.

Overall, the results are consistent with the clinical practice that protective adipokines are inversely associated with CA, whereas the other adipokines are positively associated with CA. Due to the limited number of studies, we did not assess how adjustment for confounders affected the results. Future large cohort studies should predefine carotid phenotypes and stratify analyses by underlying disease to more precisely delineate adipokine–CA associations across populations. In addition, machine learning (ML) and artificial intelligence (AI) are increasingly valuable in CVD diagnosis and prognosis by extracting complex patterns from clinical and imaging data. Integrating precise CA scores (e.g., cIMT percentiles and plaque burden/score) with clinical covariates within ML/AI frameworks may improve risk prediction, as suggested by recent ML/AI studies [51,52].

Admittedly, this study still has limitations that cannot be resolved at present. First, definitive conclusions for other adipokines could not be drawn because the available literature was limited or nonexistent. Second, phenotype and measurement heterogeneity were evident—definitions of increased cIMT were non-uniform and ultrasound protocols varied. And moreover, the increased cIMT ± plaque endpoint conflates distinct carotid phenotypes. Third, statistical power was limited in several subgroups, restricting sensitivity analyses and precluding adequately powered tests for publication bias.

5. Conclusion

According to the available evidence, circulating adipokines showed phenotype-specific relationships with CA. Adiponectin was inversely associated with increased cIMT, leptin levels were higher in cases and reached significance in the plaque phenotype, and omentin was consistently lower in cases across phenotypes. These findings support separate analysis of carotid phenotypes and careful consideration of clinical context when interpreting adipokine–CA associations. Future studies require large prospective cohorts stratified by phenotype and underlying diseases to define the prognostic value of adipokines for CA and downstream cardiovascular risk. In conclusion, this study indicated that circulating adipokine levels can serve as phenotype-specific biomarkers in CA.

CRediT authorship contribution statement

Shuo Yang: Writing – original draft, Visualization, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Hong Shen: Resources, Project administration, Methodology, Investigation, Data curation. Yukun You: Methodology, Formal analysis, Data curation. Zhenyue Fu: Supervision, Methodology. Shuaijie Guo: Supervision, Methodology. Yifan Zhang: Writing – review & editing, Funding acquisition. Qincheng Liu: Methodology. Ying Yang: Methodology. Ye Li: Methodology. Ji Qin: Methodology. Ping Liu: Writing – review & editing, Supervision, Project administration, Funding acquisition.

Financial support

This research was funded by Project of Pudong New Area Health Committee (PW2022E-04) and Shanghai Daystar Project (Sailing Special) [23YF1448100].

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the supplementary data to this article.

Data availability

The raw data analyzed in this study are available from the corresponding author on reasonable request.