Association between atherogenic index of plasma and various metabolic conditions: an umbrella review on meta-analyses

Abstract

Background and objective

The Atherogenic Index of Plasma (AIP) is increasingly recognized as a key indicator of lipid disturbances and an important predictor of cardiovascular disease (CVD) risk. Due to its ability to reflect outcomes linked to insulin resistance (IR), dyslipidemia, and atherosclerosis, this umbrella review seeks to compile evidence from multiple meta-analyses to assess clinical significance of AIP across different conditions.

Methods

A comprehensive search was conducted in PubMed, Scopus, and Web of Science. Statistical analyses were performed using Comprehensive Meta-Analysis (CMA) software to aggregate results and assess the strength of the associations.

Results

The results revealed substantial associations between AIP and various health outcomes. AIP was significantly associated with major adverse cardiovascular events (MACE) in acute coronary syndrome (ACS) (RR: 1.54, 95% CI: 1.30–1.82, P < 0.01). AIP was significantly associated with coronary artery disease (CAD), both as a categorical variable (OR: 2.74, 95% CI: 2.05–3.66, P < 0.01) and as a continuous variable (OR: 2.94, 95% CI: 1.85–4.66, P < 0.01). A higher AIP was significantly associated with myocardial infarction (MI) in CAD (RR: 2.21, 95% CI: 1.55–3.13, P < 0.01), coronary artery plaque (CAP) progression (OR: 1.49, 95% CI: 1.17–1.90, P < 0.01), and the development of multivessel lesions (OR: 2.04, 95% CI: 1.50–2.77, P < 0.01). Furthermore, revascularization in CAD was significantly associated with AIP (RR: 1.63, 95% CI: 1.34–1.97, P < 0.01). AIP was significantly associated with MACE in CAD, both as a categorical variable (RR: 1.66, 95% CI: 1.38–2.00, P < 0.01) and as a continuous variable (RR: 1.54, 95% CI: 1.30–1.82, p < 0.01). A higher AIP was significantly associated with CVD death in CAD (RR: 1.74, 95% CI: 1.09–2.77, p = 0.02). Additionally, the no-reflow phenomenon in CAD was significantly associated with AIP (RR: 3.12, 95% CI: 1.09–8.97, P = 0.03). For metabolic outcomes, a higher AIP was significantly associated with obstructive sleep apnea (OSA) (SMD: 0.71, 95% CI: 0.45–0.98, P < 0.01), type 2 diabetes mellitus (T2DM) (SMD: 1.78, 95% CI: 1.05–2.51, P< 0.01), and metabolic syndrome (MetS) (SMD: 0.78, 95% CI: 0.53–1.03, P< 0.01). All-cause mortality in CAD, stroke in CAD, and non-alcoholic fatty liver disease (NAFLD) were not significantly associated with AIP (RR: 1.15, 95% CI: 0.56–2.36, P = 0.69; RR: 1.03, 95% CI: 0.69–1.52, P = 0.90; SMD: 0.16, 95% CI: -0.18–0.50, P = 0.36, respectively).

Conclusion

AIP is significantly associated with a range of CVD and metabolic disorders. These findings suggest that AIP could serve as a valuable biomarker for diagnosing and assessing risk in CVD and metabolic conditions. However, AIP was not significantly associated with all-cause mortality in CAD, stroke in CAD, or NAFLD, highlighting the need for further research to evaluate its clinical utility in diverse patient populations. Clinicians may consider incorporating AIP into broader risk assessment strategies, particularly for patients with existing CVD or metabolic conditions. Additionally, AIP holds potential as a screening tool for large populations, offering clinicians a simple and cost-effective way to identify individuals at higher risk.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-025-05409-w.

Introduction

The Atherogenic Index of Plasma (AIP), defined as the logarithm of the ratio of triglycerides (TG) to high-density lipoprotein cholesterol (HDL), has gained attention as a robust marker of lipid abnormalities and cardiovascular risk [1–3]. By integrating both pro-atherogenic and protective lipoprotein fractions into a single measure, AIP provides a more comprehensive picture of lipid homeostasis compared with traditional lipid parameters [4–6]. Elevated AIP values are thought to reflect smaller, denser low-density lipoprotein (LDL) particles, which are strongly implicated in the pathogenesis of atherosclerosis and related metabolic complications [7–10].

Metabolic conditions such as type 2 diabetes mellitus (T2DM), obesity, metabolic syndrome (MetS), nonalcoholic fatty liver disease (NAFLD), hypertension (HTN), obstructive sleep apnea (OSA), and cardiovascular diseases (CVDs) share overlapping mechanisms involving insulin resistance (IR), chronic inflammation, and dysregulated lipid metabolism [11–19]. These shared pathways not only contribute to metabolic dysfunction but also create a strong rationale for studying lipid indices like AIP as predictive biomarkers. Identifying reliable and cost-effective markers for early detection and monitoring is particularly valuable in clinical practice, where timely intervention may reduce long-term complications [4, 20–23].

In recent years, several meta-analyses have investigated associations between AIP and diverse metabolic conditions [23–26]. However, the available evidence remains fragmented. An umbrella review offers a rigorous strategy to address this gap. By consolidating evidence across conditions and critically appraising methodological quality, umbrella reviews provide the highest level of evidence synthesis [27–30]. The present umbrella review aims to deliver a comprehensive evaluation of the associations between AIP and a broad spectrum of metabolic conditions. In particular, it seeks to summarize the strength and consistency of the existing evidence, assess the methodological rigor of published reviews, and explore the clinical relevance of AIP as a potential biomarker.

Methods

In conducting this umbrella review, we followed the Cochrane Handbook for Systematic Reviews of Interventions to ensure methodological rigor and consistency [31]. Findings were reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [32]. The review protocol was prospectively registered in PROSPERO (CRD420251147536).

Search strategy

An extensive search of the PubMed, Scopus, and Web of Science databases was conducted through May 21, 2025, aiming to identify systematic reviews and meta-analyses that explored the relationship between AIP and various metabolic conditions. The search utilized keywords such as “atherogenic index of plasma,” “AIP,” and “meta-analysis,” without imposing any language or publication date restrictions. The detailed search strategies for each database are provided in Table S1.

Criteria for study selection and eligibility

Studies considered for inclusion in this umbrella review were required to meet the following criteria: (1) the study must be a meta-analysis; (2) the review must focus on the association between AIP and one or more metabolic conditions; (3) the review must report effect sizes (e.g., odds ratios, risk ratios, or mean differences) with corresponding confidence intervals (CI). The review excluded original research studies, case reports, commentaries, editorials, and narrative reviews. Additionally, studies that did not provide adequate data for meaningful synthesis, such as those lacking clear effect sizes or CI, were also omitted. Two reviewers independently screened the identified articles, and any disagreements were resolved through discussion or with assistance from a third reviewer.

Assessment of study quality

The methodological quality of the systematic reviews and meta-analyses included in this umbrella review was evaluated using the AMSTAR 2 checklist [33]. This comprehensive tool assesses multiple dimensions of review quality, including the transparency of the review process, potential sources of bias, and the robustness of statistical approaches used. Based on the evaluation, each study was classified into one of four quality categories: high, moderate, low, or critically low. Any disagreements in quality ratings were resolved through consensus among the reviewers or by consulting an additional third reviewer for an objective assessment. Two reviewers independently evaluated the quality of the included meta-analyses, and any disagreements were resolved through discussion or consultation with a third reviewer.

Data collection and extraction procedures

Two reviewers independently extracted relevant data using a standardized form. The extracted data included the following: (1) study details, such as the first author, publication year, and sample size; (2) effect sizes (e.g., odds ratios, risk ratios, or mean differences) with their 95% CI for AIP and the studied metabolic conditions; (3) information about the statistical methods used, including heterogeneity tests and bias evaluations. In cases where data were unclear or incomplete, the corresponding authors were contacted for clarification. Two reviewers independently extracted the data from the included meta-analyses.

Statistical analysis and interpretation

The CMA software (version 4) was used to aggregate the findings from the individual studies. Given the anticipated heterogeneity among the studies, a random-effects model was applied to conduct the pooled analysis. The degree of variability between studies was assessed using the I² statistic, where a value above 50% indicated considerable heterogeneity. To evaluate publication bias, both visual inspection of the funnel plot and statistical tests, including Egger’s and Begg’s tests were conducted [34, 35].

The trim-and-fill method was applied to account for possible publication bias, thereby providing a more precise and unbiased pooled estimate. In addition, prediction intervals (PI) were computed to reflect the expected range of true effects across different study populations. To examine the stability of the findings, sensitivity analyses were performed, testing whether the overall results were influenced by the exclusion of individual studies.

Results

Study selection

The initial literature search identified 304 records. After removing duplicates (n = 140), 164 unique records remained for screening. Of these, 127 were excluded based on title and abstract review. Thirty-seven full-text articles were assessed for eligibility, and 28 were excluded for not meeting the inclusion criteria (e.g., non–meta-analyses, incomplete data, or lack of relevance). Ultimately, nine studies met the eligibility criteria and were included in the final review. Citation tracking yielded an additional seven records, but these were excluded due to overlapping data with the included studies. The final synthesis, therefore, comprised nine non-duplicate studies, as illustrated in Fig. 1.

Fig. 1.

Study selection process

Characteristics of included studies

Table 1 outlines the main characteristics of the systematic reviews and meta-analyses included in this umbrella review. Most studies were conducted in China [26, 36, 37] and Iran [24, 38, 39], with additional work from Romania [40, 41] and Peru [42]. Publication years ranged from 2014 to 2024.

Table 1.

Characteristics of included studies

Author (year)	Published journal	Searched databased	Date of search	Outcome	Number of included studies	Total sample size	Model of analysis	Quality of assessment checklist	Registration ID	Statistical software	Publication bias assessment	Country
Jiang (2025) [36]	Annals of Medicine	PubMed, Web of Science, and Embase	From inception up to 16 January 2024	Clinical outcome of patients with ACS	9	10,861	REM	NOS	NR	RevMan and STATA	Visual inspection of funnel plots, and Egger’s test	China
Assempoor (2025) [24]	Cardiovascular Diabetology	PubMed, Embase, and Web of Science	From inception up to 13 August 2024	CAD	51	1,230,991	REM	NOS	PROSPERO (CRD42024610676)	Review Manager and R	Visual inspection of funnel plots, and Egger’s test	Iran
Andraschko (2025) [41]	Medicina	PubMed, EMBASE, and Scopus	From inception up to 3 December 2023	Metabolic Syndrome	13	17,689	REM	NOS	NR	R	NR	Romania
Rabiee Rad (2024) [39]	Cardiovascular Diabetology	PubMed, Scopus, and Web of Science	From inception up to 31 December 2023	Cardiovascular outcomes in patients with CAD	16	20,833	REM	NOS	NR	STATA	Visual inspection of funnel plots, and Egger’s test	Iran
Behnoush (2023) [38]	Lipids in Health and Disease	PubMed, Scopus, Web of Science, and Embase	From inception up to 11 May 2023	OSA	14	14,943	REM	NOS	PROSPERO (CRD42023422039)	STATA	Visual inspection of funnel plots, Egger’s test, and Begg’s test	Iran
Ulloque-Badaracco (2022) [40]	Open Medicine	PubMed, Scopus, Web of Science, and Ovid-Medline	From inception up to 15 October 2021	CAD	14	40,902	REM	NOS	PROSPERO (CRD42021289308)	RevMan	Visual inspection of funnel plots and Begg’s test	Perú
Ismaiel (2022) [42]	biomedicines	PubMed, EMBASE, and Scopus	From inception up to 11 May 2022	NAFLD	8	81,178	REM	QUADAS-2	INPLASY (INPLASY202280043)	R	NR	Romania
Wu (2021) [37]	Frontiers in Cardiovascular Medicine	PubMed, Embase, and Web of Science	From inception up to 10 May 2021	CAD	10	29,847	REM	NOS	NR	RevMan	Visual inspection of funnel plots, and Egger’s test	China
Zhu (2015) [26]	Primary Care Diabetes	PubMed Database, CNKI, and Wanfang Database	From inception up to February 2014	T2DM	15	4010	REM	NR	NR	STATA	Visual inspection of funnel plots, Egger’s test, and Begg’s test	China

Sample sizes varied widely—from 4,010 participants in Zhu et al. [26] to over 1.2 million in Assempoor et al. [24]. All reviews used random-effects models, and the Newcastle–Ottawa Scale (NOS) was the primary quality assessment tool. Several reviews reported protocol registration, and analyses were performed using RevMan, STATA, or R. Publication bias was typically assessed using funnel plots and Egger’s or Begg’s tests.

Sample sizes ranged from as few as 4,010 participants in Zhu et al. [26] to more than 1.2 million individuals in Assempoor et al. [24]. All included reviews adopted a random-effects model for statistical synthesis, while the Newcastle–Ottawa Scale (NOS) was the most frequently employed tool for quality assessment. Data analysis was performed using established statistical packages such as RevMan, STATA, and R, with publication bias most commonly assessed by funnel plot inspection and Egger’s or Begg’s tests. Importantly, the included reviews evaluated AIP across a wide spectrum of outcomes, ranging from acute coronary syndrome (ACS), coronary artery disease (CAD), and other CVD outcomes [24, 36, 37, 39, 40] to major metabolic disorders such as MetS [41], T2DM [26], OSA [38], and NAFLD [42]. The quality of included studies is presented in Fig. 2.

Fig. 2.

Quality of included studies

Results of meta-analysis

Occurrence of major adverse cardiac events (MACE) in patients with ACS

The meta-analysis showed that higher AIP values were significantly associated with an increased incidence of MACE in patients with ACS (RR = 1.54, 95% CI: 1.30–1.82, P < 0.01, I² = 47.92%) (Fig. 3.A). Sensitivity analysis further demonstrated that the pooled effect size remained robust, with no substantial variation observed after the stepwise removal of individual studies (Fig. 3.B). In addition, the prediction interval extended from 0.98 to 2.41 (Fig. 3.C). Tests for publication bias, including Egger’s (P = 0.15) and Begg’s (P = 0.50), did not indicate meaningful asymmetry. However, application of the trim-and-fill method, which accounted for three imputed studies positioned on the left side of the funnel plot, adjusted the OR to 1.35 (95% CI: 1.12–1.63) (Fig. 3.D).

Fig. 3.

The association between AIP and the occurrence of MACE in patients with ACS. A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

Risk of CAD

In the analysis of AIP as a continuous variable, a significant association with CAD was observed (OR = 2.94, 95% CI: 1.85–4.66, P < 0.01, I² = 82.37%) (Fig. 4.A). The consistency of this effect was evaluated through sensitivity testing, and the overall pooled estimate remained stable (Fig. 4.B). The prediction interval extended from 0.62 to 13.78 (Fig. 4.C). Evaluation of publication bias using Egger’s test (P = 0.25) and Begg’s test (P = 0.72) did not reveal considerable asymmetry. The trim-and-fill procedure did not add any imputed studies, and the original effect size remained unchanged (Fig. 4.D).

Fig. 4.

The association between AIP and CAD (continuous). A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

When AIP was examined as a categorical variable, the association with CAD also reached statistical significance (OR = 2.74, 95% CI: 2.05–3.66, P < 0.01, I² = 79.07%) (Fig. 5.A). Sensitivity analysis confirmed the robustness of this outcome, as removal of each study one at a time did not materially influence the pooled result (Fig. 5.B). The prediction interval for this model ranged from 0.99 to 7.56, suggesting variability across included datasets (Fig. 5.C). Publication bias tests showed potential distortion, although Egger’s (P = 0.05) and Begg’s (P = 0.06) results did not reach statistical significance. Application of the trim-and-fill method resulted in the imputation of four studies on the right side of the funnel plot, yielding an adjusted OR of 3.31 (95% CI: 2.54–4.30) (Fig. 5.D).

Fig. 5.

The association between AIP and CAD (categorized). A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

Coronary artery plaque (CAP) progression

AIP was significantly associated with CAP progression when evaluated as a categorical variable (OR = 1.49, 95% CI: 1.17–1.90, P < 0.01, I² = 0.00%) (Fig. 6.A). Sensitivity analysis showed that after removal of the study by Won (2021), the association was no longer statistically significant (OR = 1.49, 95% CI: 0.96–2.31, P = 0.07) (Fig. 6.B). The calculated prediction interval extended from 1.17 to 1.90 (Fig. 6.C). Publication bias was not evident, based on Egger’s (P = 0.55) and Begg’s (P = 1.00) results. The trim-and-fill analysis, which imputed one study on the left side of the funnel plot, adjusted the OR to 1.36 (95% CI: 1.04–1.76) (Fig. 6.D). There were not enough eligible studies to evaluate AIP as a continuous variable, and therefore this analysis was not performed.

Fig. 6.

The association between AIP and CAP progression (categorized). A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

Multivessel lesion

AIP exhibited a statistically significant association with the occurrence of multivessel lesions when evaluated in categorical form (OR = 2.04, 95% CI: 1.50–2.77, P < 0.01, I² = 0.00%) (Fig. 7). Because of the small number of eligible investigations (only two studies), neither publication bias assessment nor prediction interval estimation could be carried out. Moreover, the evidence base was insufficient to conduct an assessment of AIP in its continuous form, and thus this analysis was not performed.

Fig. 7.

The association between AIP and multivessel lesions

MACE in CAD

In the categorical assessment, AIP demonstrated a statistically significant association with the occurrence of MACE in patients with CAD (RR = 1.66, 95% CI: 1.38–2.00, P < 0.01, I² = 40.99%) (Fig. 8.A). The robustness of this effect was examined through sensitivity analysis, and the pooled estimate remained consistent, with no meaningful changes observed when individual studies were removed in turn (Fig. 8.B). The prediction interval extended from 1.03 to 2.67 (Fig. 8.C). Evaluation of publication bias using Egger’s (P = 0.15) and Begg’s (P = 0.25) statistics did not reveal significant evidence of asymmetry. Furthermore, the trim-and-fill method did not add any additional studies, supporting the stability of the association (Fig. 8.D).

Fig. 8.

The association between AIP and MACE in CAD (categorized). A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

In the continuous assessment, AIP likewise showed a statistically significant association with the occurrence of MACE in CAD (RR = 1.54, 95% CI: 1.30–1.82, P < 0.01, I² = 61.68%) (Fig. 9.A). Sensitivity testing again confirmed that the overall effect size remained steady, and sequential exclusion of single studies did not materially alter the findings (Fig. 9.B). The predicted interval for this association ranged from 0.99 to 2.37 (Fig. 9.C). Publication bias was not evident, as indicated by Egger’s (P = 0.18) and Begg’s (P = 0.53) results. However, when the trim-and-fill approach imputed one study on the left side of the funnel plot, the adjusted RR was 1.51 (95% CI: 1.28–1.79), while maintaining statistical significance (Fig. 9.D).

Fig. 9.

The association between AIP and MACE in CAD (continuous). A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

All-cause mortality in CAD

The analysis indicated that AIP did not present a statistically significant association with all-cause mortality in CAD (RR = 1.15, 95% CI: 0.56–2.36, P = 0.69, I² = 64.03%) (Fig. 10.A). When sensitivity testing was applied, a significant effect emerged after exclusion of the study by Zheng (2022), yielding an adjusted estimate of RR = 1.66 (95% CI: 1.07–2.58, P = 0.02) (Fig. 10.B). The interval estimate extended widely, ranging from 0.00 to 3087.32 (Fig. 10.C). Assessment of publication bias did not reveal evidence of asymmetry, as suggested by Egger’s (P = 0.75) and Begg’s (P = 1.00) results. Moreover, the trim-and-fill approach did not add any hypothetical studies to the model, and the effect size remained unchanged (Fig. 10.D).

Fig. 10.

The association between AIP and all-cause mortality in CAD. A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

CVD death in CAD

The pooled analysis indicated a significant association between AIP and CVD death in patients with CAD (RR = 1.74, 95% CI: 1.09–2.77, P = 0.02, I² = 0.00%) (Fig. 11.A). When sensitivity testing was performed, this association was no longer evident after exclusion of the study by Qin et al. (2020) (RR = 1.24, 95% CI: 0.59–2.59, P = 0.57) and Wang (2023) (RR = 1.69, 95% CI: 0.88–3.23, P = 0.11) (Fig. 11.B). The range of predicted outcomes was estimated to lie between 1.09 and 2.77 (Fig. 11.C). Assessment of potential publication bias did not reveal significant evidence of asymmetry, as reflected by Egger’s (P = 0.56) and Begg’s (P = 1.00) statistics. Application of the trim-and-fill procedure, which imputed two studies on the right side of the funnel plot, produced an adjusted RR of 2.18 (95% CI: 1.38–3.43) (Fig. 11.D).

Fig. 11.

The association between AIP and CV death in CAD. A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

Myocardial infarction (MI) in CAD

The pooled analysis indicated a significant association between AIP and MI in patients with CAD (RR = 2.21, 95% CI: 1.55–3.13, P < 0.01) (Fig. 12.A). The stability of this association was confirmed through sensitivity testing, as the overall effect size remained consistent and no substantial changes were observed after the sequential exclusion of individual studies (Fig. 12.B). The estimated prediction ranges for this association extended from 1.55 to 3.13 (Fig. 12.C). Evaluation of potential publication bias showed no evidence of asymmetry, based on Egger’s (P = 0.22) and Begg’s (P = 0.29) results. Application of the trim-and-fill method, which imputed two additional studies on the left side of the funnel plot, adjusted the RR to 1.96 (95% CI: 1.43–2.67) (Fig. 12.D).

Fig. 12.

The association between AIP and MI in CAD. A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

Revascularization in CAD

The meta-analysis indicated a significant association between AIP and the need for revascularization in patients with CAD (RR = 1.63, 95% CI: 1.34–1.97, P < 0.01, I² = 0.00%) (Fig. 13.A). The robustness of this finding was assessed through sensitivity testing, and the pooled effect size remained consistent, with no meaningful alterations after the sequential exclusion of individual studies (Fig. 13.B). The estimated prediction ranges for outcomes related to this association extended from 1.34 to 1.97 (Fig. 13.C). Evaluation of publication bias using Egger’s (P = 0.96) and Begg’s (P = 1.00) statistics did not reveal evidence of asymmetry. In addition, the trim-and-fill procedure did not impute any additional studies (Fig. 13.D).

Fig. 13.

The association between AIP and revascularization in CAD. A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

Stroke in CAD

The analysis did not reveal a statistically significant association between AIP and the risk of stroke among patients with CAD (RR = 1.03, 95% CI: 0.69–1.52, P = 0.90, I² = 0.00%) (Fig. 14.A). The consistency of this result was examined through sensitivity testing, and the pooled effect size remained stable, showing no relevant variation after the sequential exclusion of individual studies (Fig. 14.B). The estimated outcome range for this association was calculated between 0.69 and 1.51 (Fig. 14.C). Publication bias was not evident, as indicated by Egger’s (P = 0.22) and Begg’s (P = 1.00) statistics. Furthermore, application of the trim-and-fill procedure, which imputed one additional study on the right side of the funnel plot, adjusted the RR to 1.06 (95% CI: 0.72–1.55) (Fig. 14.D).

Fig. 14.

The association between AIP and stroke in CAD. A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

No-reflow phenomenon in CAD

The pooled analysis indicated that AIP had a statistically significant association with the occurrence of the no-reflow phenomenon in CAD (RR = 3.12, 95% CI: 1.09–8.97, P = 0.03, I² = 89.71%) (Fig. 15.A). When sensitivity testing was performed, the association was no longer evident after exclusion of the studies by Toprak (2024) (RR = 1.99, 95% CI: 0.76–5.20, P = 0.16) and Refaat (2021) (RR = 2.57, 95% CI: 0.86–7.68, P = 0.09) (Fig. 15.B). The calculated interval estimates for possible outcomes extended widely, ranging from 0.04 to 263.25 (Fig. 15.C). Assessment of publication bias did not reveal evidence of asymmetry, as suggested by Egger’s (P = 0.30) and Begg’s (P = 0.73) results. Moreover, the trim-and-fill method, which imputed one study on the left side of the funnel plot, yielded an adjusted RR of 2.57 (95% CI: 0.95–6.94) (Fig. 15.D).

Fig. 15.

The association between AIP and the no-reflow phenomenon in CAD. A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

OSA

The analysis showed that AIP exhibited a statistically significant association with OSA (SMD = 0.71, 95% CI: 0.45–0.98, P < 0.01, I² = 53.95%) (Fig. 16.A). Examination of robustness through sensitivity testing confirmed that the pooled effect size remained stable, with no substantial alterations after sequential removal of individual studies (Fig. 16.B). The estimated range of predicted outcomes for this association extended from − 0.30 to 1.73 (Fig. 16.C). Evaluation of publication bias did not reveal significant asymmetry, as indicated by Egger’s (P = 0.24) and Begg’s (P = 0.30) tests. Furthermore, application of the trim-and-fill method, which imputed one study on the left side of the funnel plot, resulted in an adjusted SMD of 0.63 (95% CI: 0.38–0.89) (Fig. 16.D).

Fig. 16.

The association between AIP and OSA. A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

NAFLD

The analysis indicated that AIP did not display a statistically significant association with NAFLD (SMD = 0.16, 95% CI: − 0.18 to 0.50, P = 0.36, I² = 98.97%) (Fig. 17.A). When sensitivity testing was performed, a significant association emerged after exclusion of the study by Turecky (2021), yielding an adjusted estimate of SMD = 0.78 (95% CI: 0.56–1.00, P < 0.01) (Fig. 17.B). The calculated interval of predicted outcomes for this association extended from − 1.35 to 1.67 (Fig. 17.C). Assessment of publication bias suggested no meaningful indication of bias based on Egger’s (P = 0.94) and Begg’s (P = 1.00) results. Moreover, the trim-and-fill procedure, which imputed one study on the left side of the funnel plot, yielded an adjusted SMD of − 0.48 (95% CI: − 0.87 to − 0.08) (Fig. 17.D).

Fig. 17.

The association between AIP and NAFLD. A: Forest plot, (B): Sensitivity analysis, (C): Prediction interval analysis, (D): Trim and fill analysis

T2DM

The pooled findings indicated a significant association between AIP and T2DM (SMD = 1.78, 95% CI: 1.05–2.51, P < 0.01, I² = 98.55%) (Fig. 18.A). Robustness was assessed through sensitivity testing, which showed that the overall effect size remained stable, with no substantial changes after the sequential exclusion of individual studies (Fig. 18.B). The estimated outcome range for this association was broad, spanning from − 1.42 to 4.97 (Fig. 18.C). Evidence of publication bias was suggested by Begg’s test (P = 0.02), although Egger’s test (P = 0.57) did not indicate bias. Application of the trim-and-fill method, which added four imputed studies on the right side of the funnel plot, resulted in an adjusted SMD of 2.12 (95% CI: 1.46–2.79) (Fig. 18.D).

Fig. 18.

The association between AIP and T2DM: A: Forest plot, (B): SA, (C): PI analysis, (D): Trim and fill analysis

MetS

The analysis indicated a significant association between AIP and MetS (SMD = 0.78, 95% CI: 0.53–1.03, P < 0.01, I² = 99.36%) (Fig. 19.A). Robustness was evaluated using sensitivity testing, and the pooled effect estimate remained stable, with no meaningful alterations observed after the stepwise removal of individual studies (Fig. 19.B). The calculated range of predicted outcomes for this association extended from − 0.23 to 1.80 (Fig. 19.C). Publication bias was not evident, as suggested by Egger’s (P = 0.23) and Begg’s (P = 0.16) statistics. Additionally, application of the trim-and-fill procedure, which imputed three additional studies on the right side of the funnel plot, produced an adjusted SMD of 0.98 (95% CI: 0.56–1.40) (Fig. 19.D).

Fig. 19.

The association between AIP and MetS: A: Forest plot, (B): SA, (C): PI analysis. (D): Trim and fill analysis

Discussion

In this umbrella review of meta-analyses, we found that AIP was consistently associated with a broad range of cardiovascular and metabolic outcomes. Elevated AIP showed a significant relationship with MACE in patients with ACS and CAD. It was also linked to higher associated risks of CAD, MI, multivessel disease, CAP progression, revascularization, CVD-related mortality, and the no-reflow phenomenon. In contrast, no significant associations were observed with all-cause mortality, stroke, or NAFLD. Notably, AIP was strongly associated with key metabolic disorders, including T2DM, MetS, and OSA, supporting its value as a unified biomarker for cardiometabolic risk stratification.

The predictive value of AIP in CVD is closely linked to its reflection of an atherogenic lipid profile. A higher AIP indicates elevated triglycerides and reduced HDL-C, a combination that promotes the formation of small dense low-density lipoprotein (sdLDL), a subtype of LDL-C with strong atherogenic potential [3, 43, 44]. SdLDL particles are smaller, more easily penetrate the arterial wall, and are highly susceptible to oxidative modification [45]. Once oxidized, they are taken up by macrophages, leading to foam cell formation and the progression of atherosclerotic CAP [46]. In addition, sdLDL contributes to lipid peroxidation, stimulates endothelial adhesion molecule expression, and enhances oxidative stress, thereby promoting vascular inflammation and injury [47, 48]. Reduced HDL-C further aggravates these processes by impairing reverse cholesterol transport and weakening antioxidant and anti-inflammatory protection [49]. Also, chronically elevated AIP may indicate excessive cholesterol crystal deposition in the arterial intima, which, when maintained over time, can lead to progressive CVD damage [50, 51].

Beyond CVD outcomes, AIP also reflects the metabolic abnormalities that are central to systemic disorders. The combination of elevated triglycerides and reduced HDL-C, which defines AIP, is a hallmark of IR and diabetic dyslipidemia [52, 53]. In states of IR, excessive release of free fatty acids from adipose tissue promotes hepatic triglyceride synthesis and very-low-density lipoprotein overproduction, leading to further triglyceride accumulation and depletion of HDL-C [54–57]. This dyslipidemic environment favors the predominance of small dense LDL particles and contributes to oxidative stress, chronic inflammation, and adipocyte dysfunction [58]. Also, in OSA, intermittent hypoxia and sleep fragmentation promote hepatic triglyceride synthesis, impair HDL metabolism, and enhance oxidative stress, creating an atherogenic lipid profile that is well reflected by AIP [59, 60]. Collectively, these mechanisms impair metabolic balance, ultimately contributing to the development of metabolic diseases such as type 2 diabetes, metabolic syndrome, and obstructive sleep apnea.

Our findings are consistent with several recently published systematic reviews and meta-analyses, which have also reported strong associations between elevated AIP and the associated risk of CVD and metabolic diseases. For instance, a comprehensive meta-analysis by Assempoor et al., including 51 observational studies, demonstrated that elevated AIP is strongly associated with both the presence and severity of CAD, as well as with adverse prognosis. Specifically, higher AIP was linked to a nearly threefold increased associated risk of CAD, greater coronary calcification, multivessel involvement, and accelerated CAP progression. Moreover, in populations without established CAD, elevated AIP predicted future MACE, with consistent effects observed across both ACS and stable CAD subgroups [24]. Similarly, Jiang et al. conducted a meta-analysis including 13 datasets from nine cohort studies with more than 10,000 ACS patients, showing that higher AIP at admission predicted an increased incidence of MACE during follow-up. Subgroup analyses in this study further revealed stronger associations in older patients and in those with diabetes, highlighting the prognostic utility of AIP in high-risk populations [36]. Together, studies reinforce our findings and support the role of AIP as a robust and clinically relevant biomarker across both stable CAD and ACS settings [24, 36, 37, 39, 40].

Recent studies have demonstrated that, beyond its CVD implications, elevated AIP is also closely associated with major metabolic disorders. A recent meta-analysis systematically evaluated the association between AIP and diabetic nephropathy (DN) in patients with type 2 diabetes. Drawing on 11 datasets from 10 observational studies encompassing 25,773 individuals, the pooled analysis demonstrated that higher AIP levels were significantly associated with an increased risk of DN (RR = 1.51, 95% CI: 1.36–1.67). These findings suggest that AIP may function not only as a marker of CVD risk in diabetes but also as an important indicator of microvascular complications such as DN [61]. Mechanistically, excess triglyceride-rich lipoproteins contribute to glomerular damage by inducing lipotoxic effects, triggering inflammatory pathways, and enhancing oxidative stress [62, 63].

Importantly, our study further revealed a close relationship between AIP and MetS. A recently published meta-analysis reported that individuals with MetS consistently showed higher AIP values than healthy controls (MD = 0.309, 95% CI: 0.214–0.405). Beyond this, AIP also demonstrated good discriminatory capacity for identifying MetS, with an AUC of 0.864 (95% CI: 0.857–0.871), suggesting its potential value as a simple and reliable diagnostic marker [41]. In addition, pooled evidence from eight studies with almost 4,000 participants indicated that patients with OSA had significantly higher AIP values than healthy controls (SMD = 0.71, 95% CI: 0.45–0.97), and these values tended to rise with increasing disease severity [38]. These findings underscore the utility of AIP as an integrative biomarker across diverse metabolic disorders.

Consistent with our findings, recent evidence suggests that AIP does not have a clear predictive role for all-cause mortality in patients with CAD. By contrast, some studies link AIP with mortality [64–66]. These conflicting results may stem from the limited number of available studies and differences in follow-up durations. Similarly, no significant association was observed between AIP and NAFLD in previous meta-analyses [42]. This lack of significance is likely attributable to methodological limitations, including reliance on ultrasound rather than liver biopsy for diagnosis, imbalanced case–control designs, and marked heterogeneity across studies. Moreover, sex- and ethnicity-related differences in lipid metabolism, particularly in the TG/HDL-C ratio, may have further contributed to the inconsistent results. Therefore, the absence of a clear link between AIP and NAFLD should be interpreted with caution, and future well-designed studies are needed to better clarify this relationship.

In line with previous literature, AIP has consistently been shown to outperform conventional lipid ratios such as TC/HDL-C and LDL-C/HDL-C in predicting CVD and metabolic risk. Karimpour Reyhan et al. investigated the relationship between lipid indices and obesity in patients with T2D and reported that AIP was more strongly associated with both obesity and overweight than traditional lipid markers. With an AUC of 0.770, AIP proved to be an accurate and independent predictor of obesity, underscoring its potential for use in clinical screening and risk stratification of T2D patients [67]. Similarly, Babaahmadi-Rezaei et al. highlighted the superior predictive power of AIP compared with traditional lipid indices in identifying metabolic risks, including MetS, supporting its role as a reliable and effective tool in clinical practice [68]. Furthermore, Cai et al. demonstrated that AIP had superior predictive value not only over traditional lipid indices (TC, LDL-C, HDL-C, TG) but also over non-traditional ratios (TC/HDL-C, LDL-C/HDL-C, non-HDL-C). In logistic regression analyses, AIP exhibited the strongest association with CAD, with an odds ratio (OR = 1.719, 95% CI: 1.388–2.128, p < 0.001), reinforcing its utility as an independent and robust risk factor for CVD [69].

In addition to the outcomes synthesized in this umbrella review, recent evidence provides further insight into the broader cardiometabolic relevance of the AIP and other lipid indices. Findings from the Pressioni Arteriose Monitorate E Loro Associazioni (PAMELA) study demonstrated a bidirectional relationship between AIP and serum uric acid, as well as a significant association between AIP and elevated blood pressure [70]. These data strengthen the concept that AIP reflects deeper metabolic disturbances that may influence hemodynamic regulation. Furthermore, while our umbrella review highlights strong links between AIP and major cardiovascular and metabolic outcomes, Additional evidence suggests that lipid indices used in AIP calculation may not be consistently associated with markers of early vascular structural changes. Specifically, studies examining carotid intima–media thickness (IMT) found no independent association between IMT and lipid indices incorporating triglyceride–HDL interactions [71]. Similarly, investigations of arterial stiffness using pulse wave velocity (PWV) and longitudinal changes in PWV (ΔPWV) reported no independent relationship with AIP-related indices [72]. Collectively, these findings suggest that although AIP is strongly linked with metabolic dysfunction and adverse cardiovascular events, its relationship with subclinical vascular remodeling—such as IMT thickening or arterial stiffening—remains uncertain, highlighting the need for additional mechanistic and long-term studies.

In interpreting the non-significant associations observed for NAFLD, stroke, and all-cause mortality, several methodological issues should be acknowledged. These outcomes were supported by a relatively small number of available studies, limiting the statistical power to detect meaningful relationships. In addition, these analyses frequently exhibited substantial heterogeneity and wide prediction intervals, suggesting considerable between-study variability. Such high heterogeneity raises the possibility that the observed non-significant results may be unstable and should not be interpreted as definitive evidence of no association. Differences in diagnostic criteria, population characteristics, and adjustment strategies across studies likely contributed to this inconsistency. Taken together, these factors highlight the need for future research with standardized diagnostic definitions, larger and more homogeneous cohorts, and longer follow-up durations to more reliably evaluate whether AIP plays a role in these outcomes.

From a clinical perspective, AIP represents a simple, inexpensive, and accessible biomarker that can be derived from routine lipid panels. Unlike traditional single lipid measures, AIP captures the interplay between triglycerides and HDL-C, thereby providing a more integrative assessment of atherogenic risk. This makes it a valuable tool for identifying individuals at heightened risk of adverse outcomes, including coronary artery disease, ACS, metabolic syndrome, type 2 diabetes, and their related complications. Importantly, AIP holds particular promise in resource-limited settings, where more advanced or costly biomarkers may not be feasible. By incorporating AIP into existing risk assessment frameworks, clinicians could improve early detection, refine risk stratification, and guide targeted interventions in high-risk patients. Nonetheless, standardized cut-off values and further validation in diverse populations remain necessary before AIP can be fully integrated into routine clinical decision-making.

Strengths, limitations, and future suggestions

This umbrella review synthesizes evidence from a broad range of systematic reviews and meta-analyses, covering diverse cardiometabolic outcomes such as ACS, CAD, MetS, T2DM, OSA, and NAFLD. By pooling findings from large-scale datasets, it provides a comprehensive evaluation of the clinical and prognostic significance of AIP.

Despite its comprehensiveness, several limitations should be acknowledged. First, substantial heterogeneity was observed across included studies, partly due to differences in populations, diagnostic criteria, and cut-off thresholds for AIP. Second, while most outcomes showed significant associations, others did not, possibly reflecting methodological limitations. Additionally, most included reviews were observational in design, raising the possibility of residual confounding. Geographic concentration of studies in Asia (notably China and Iran) also limits the generalizability of findings to other populations. Another limitation concerns the variability in lipid indices reported across the broader literature. Several published meta-analyses have evaluated the TG/HDL-C ratio rather than the true Atherogenic Index of Plasma. Because the TG/HDL-C ratio is not equivalent to AIP and does not reflect the same physiological construct, we did not include those studies in this umbrella review [73, 74]. Moreover, as an umbrella review, our unit of analysis was the meta-analysis itself, not the individual primary studies. Consequently, we relied on the effect estimates, heterogeneity values, and methodological decisions reported by the original authors and were unable to reanalyze study-level data or perform additional subgroup or meta-regression analyses. Our aim was to provide a broad overview of the evidence regarding the Atherogenic Index of Plasma rather than to conduct new stratified or exploratory analytic modeling.

Future investigations should prioritize large-scale, multi-ethnic prospective cohort studies to validate the prognostic utility of AIP across diverse populations. An urgent priority is the standardization of AIP cut-off values, which would allow for consistent interpretation and facilitate its translation into clinical practice. For outcomes where evidence remains inconsistent or non-significant, well-designed studies employing gold-standard diagnostic criteria and extended follow-up durations are warranted. Moreover, integrating AIP with other emerging biomarkers and established risk prediction models may enhance its discriminative accuracy and clinical applicability. Finally, interventional studies are needed to determine whether therapeutic strategies aimed at lowering AIP—through lipid-modifying agents or lifestyle interventions—can lead to measurable improvements in CVD and metabolic outcomes.

Conclusion

This umbrella review provides comprehensive evidence that the AIP is strongly associated with a wide range of CVD and metabolic outcomes. Elevated AIP is significantly associated with the risk, of CAD and ACS, while also showing strong links with metabolic disorders such as T2DM, MetS, and OSA. Importantly, AIP emerges as a simple, inexpensive, and readily available biomarker that integrates lipid-related risk into a single index, offering added value beyond traditional lipid parameters. Although the findings highlight its potential role in clinical risk stratification and preventive strategies, heterogeneity across studies, and lack of standardized cut-off values remain important gaps. Future large-scale, multi-ethnic prospective studies and interventional trials are needed to confirm whether targeting AIP reduction can translate into improved clinical outcomes.

Abbreviations

ACS

Acute Coronary Syndrome

AIP

Atherogenic Index of Plasma

CAD

Coronary Artery Disease

CAP

Coronary artery plaque

CVD

Cardiovascular disease

CCS

Chronic Coronary Syndrome

CI

Confidence Interval

DN

Diabetic Nephropathy

HDL-C

High-Density Lipoprotein Cholesterol

LDL-C

Low-Density Lipoprotein Cholesterol

MACE

Major Adverse Cardiovascular Events

MD

Mean Difference

MetS

Metabolic Syndrome

NAFLD

Non-Alcoholic Fatty Liver Disease

NOS

Newcastle–Ottawa Scale

OR

Odds Ratio

OSA

Obstructive Sleep Apnea

PROSPERO

International Prospective Register of Systematic Reviews

QUADAS-2

Quality Assessment of Diagnostic Accuracy Studies (Version 2)

REM

Random Effects Model

RLP-C

Remnant Lipoprotein Cholesterol

PI

Prediction intervals

RR

Risk Ratio/Relative Risk

sdLDL

Small Dense Low-Density Lipoprotein

SMD

Standardized Mean Difference

T2DM

Type 2 Diabetes Mellitus

TG

Triglycerides

Authors’ contributions

Concept development: E.A.S., P.R. and N.L. Study design: S.M.H, N.L, and F.J. Supervision: E.AS., and P.R. Data collection/processing: M.M, K.P.S, R.E, S.R.A.J, and A.N. Data analysis/interpretation (performed statistical analysis, evaluation, and result presentation): E.A.S., PR, A.M, and P.R. Manuscript drafting: all authors.

Funding

None.

Data availability

The datasets used and/or analyzed during the current study can be provided by the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.