Abstract
Abstract
Objectives
Small dense low-density lipoprotein cholesterol (sdLDL-C) is considered one of the most atherogenic lipid fractions, but its association with cardiovascular events and whether this differs by assessment method remain uncertain. We conducted a systematic review and meta-analysis of prospective cohort studies to quantify this association.
Design
Systematic review and meta-analysis of prospective cohorts; PROSPERO registered (CRD420261411427).
Data sources
PubMed, ProQuest Health & Medical Collection, MEDLINE (Ovid) and Web of Science Core Collection were searched from inception to June 2026.
Eligibility criteria
Prospective cohort studies of community-based or population-based adults reporting HRs for incident cardiovascular events across baseline sdLDL-C categories were eligible.
Data extraction and synthesis
Data were extracted independently by two reviewers. Random-effects restricted maximum likelihood models pooled HRs for the highest versus lowest sdLDL-C categories and for quartile (Q)2, Q3 and Q4 versus Q1 comparisons, stratified by sdLDL-C assessment method. Sensitivity analyses, risk-of-bias assessment, Bayesian hierarchical relative-to-an-identified-distribution (RIDIT)-based modelling and Grading of Recommendations, Assessment, Development and Evaluations (GRADE) assessment were performed.
Results
Fourteen studies were included, comprising 738 114 participants and 77 839 incident cardiovascular events. In the highest-versus-lowest comparison, the pooled HR was 1.47 (95% CI 1.32 to 1.63). The association was stronger for directly measured than calculated sdLDL-C (HR 1.64, 95% CI 1.38 to 1.94 vs HR 1.37, 95% CI 1.21 to 1.54). Between-study heterogeneity was substantial in several analyses. Pairwise quantile and Bayesian RIDIT-based analyses supported a graded positive exposure–risk pattern. Leave-one-out analyses did not materially change the estimates, and Egger’s tests showed no small-study effects. Risk of bias was low to moderate. GRADE certainty was moderate for calculated sdLDL-C and low for the overall and directly measured analyses.
Conclusion
Higher baseline sdLDL-C is associated with increased incident cardiovascular event risk in general adult populations and may serve as an accessible marker of atherogenic cardiovascular risk.
Keywords: Cardiovascular Disease, EPIDEMIOLOGIC STUDIES, Mortality
STRENGTHS AND LIMITATIONS OF THIS STUDY.
This meta-analysis pooled 14 prospective cohorts comprising 738 114 community-based adults with 77 839 incident cardiovascular events, drawn from internationally established cohorts including ARIC, MESA, UK Biobank and CHARLS.
Prespecified stratification by small dense low-density lipoprotein cholesterol (sdLDL-C) assessment method allowed separate evaluation of directly measured and calculated sdLDL-C, which is clinically important because these methods may not be fully interchangeable.
The exposure–risk pattern was assessed using multiple complementary approaches, including highest-versus-lowest comparisons, pairwise quantile analyses, leave-one-out sensitivity analyses, Bayesian hierarchical relative-to-an-identified-distribution-based modelling and Grading of Recommendations, Assessment, Development and Evaluations certainty assessment.
Interpretation should consider the substantial between-study heterogeneity and possible residual confounding. In addition, most studies measured sdLDL-C only once at baseline and therefore could not fully account for changes in lipid levels, lipid-lowering treatment or lifestyle during follow-up.
Introduction
Cardiovascular disease (CVD) remains the leading cause of morbidity and mortality worldwide. Globally, CVD cases increased from 320 million in 1990 to 437 million in 2023, while CVD deaths rose from 13.1 million to 19.2 million over the same period.1 Although low-density lipoprotein cholesterol (LDL-C) is the primary target for lipid-lowering therapy, many individuals with apparently normal LDL-C levels still experience major adverse cardiovascular events.2 This clinical reality has prompted research into specific lipoprotein subfractions that may better reflect atherogenic particle burden and vascular risk beyond conventional lipid measures.
Small dense LDL particles (sdLDL) are considered a highly atherogenic LDL subfraction because of their prolonged circulation time, enhanced endothelial penetration and retention and greater susceptibility to oxidative modification.3 Gold-standard methods for characterising lipoprotein particle size and density are nuclear magnetic resonance spectroscopy and gradient ultracentrifugation, techniques that are not readily applicable to large population study or routine clinical testing due to high cost, technical complexity and labour-intensive nature. As an alternative, automated biochemical assays can measure sdLDL-C and correlate well with gold-standard methods.4 More recently, estimation formulas, such as the Sampson equation, have enabled sdLDL-C calculation from standard lipid panels and shown strong correlation with direct laboratory measurements of sdLDL-C.5
Previous epidemiological studies have reported positive associations between circulating sdLDL-C and incident cardiovascular events6–8; however, several important questions remain unclear. First, although sdLDL-C is widely regarded as an atherogenic lipoprotein parameter, studies from ARIC and the Beijing community population suggest that its predictive value may be attenuated or may no longer remain significant after adjustment for traditional lipid measures such as total cholesterol and HDL-C.9 10 Second, whether calculated sdLDL-C is clinically comparable to directly measured sdLDL-C for cardiovascular risk prediction remains unclear. Prior validation studies indicate that calculated sdLDL-C explains approximately 67%–74% of the variability in direct measurements,5 11 suggesting that the two approaches may not be interchangeable in fully adjusted risk models. Indeed, evidence from large prospective cohorts has suggested that directly measured sdLDL-C may remain a more robust risk enhancer, whereas calculated sdLDL-C may lose statistical significance after adjustment for standard cardiovascular risk factors.11 Finally, associations may differ across vascular outcomes. Evidence for ischaemic heart disease appears relatively consistent,8 12 whereas findings for stroke are more variable. For example, the Copenhagen General Population Study reported a log-linear association between sdLDL-C and stroke,6 whereas the Women’s Health Study found no significant association.13
To our knowledge, no meta-analysis has yet synthesised prospective cohort evidence on sdLDL-C and cardiovascular event risk. Therefore, to address this gap, we conducted a systematic review and meta-analysis to evaluate this association in general adult populations. By stratifying analyses into calculated and directly measured sdLDL-C, we aimed to determine whether the strength and consistency of the association differed by assessment method.
Material and methods
Protocol and registration
This systematic review and meta-analysis was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement and was conducted following the methodological guidance of the Cochrane Handbook. The protocol was prospectively registered with PROSPERO (registration number (CRD420261411427)).
Data sources and search strategy
We systematically searched PubMed (National Library of Medicine), ProQuest Health & Medical Collection (ProQuest), MEDLINE (Ovid) and Web of Science Core Collection (Clarivate) from inception to 10 June 2026, without language restrictions. Search terms combined the exposure and outcome domains using both controlled vocabulary and free-text keywords, including ‘small dense low-density lipoprotein’, ‘small dense LDL’, ‘sdLDL-C’ and ‘low-density lipoprotein subfractions’, combined with ‘cardiovascular disease’, ‘coronary heart disease’, ‘myocardial infarction’, ‘stroke’, ‘atherosclerotic cardiovascular disease’ and ‘cardiovascular mortality’. The complete search strategy for each database is provided in online supplemental file S1. Reference lists of included studies and relevant reviews were manually screened to identify additional eligible records. Records from all databases were exported to EndNote 21 (Clarivate Analytics, US) for automated de-duplication, then imported into Microsoft Excel for Microsoft 365 (Microsoft Corporation, Redmond, WA, US), where the remaining duplicates were automatically flagged and manually verified by two independent reviewers (DDK and NDK).
Eligibility criteria
Studies were eligible if they (1) used a prospective cohort, nested case-control or case-cohort design; (2) were conducted in community- or population-based adult cohorts free of CVD at baseline; (3) quantified sdLDL-C as a cholesterol concentration, determined either by direct measurement (eg, a homogeneous assay (Denka Seiken)) or by estimation from the standard lipid panel (eg, the Sampson equation) and (4) reported HRs, relative risks or ORs with 95% CIs for incident cardiovascular events.
We excluded studies that (1) quantified LDL particle size or diameter, rather than sdLDL-C concentration; (2) were restricted to selected clinical populations (eg, established coronary artery disease); (3) used a cross-sectional or case-control design assessing prevalent disease or (4) did not provide extractable risk estimates. When multiple reports described the same cohort, the publication with the longest follow-up and most complete outcome ascertainment was retained, unless the reports addressed non-overlapping outcomes. The detailed PICOS criteria table is presented in online supplemental file S2.
The exposure of interest was baseline sdLDL-C concentration. The primary effect measure was the HR and its 95% CI for incident CVD, comprising coronary heart disease, stroke and composite atherosclerotic CVD endpoints as defined by the included studies.
Study selection and data extraction
After de-duplication, two reviewers (DDK and NDK) independently screened titles/abstracts and full texts, with disagreements resolved by discussion or, if needed, by a third reviewer (TQD). Data were independently extracted in duplicate by DDK and NDK using a piloted standardised form and reconciled against the included articles; discrepancies were resolved by discussion or adjudication by TQD. From each eligible study, we extracted the first author, publication year, country, cohort name, study design, sample size, number of incident events, follow-up duration, participant age and sex distribution, sdLDL-C measurement method, exposure units and category cut-offs and adjusted effect estimates with 95% CIs for each exposure category. When several multivariable models were reported, we extracted the estimate from the most fully adjusted model; no unadjusted estimates were pooled. Across studies, adjustment covariates consistently included age and sex, and most models additionally adjusted for smoking, body mass index, blood pressure or hypertension, diabetes and one or more conventional lipid measures (total cholesterol, HDL-C, LDL-C or triglycerides). All data required for the meta-analysis (adjusted HRs, 95% CIs, sample sizes and event counts) were available; therefore, no imputation or author contact was undertaken.
Risk of bias and quality of evidence assessment
Two reviewers (NDK and DDK) independently assessed the risk of bias of the included studies using the Quality in Prognosis Studies (QUIPS) tool. The tool comprises six domains: study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding and statistical analysis and reporting. Each domain and the overall study were rated as low, moderate or high risk of bias. The same two reviewers evaluated the certainty of the pooled evidence using the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) approach, considering risk of bias, inconsistency, indirectness, imprecision, publication bias and evidence of an exposure–risk gradient. Disagreements in either assessment were resolved through discussion.
Statistical analysis
Effect estimates were pooled on the log scale using the inverse-variance method, with standard errors derived from the reported 95% CIs. Pooling used random-effects models with the restricted maximum-likelihood estimator for the between-study variance (τ²), chosen a priori to accommodate the anticipated clinical and methodological diversity across cohorts. The primary analysis compared the highest versus the lowest sdLDL-C category. To characterise the exposure–risk gradient, we conducted pairwise meta-analyses comparing each higher category with the lowest (Q1) category in separate analyses (Q2 vs Q1, Q3 vs Q1 and Q4 vs Q1). In these analyses, results were stratified by sdLDL-C measurement approach (calculated vs directly measured), and differences between subgroups were assessed using a test based on the between-subgroup Q statistic. This stratification was prespecified because calculated and directly measured sdLDL-C may yield different associations with cardiovascular risk.
Statistical heterogeneity was quantified using the I² statistic and τ², with the Cochran Q test used to assess significance; I² values of approximately 25%, 50% and 75% were interpreted as low, moderate and high, respectively.
To explore potential sources of heterogeneity, subgroup analyses were conducted for the highest-versus-lowest comparison by geographic region, outcome subtype, follow-up duration, sample size and overall QUIPS risk-of-bias category. Between-subgroup differences were assessed using a test based on the between-subgroup Q statistic.
The robustness of the pooled estimates was examined through leave-one-out sensitivity analyses, in which each study was omitted in turn. Publication bias was evaluated by visual inspection of contour-enhanced funnel plots and Egger’s linear regression test for funnel-plot asymmetry. Where asymmetry was suggested, the Duval and Tweedie trim-and-fill method was applied to derive a bias-adjusted estimate.
To further model the shape of the exposure–risk relationship, we fitted a Bayesian hierarchical dose–response model using the brms interface to Stan, performed separately for overall, calculated and directly measured sdLDL-C. As previously described,14 within each study, exposure categories were converted to a centred relative rank (RIDIT) score derived from the category sample sizes. Study-specific log-HRs, weighted by their standard errors, were then modelled as a function of the centred RIDIT score, with random intercepts and slopes across studies. Weakly informative priors were specified (normal (0, 1) for the intercept and regression coefficient; exponential (1) for the between-study SD). Models were fitted with four chains of 4000 iterations (2000 warmup); convergence was assessed using R̂ (<1.01) and effective sample sizes and model adequacy was examined through posterior predictive checks. Pooled HRs across the exposure distribution were obtained from the population-level posterior and expressed relative to the lowest quantile (HR=1).
Analyses and visualisations were performed using R V.4.2.3 (R Foundation for Statistical Computing, Vienna, Austria) in RStudio V.2026.04.0+526 (Posit Software, Boston, MA, US), with the meta, metafor, brms and ggplot2 packages. A two-sided P value<0.05 was considered significant; Bayesian estimates are reported with 95% credible intervals (CrI).
Patient and public partner involvement
No patients were involved in setting the research question and in developing plans for design, interpretation, reporting or implementation of the study.
Results
Study selection
As shown in online supplemental file S3, the database search identified 2644 records. After removal of 1807 duplicates, 837 unique records underwent title and abstract screening, of which 708 were excluded. A total of 129 full-text articles were assessed for eligibility, and 115 were subsequently excluded, leaving 14 studies for inclusion in the review.6–13 15–20 The full list of excluded articles with reasons for exclusion is provided in online supplemental file S9.
Characteristics of included studies
As shown in online supplemental S4, 14 prospective studies were included, comprising 738 114 participants: 3 18 479 men and 4 19 635 women. Individual study sample sizes ranged from 1325 to 3 21 310, and 77 839 cardiovascular events were recorded. The mean baseline age ranged from approximately 48 to 63 years, and the duration of follow-up ranged from 3.1 to 16 years. The included cohorts were conducted across diverse geographic settings, comprising studies from the US (ARIC,7 MESA,16 17 the Framingham Offspring Study,10 the Women’s Health Study13 and a pooled cohort11), East Asia (the Suita Study,20 the Hisayama Study12 and the Sapporo Health Check-up Cohort in Japan,8 and CHARLS and a Beijing community cohort in China9 15 18), Europe (the Copenhagen General Population Study in Denmark6) and the UK (UK Biobank18 19). sdLDL-C was directly measured by a homogeneous assay in nine cohorts6 7 9 10 12 13 16 17 20 and estimated using the Sampson equation in four,8 15 18 19 while one pooled study applied both approaches.11 The included studies evaluated a broad range of cardiovascular outcomes, including coronary heart disease, ischaemic heart disease, stroke and other cerebrovascular events, atherosclerotic CVD, major adverse cardiovascular events and other composite cardiovascular outcomes. Detailed definitions of the outcomes used in each study are provided in online supplemental table S4.
Pooled risk of cardiovascular events across sdLDL-C categories
Figure 1 shows that higher sdLDL-C categories were associated with increased cardiovascular event risk across all studies, with pooled estimates increasing across pairwise quantile comparisons. In the highest-versus-lowest comparison, the overall pooled HR was 1.47 (95% CI 1.32 to 1.63; I² = 97.2%). Stratification by assessment method showed a stronger association for directly measured sdLDL-C than for calculated sdLDL-C, with pooled HRs of 1.64 (95% CI 1.38 to 1.94; I² = 85.7%) and 1.37 (95% CI 1.21 to 1.54; I² = 91.9%), respectively. In the pairwise quantile comparisons, the overall pooled HRs increased from 1.06 for Q2 versus Q1 (95% CI 1.03 to 1.09; I² = 38%) to 1.18 for Q3 versus Q1 (95% CI 1.13 to 1.24; I² = 70.7%) and 1.43 for Q4 versus Q1 (95% CI 1.29 to 1.58; I² = 96.4%). Across all quantile comparisons, directly measured sdLDL-C was associated with a consistently higher risk of cardiovascular events than calculated sdLDL-C. Leave-one-out analyses showed that no individual study materially influenced the pooled estimates (online supplemental S5. Egger’s regression test showed no evidence of small-study effects in the highest-versus-lowest comparisons for either calculated or directly measured sdLDL-C (online supplemental S6).
Figure 1. Pooled association between sdLDL-C categories and cardiovascular events. Forest plots present pooled HRs and 95% CIs for cardiovascular events across pairwise sdLDL-C quantile comparisons: (A) highest versus lowest category, (B) Q2 versus Q1, (C) Q3 versus Q1 and (D) Q4 versus Q1. Results are shown overall and stratified by sdLDL-C assessment method, including calculated and directly measured sdLDL-C. Between-study heterogeneity is reported using the I² statistic. Q, quartile; sdLDL-C, small dense low-density lipoprotein cholesterol.
Bayesian hierarchical dose–response analysis of sdLDL-C and cardiovascular events
The Bayesian RIDIT-based hierarchical dose–response analysis showed a graded positive association between sdLDL-C quantiles and cardiovascular event risk (figure 2). In the overall analysis, the estimated HR increased from 1.11 in Q2 (95% CrI 1.04 to 1.19), to 1.22 in Q3 (95% CrI 1.12 to 1.35) and 1.35 in Q4 (95% CrI 1.21 to 1.54). A similar increasing pattern was observed in the calculated sdLDL-C, with HRs of 1.09 (95% CrI 1.01 to 1.20), 1.18 (95% CrI 1.07 to 1.32) and 1.28 (95% CrI 1.12 to 1.48) for Q2, Q3 and Q4, respectively. The directly measured sdLDL-C showed corresponding HRs of 1.11 (95% CrI 0.98 to 1.27), 1.23 (95% CrI 1.06 to 1.45) and 1.36 (95% CrI 1.13 to 1.68). Overall, the findings support a positive exposure–risk gradient across sdLDL-C quantiles, with broadly consistent patterns across both assessment methods.
Figure 2. Bayesian dose–response association between sdLDL-C quantiles and cardiovascular events. Bayesian RIDIT-based hierarchical dose–response plots show estimated HRs and 95% credible intervals (CrIs) for cardiovascular events across sdLDL-C quantiles. Results are presented for (A) overall sdLDL-C, (B) calculated sdLDL-C and (C) directly measured sdLDL-C. Q1 was used as the reference category. The red dashed line represents HR=1.0. CrI, credible interval; Q, quartile; RIDIT, relative-to-an-identified-distribution; sdLDL-C, small dense low-density lipoprotein cholesterol.
Subgroup analysis
Subgroup analyses showed generally positive associations between the highest versus lowest sdLDL-C category and cardiovascular event risk across most strata (figure 3). In the overall analysis, pooled estimates were higher in Asian populations, shorter follow-up studies, older age groups, coronary events and the low-risk QUIPS subgroup. In the calculated sdLDL-C, associations were generally consistent but smaller in magnitude. Higher estimates were observed in Asian populations, larger studies, shorter follow-up studies, younger age groups, stroke/cerebrovascular outcomes and the low-risk QUIPS subgroup. In the directly measured sdLDL-C analysis, pooled estimates were generally larger than those for calculated sdLDL-C, with the most pronounced associations observed in Asian populations, for coronary events and in the low-risk QUIPS subgroup.
Figure 3. Forest plots show subgroup-specific pooled HRs and 95% CIs for the risk of cardiovascular events comparing the highest versus the lowest sdLDL-C category. Results are presented for (A) overall sdLDL-C, (B) calculated sdLDL-C and (C) directly measured sdLDL-C. Each pooled estimate is annotated with the number of effect sizes (k) and within-subgroup heterogeneity (I²). CVD, cardiovascular disease; I², inconsistency statistic; k, number of effect sizes; sdLDL-C, small dense low-density lipoprotein cholesterol; QUIPS, Quality in Prognosis Studies.
Risk of bias and certainty of evidence
As shown in online supplemental S7, the QUIPS assessment indicated generally low risk of bias across the 14 included studies. Nine studies were rated as overall low risk and five as moderate risk, with no studies rated as high risk. The moderate ratings primarily reflected some concerns in study participation, including potential selection or representativeness issues and differences in outcome ascertainment in a small number of studies. In online supplemental S8, GRADE certainty was moderate for calculated sdLDL-C and low for the overall and directly measured analyses, mainly due to substantial between-study heterogeneity.
Discussion
This meta-analysis of 14 prospective cohort studies involving 738 114 adults identifies 3 key findings regarding sdLDL-C as a cardiovascular risk marker. First, higher sdLDL-C was associated with an increased risk of cardiovascular events in pooled analyses of fully adjusted models. Second, this association was observed for both directly measured and calculated sdLDL-C, although the magnitude was stronger for directly measured sdLDL-C. Third, sdLDL-C was consistently associated with different cardiovascular events using both measurement methods, with the strongest association observed for coronary events.
In line with previous evidence,11 our findings indicate that directly measured sdLDL-C yielded a stronger pooled association than calculated sdLDL-C. Several factors may explain this result. First, although equation-derived sdLDL-C correlates reasonably well with direct laboratory measurement,5 11 its agreement with direct measurement is reduced in individuals with diabetes21 or hypertriglyceridaemia,22 whereas sdLDL-C levels increase with the number of MetS components.23 Therefore, differences in baseline cardiometabolic risk across subgroup cohorts may have contributed to the observed findings. As shown in online supplemental S4, studies using calculated sdLDL-C had lower mean prevalences of hypertension and diabetes, at approximately 24.5% and 5.0%, respectively, whereas studies using directly measured sdLDL-C had substantially higher prevalences, at approximately 42.0% and 15.1%, respectively. The calculated sdLDL-C subgroup was therefore drawn from comparatively lower-risk populations, potentially with fewer events and a narrower distribution of cardiovascular risk, which may have attenuated the observed HRs toward the null and contributed to the smaller pooled estimate in this subgroup. Second, the two approaches also appear to capture overlapping but not identical aspects of atherogenic lipoprotein metabolism. Direct homogeneous assays quantify the cholesterol content within the sdLDL fraction itself, whereas estimation equations derive sdLDL-C from the standard lipid panel.5 11 Therefore, calculated sdLDL-C may partly reflect information contained in triglycerides and non-HDL cholesterol, rather than purely the cholesterol content within the sdLDL fraction.
A consistent finding across both the overall and the directly measured analyses was that the largest risk was observed for coronary and cardiac endpoints (figure 3A–3C). Two considerations may explain why sdLDL-C appears more strongly associated with coronary events than with stroke. First, coronary events are predominantly atherosclerotic in origin, with approximately 90% of myocardial infarctions resulting from acute coronary thrombosis triggered by rupture or erosion of an atherosclerotic plaque.24 These processes involve apoB-containing LDL retention, oxidation and macrophage uptake.25 Stroke, by contrast, is etiologically heterogeneous; only 20–25% of ischaemic strokes are attributable to large-artery atherosclerosis,26 whereas many others arise from cardioembolism, small vessel occlusion or haemorrhagic mechanisms that are less directly LDL-driven. Therefore, the pro-atherogenic properties of sdLDL particles, including greater arterial wall retention, susceptibility to oxidation and uptake by inflammatory cells, are more closely aligned with the dominant pathway leading to coronary events than with the broader spectrum of stroke. Second, coronary endpoints are typically diagnosed with higher precision than cerebrovascular endpoints in large prospective cohorts. Acute myocardial infarction is defined by internationally standardised criteria combining a rise or fall in cardiac troponin, ischaemic symptoms or electrocardiographic changes and angiographic or imaging confirmation.27 Coronary outcomes in the included cohorts were ascertained either through formal event committees applying these criteria (ARIC,7 MESA,16 the Framingham Offspring Study10) or through linkage to hospital discharge records and death registries using validated coding algorithms with high diagnostic specificity (UK Biobank,18 CHARLS18). In contrast, stroke outcomes in the Women’s Health Study were based on annual questionnaires.13 Previous validation studies have shown that administrative coding algorithms identify acute myocardial infarction more accurately than stroke. Indeed, reported positive predictive values exceed 90% for myocardial infarction compared with approximately 80% for stroke in the UK,28 are 78.9% versus 64.3% in Norway,29 and 88% versus 80%, respectively, in a meta-analysis of 81 studies.30 This lower diagnostic accuracy for stroke, particularly for subtype classification, may introduce outcome misclassification and attenuate associations toward the null, partly explaining the weaker pooled association for stroke than for coronary endpoints.
From a clinical perspective, our population-based findings have two clinical implications for cardiovascular risk assessment and prevention. First, sdLDL-C concentration adds prognostic information beyond conventional LDL-C in adults free of CVD. A substantial proportion of patients hospitalised with acute coronary syndromes present with LDL-C levels considered ‘normal’,2 and even among patients achieving aggressive LDL-C lowering with high-intensity statins or PCSK9 inhibitors, a significant residual cardiovascular risk persists.31 In the FOURIER trial, 9.8% of patients receiving evolocumab experienced a cardiovascular event despite achieving a mean LDL-C of 30 mg/dL,32 and similar residual event rates have been reported in ODYSSEY OUTCOMES33 and IMPROVE-IT.34 In such cases, sdLDL-C may capture a component of this residual atherogenic risk that conventional LDL-C does not, supporting its evaluation as a refinement of standard lipid-based risk assessment. Second, the calculated form of sdLDL-C is obtainable from the standard lipid panel that is already routinely measured in clinical practice, making it immediately deployable in any setting that performs lipid screening, without requiring specialised assays. The consistency of the calculated estimate across UK Biobank and CHARLS18 19 supports generalisability across populations. Further studies are needed to determine whether incorporating sdLDL-C into routine clinical practice improves cardiovascular risk prediction and clinical decision-making beyond established risk factors.
This meta-analysis has several strengths. It provides a large and comprehensive synthesis of prospective evidence on sdLDL-C and cardiovascular event risk in the general adult population. Prespecified stratification by sdLDL-C measurement method enhanced interpretability, while dose–response assessment across Q2–Q4 contrasts, complemented by a Bayesian hierarchical RIDIT model, clarified the exposure–risk gradient.
However, this study has several limitations. First, reliance on a single baseline measurement of sdLDL-C introduces regression dilution bias, as one-time measurement does not account for longitudinal fluctuation in lipid levels or for changes in treatment, such as the initiation of statin therapy, over follow-up. Second, residual confounding remains a persistent concern across these observational designs, as detailed data on dietary habits, physical activity and socioeconomic status were often unavailable or incompletely adjusted for in the primary models. Third, the high collinearity between sdLDL-C and conventional lipid markers, particularly triglycerides and non-HDL-C, complicates the estimation of its independent prognostic value, which frequently diminished or lost statistical significance after comprehensive multivariable adjustment. Fourth, the use of archived plasma samples stored for several years at temperatures ranging from −70 °C to −80 °C raises potential concerns regarding sample stability and measurement accuracy relative to fresh specimens. Fifth, a limited number of specific clinical events (such as stroke or cardiovascular mortality) and relatively short follow-up durations in some cohorts reduced the statistical power for detailed subgroup analyses and limited the ability to infer causality. Finally, the current review focused on cholesterol mass (sdLDL-C) rather than direct sdLDL-P. This distinction is important because it remains unclear whether the absolute number of circulating particles or their cholesterol content is the principal determinant of vascular injury and atherosclerotic risk. Further studies directly comparing the predictive value of sdLDL-C and sdLDL-P for cardiovascular outcomes are needed to clarify their respective pathophysiological roles and to determine which measure provides superior risk stratification and clinical utility.
Conclusion
This meta-analysis demonstrates that higher sdLDL-C is consistently associated with an increased risk of cardiovascular events in general adult populations. The association exhibited a graded exposure–response relationship, was stronger in studies using direct rather than calculated sdLDL-C measurements and was most pronounced for coronary events. These findings support sdLDL-C as a clinically informative marker of atherogenic risk, with direct measurement preferred when available and calculated sdLDL-C representing a practical alternative for routine cardiovascular risk assessment.
Supplementary material
Footnotes
Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online (https://doi.org/10.1136/bmjopen-2026-125031).
Provenance and peer review: Not commissioned; externally peer reviewed.
Patient consent for publication: Not applicable.
Ethics approval: Ethics approval was not required because this systematic review and meta-analysis used published aggregate data and did not involve recruitment of human participants or access to identifiable individual-level data.
Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting or dissemination plans of this research.
Data availability free text: The data underlying this article will be shared on reasonable request to the corresponding.
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
References
- 1.Stark BA, DeCleene NK, Desai EC, et al. Global, Regional, and National Burden of Cardiovascular Diseases and Risk Factors in 204 Countries and Territories, 1990-2023. JACC . 2025;86:2167–243. doi: 10.1016/j.jacc.2025.08.015. [DOI] [PubMed] [Google Scholar]
- 2.Sachdeva A, Cannon CP, Deedwania PC, et al. Lipid levels in patients hospitalized with coronary artery disease: an analysis of 136,905 hospitalizations in Get With The Guidelines. Am Heart J. 2009;157:111–7. doi: 10.1016/j.ahj.2008.08.010. [DOI] [PubMed] [Google Scholar]
- 3.Krauss RM. Small dense low-density lipoprotein particles: clinically relevant? Curr Opin Lipidol. 2022;33:160–6. doi: 10.1097/MOL.0000000000000824. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Albers JJ, Kennedy H, Marcovina SM. Evaluation of a new homogenous method for detection of small dense LDL cholesterol: Comparison with the LDL cholesterol profile obtained by density gradient ultracentrifugation. Clinica Chimica Acta . 2011;412:556–61. doi: 10.1016/j.cca.2010.12.003. [DOI] [PubMed] [Google Scholar]
- 5.Sampson M, Wolska A, Warnick R, et al. A New Equation Based on the Standard Lipid Panel for Calculating Small Dense Low-Density Lipoprotein-Cholesterol and Its Use as a Risk-Enhancer Test. Clin Chem. 2021;67:987–97. doi: 10.1093/clinchem/hvab048. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Balling M, Nordestgaard BG, Varbo A, et al. Small Dense Low-Density Lipoprotein Cholesterol and Ischemic Stroke. Ann Neurol. 2023;93:952–64. doi: 10.1002/ana.26598. [DOI] [PubMed] [Google Scholar]
- 7.Hoogeveen RC, Gaubatz JW, Sun W, et al. Small dense low-density lipoprotein-cholesterol concentrations predict risk for coronary heart disease: the Atherosclerosis Risk In Communities (ARIC) study. Arterioscler Thromb Vasc Biol. 2014;34:1069–77. doi: 10.1161/ATVBAHA.114.303284. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Inyaku M, Tanaka M, Sato T, et al. Calculated Small Dense Low-Density Lipoprotein Cholesterol Level is A Predominant Predictor for New Onset of Ischemic Heart Disease. J Atheroscler Thromb. 2024;31:232–48. doi: 10.5551/jat.64369. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Wang X, Wang L, Cao R, et al. Correlation between small and dense low-density lipoprotein cholesterol and cardiovascular events in Beijing community population. J Clin Hypertens (Greenwich) 2021;23:345–51. doi: 10.1111/jch.14150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ikezaki H, Lim E, Cupples LA, et al. Small Dense Low-Density Lipoprotein Cholesterol Is the Most Atherogenic Lipoprotein Parameter in the Prospective Framingham Offspring Study. J Am Heart Assoc. 2021;10:e019140. doi: 10.1161/JAHA.120.019140. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Schaefer EJ, Ikezaki H, Diffenderfer MR, et al. Atherosclerotic cardiovascular disease risk and small dense low-density lipoprotein cholesterol in men, women, African Americans and non-African Americans: The pooling project. Atherosclerosis. 2023;367:15–23. doi: 10.1016/j.atherosclerosis.2023.01.015. [DOI] [PubMed] [Google Scholar]
- 12.Higashioka M, Sakata S, Honda T, et al. Small Dense Low-Density Lipoprotein Cholesterol and the Risk of Coronary Heart Disease in a Japanese Community. J Atheroscler Thromb. 2020;27:669–82. doi: 10.5551/jat.51961. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Duran EK, Aday AW, Cook NR, et al. Triglyceride-Rich Lipoprotein Cholesterol, Small Dense LDL Cholesterol, and Incident Cardiovascular Disease. J Am Coll Cardiol. 2020;75:2122–35. doi: 10.1016/j.jacc.2020.02.059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Khuc THH, Duc TQ, Di Khanh N, et al. Beyond binary comparisons: A Bayesian dose–response meta-analysis of adherence to the planetary health diet and risks of all-cause mortality and cardiovascular outcomes in adults. Prev Med. 2026;207:108551. doi: 10.1016/j.ypmed.2026.108551. [DOI] [PubMed] [Google Scholar]
- 15.Zhao Y, Huang W, Li W, et al. Estimation of small dense low-density lipoprotein cholesterol and cardiometabolic multimorbidity risk in middle-aged and elderly Chinese: findings from a nationwide prospective cohort. BMC Cardiovasc Disord. 2026;26:374. doi: 10.1186/s12872-026-05734-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Tsai MY, Steffen BT, Guan W, et al. New automated assay of small dense low-density lipoprotein cholesterol identifies risk of coronary heart disease: the Multi-ethnic Study of Atherosclerosis. Arterioscler Thromb Vasc Biol. 2014;34:196–201. doi: 10.1161/ATVBAHA.113.302401. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Nomura SO, Karger AB, Garg P, et al. Small dense low-density lipoprotein cholesterol compared to other lipoprotein biomarkers for predicting coronary heart disease among individuals with normal fasting glucose: The Multi-Ethnic Study of Atherosclerosis. Am J Prev Cardiol . 2023;13:100436. doi: 10.1016/j.ajpc.2022.100436. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ma Y, Zhai X, Mu D, et al. Superior prognostic value of estimated small dense LDL cholesterol for cardiovascular risk assessment: evidence from international cohort studies. Lipids Health Dis. 2025;24:321. doi: 10.1186/s12944-025-02717-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Zubiran R, Sampson M, Wolska A, et al. Estimated Small, Dense LDL Cholesterol and Atherosclerotic Cardiovascular Risk in the UK Biobank. Arterioscler Thromb Vasc Biol. 2025;45:e512–22. doi: 10.1161/ATVBAHA.125.323157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Arai H, Kokubo Y, Watanabe M, et al. Small dense low-density lipoproteins cholesterol can predict incident cardiovascular disease in an urban Japanese cohort: the Suita study. J Atheroscler Thromb. 2013;20:195–203. doi: 10.5551/jat.14936. [DOI] [PubMed] [Google Scholar]
- 21.Hirano T, Ito Y. Accuracy of Small Dense Low-density Lipoprotein-cholesterol Concentration Estimated via Sampson’s Equation in Healthy Subjects and Patients with Diabetes. J Atheroscler Thromb. 2023;30:979–89. doi: 10.5551/jat.63865. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Hirano T. Hypertriglyceridemia contributes significantly to high prevalence of small dense LDL-cholesterol in patients with type 2 diabetes, even when LDL-C targets are achieved. Diabetol Int. 2023;14:288–93. doi: 10.1007/s13340-023-00634-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Fan J, Liu Y, Yin S, et al. Small dense LDL cholesterol is associated with metabolic syndrome traits independently of obesity and inflammation. Nutr Metab (Lond) 2019;16:7. doi: 10.1186/s12986-019-0334-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Qiao JH, Fishbein MC. The severity of coronary atherosclerosis at sites of plaque rupture with occlusive thrombosis. J Am Coll Cardiol. 1991;17:1138–42. doi: 10.1016/0735-1097(91)90844-y. [DOI] [PubMed] [Google Scholar]
- 25.Farahi L, Sinha SK, Lusis AJ. Roles of Macrophages in Atherogenesis. Front Pharmacol. 2021;12:785220. doi: 10.3389/fphar.2021.785220. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Mechtouff L, Rascle L, Crespy V, et al. A narrative review of the pathophysiology of ischemic stroke in carotid plaques: a distinction versus a compromise between hemodynamic and embolic mechanism. Ann Transl Med. 2021;9:1208. doi: 10.21037/atm-20-7490. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Thygesen K, Alpert JS, Jaffe AS, et al. Fourth Universal Definition of Myocardial Infarction (2018) J Am Coll Cardiol. 2018;72:2231–64. doi: 10.1016/j.jacc.2018.08.1038. [DOI] [PubMed] [Google Scholar]
- 28.Arana A, Margulis AV, Varas-Lorenzo C, et al. Validation of cardiovascular outcomes and risk factors in the Clinical Practice Research Datalink in the United Kingdom. Pharmacoepidemiol Drug Saf. 2021;30:237–47. doi: 10.1002/pds.5150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Eliassen B-M, Melhus M, Tell GS, et al. Validity of self-reported myocardial infarction and stroke in regions with Sami and Norwegian populations: the SAMINOR 1 Survey and the CVDNOR project. BMJ Open. 2016;6:e012717. doi: 10.1136/bmjopen-2016-012717. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Davidson J, Banerjee A, Muzambi R, et al. Validity of Acute Cardiovascular Outcome Diagnoses Recorded in European Electronic Health Records: A Systematic Review. Clin Epidemiol. 2020;12:1095–111. doi: 10.2147/CLEP.S265619. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Pradhan AD, Aday AW, Rose LM, et al. Residual Inflammatory Risk on Treatment With PCSK9 Inhibition and Statin Therapy. Circulation. 2018;138:141–9. doi: 10.1161/CIRCULATIONAHA.118.034645. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Sabatine MS, Giugliano RP, Keech AC, et al. Evolocumab and Clinical Outcomes in Patients with Cardiovascular Disease. N Engl J Med. 2017;376:1713–22. doi: 10.1056/NEJMoa1615664. [DOI] [PubMed] [Google Scholar]
- 33.Schwartz GG, Steg PG, Szarek M, et al. Alirocumab and Cardiovascular Outcomes after Acute Coronary Syndrome. N Engl J Med. 2018;379:2097–107. doi: 10.1056/NEJMoa1801174. [DOI] [PubMed] [Google Scholar]
- 34.Cannon CP, Blazing MA, Giugliano RP, et al. Ezetimibe Added to Statin Therapy after Acute Coronary Syndromes. N Engl J Med. 2015;372:2387–97. doi: 10.1056/NEJMoa1410489. [DOI] [PubMed] [Google Scholar]