Association of Nontraditional Lipid Parameters With Long‐Term Prognosis of Acute Ischemic Stroke: A Principal Components Analysis

Abstract

Background

Lipids can adversely affect the prognosis of acute ischemic stroke (AIS). Unlike traditional lipids, nontraditional parameters can offer more extensive insights, but their significance in AIS remains unexplored. This study aimed to evaluate the association between 11 lipid parameters and AIS comprehensively.

Methods

We leveraged data from 1858 participants with AIS without intravenous thrombolysis to examine the relationship between nontraditional lipid parameters and 1‐year prognostic outcomes. The principal component analysis was used to comprehensively assess the prognostic value of nontraditional lipid parameters.

Results

Four nontraditional lipid parameters, plasma atherosclerotic index, atherosclerosis coefficient, Castelli’s index‐II, and lipoprotein combine index were selected by random forest. Among them, Castelli’s index‐II was the strongest and most comprehensive predictor of the prognosis of patients with AIS and was associated with 1‐year all‐cause mortality (hazard ratio [HR],0.571 [95% CI, 0.342–0.954], P=0.032), all‐cause hospital readmission (HR, 1.364 [95% CI, 1.004–1.852], P=0.047), and prolonged length of hospital stay (odds ratio [OR], 1.204 [95% CI, 1.097–1.321], P<0.001). Higher principal component analysis scores of nontraditional lipid parameters were associated with lower 1‐year all‐cause mortality risk after principal component analysis (HR, 0.498 [95% CI, 0.313–0.791], P=0.003).

Conclusions

Nontraditional lipid parameters, especially Castelli’s index‐II may serve as a new AIS prognostic outcomes indicator. Constructed composite principal component analysis score for lipids was associated with all‐cause mortality at 1 year.

Nonstandard Abbreviations and Acronyms

AC

atherosclerosis coefficient

AIP

atherogenic index of plasma

AIS

acute ischemic stroke

CRI‐I

Castelli’s index I

CRI‐II

Castelli’s index II

IVT

intravenous thrombolysis

LCI

lipoprotein combine index

PCA

principal components analysis

TC

total cholesterol

Clinical Perspective.

What Is New?

• This study identifies Castelli’s index II as the most robust nontraditional lipid predictor of 1‐year mortality in patients with acute ischemic stroke without thrombolysis and introduces a composite principal components analysis score of lipid parameters that significantly correlates with long‐term prognosis.

What Are the Clinical Implications?

• The use of Castelli’s index II and a principal components analysis‐derived lipid score can enhance early risk stratification in acute ischemic stroke, providing clinicians with additional tools to guide individualized treatment and follow‐up planning.

Acute ischemic stroke (AIS), with low‐ and middle‐income countries bearing the greatest burden, is a significant subtype of stroke and ranks as the second leading cause of both disability and death globally. ¹ , ² Although endovascular therapies and intravenous thrombolysis (IVT) have been used extensively, they benefit only a small proportion of patients. ³ So far, the overall IVT rate in China is still <10%, rendering the population without IVT considerable. ⁴ A more profound understanding of its risk factors for effective prevention and management is of great necessity. Consequently, finding valuable biomarkers to improve risk stratification and prevent adverse outcomes in patients with AIS is essential.

Dyslipidemia is an established risk factor for stroke, with affected individuals showing a higher likelihood of developing coronary artery disease (CAD) and IS. ⁵ Dyslipidemia plays a crucial role in 3 primary pathways to IS: atherosclerosis, the formation of thrombosis, and the inflammatory response. ⁶ , ⁷ Patients with a high risk of IS often exhibit quantitative lipoprotein, qualitative lipoprotein, and kinetic abnormalities, fostering a shift to a more atherogenic lipid profile, including higher total cholesterol (TC), triglyceride, and low‐density lipoprotein cholesterol (LDL‐C) levels, and lower high‐density lipoprotein cholesterol (HDL‐C). ⁸ , ⁹ However, the predictive power of a single lipid parameter for stroke prognosis may be limited.

Currently, recent studies highlight nontraditional lipid indicators as comprehensive lipid parameters, including plasma atherosclerotic index (AIP), atherosclerosis coefficient (AC), Castelli’s index‐I (CRI‐I), Castelli’s index‐II (CRI‐II), and lipoprotein combine index (LCI) in diabetes and CAD. ¹⁰ , ¹¹ Compared with traditional lipid parameters, these nontraditional parameters offer richer insights, quantifying risk information and balancing atherogenic and antiatherogenic lipoproteins more effectively. ¹² Previous studies have shown that, in patients with AIS receiving IVT, nontraditional lipid parameters are associated with an increased risk of symptomatic intracranial hemorrhage and a higher risk of poor outcomes. ¹³ , ¹⁴ However, the relationship between nontraditional lipid parameters and the prognosis of patients with IS without IVT remains inadequately explored, and it is still unclear which parameters may serve as the most effective predictors for AIS prognosis in patients not receiving IVT.

Principal component analysis (PCA) is a widely used dimensionality reduction technique in medical research. ¹⁵ PCA works by transforming multiple correlated variables into a smaller set of uncorrelated principal components, thereby reducing the complexity of data while retaining the maximum variance. ¹⁶ Therefore, in clinical data analysis, PCA is widely used to extract insights from high‐dimensional data. ¹⁷ This offers a chance to comprehensively assess the predictive value of nontraditional lipid indicators.

To advance the understanding of stroke cause and improve prevention strategies, more evidence on the association between lipids and the prognosis of patients with AIS who did not receive IVT. This study aimed to investigate the relationship between traditional and nontraditional lipid parameters and the long‐term prognosis of patients with AIS without IVT and to comprehensively assess their predictive value through PCA.

METHODS

Availability of Data and Materials

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Study Population

A total of 3404 participants with AIS without IVT from January to December 2020 were hospitalized in the First Affiliated Hospital of Wenzhou Medical University and the Third Affiliated Hospital of Wenzhou Medical University were retrospectively recruited. All patients were aged >18 years and diagnosed with AIS according to the World Health Organization criteria, which included either clear evidence of a responsible lesion on neuroimaging or the presence of symptoms/signs lasting >24 hours, with hemorrhagic stroke and other nonvascular brain diseases excluded. At enrollment, all patients had complete baseline clinical data, including demographic information, laboratory tests, imaging assessments, and treatment records, as well as at least 1 year of follow‐up data. The follow‐up included key clinical outcomes such as all‐cause mortality, hospital readmission, and length of hospital stay. The exclusion criteria were as follows ¹⁸ : (1) patients with severe diseases such as kidney failure, severe hepatic insufficiency, or cancer; (2) patients with autoimmune diseases; and (3) patients with prestroke disability (with a prior admission modified Rankin Scale score≥2). Finally, 1858 patients were enrolled in this study (Figure S1). The study was approved by the Institutional Ethics Committee review board of both the First and the Third Affiliated Hospital of Wenzhou Medical University and was performed following the Declaration of Helsinki. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Data Collection

All patients’ baseline data, including demographic characteristics (sex, age), personal history data (smoking, drinking), medical history data (previous stroke, hypertension, diabetes, CAD, and atrial fibrillation [AF]), clinical features (National Institutes of Health Stroke Scale, modified Rankin Scale), and therapy history data (lipid‐lowering, antihypertensive, and hypoglycemic therapy) were obtained through electronic medical records. Patients were diagnosed with hypertension if they had evidence of systolic blood pressure ≥140 mm Hg and diastolic blood pressure ≥90 mm Hg or had received any antihypertensive medication. Diabetes was defined as a fasting serum glucose level ≥126 mg/dL or a serum glucose level ≥200 mg/dL on 2 random measurements or a glycated hemoglobin level ≥6.5% or having received antidiabetic therapy (oral hypoglycemic agents or insulin). Lipid‐lowering therapy was defined as the use of lipid‐lowering medications (eg, statins, fibrates, or other lipid‐modifying agents) before admission. CAD was defined by a previous history of CAD. AF was defined by a previous history of AF or a diagnosis of AF by electrocardiography. Smoking and drinking status was defined based on current use.

Laboratory examinations, including blood biochemical examinations, were performed within 24 hours of admission. The nontraditional lipid parameters were calculated as follows: AIP=lg (triglyceride/HDL‐C) ¹⁹ ; AC=non‐HDL‐C/HDL‐C, CRI‐I=TC/HDL‐C, CRI‐II=LDL‐C/HDL‐C ²⁰ ; LCI=TC×triglyceride×LDL‐C/HDL‐C ²¹ ; remnant cholesterol =TC−HDL‐C−LDL‐C ²² ; non‐HDL‐C=TC−HDL‐C. ²³

Clinical Outcomes

The primary outcome was all‐cause mortality within 1 year following AIS onset. The secondary outcomes were all‐cause readmission, stroke recurrence at 1 year, functional outcome improvement, and prolonged length of hospital stay (LOS). Stroke recurrence was defined as a new neurological deficit lasting >24 hours or rehospitalization with a diagnosis of IS, intracerebral hemorrhage, or subarachnoid hemorrhage. Functional outcome improvement was defined as a reduction in modified Rankin Scale score of ≥1 point from 3 to 12 months. ²⁴ Prolonged LOS was defined as a hospital stay exceeding the median number (10 days) of days in patients. ²⁵ , ²⁶ Follow‐up data were collected through telephone interviews conducted by professional clinicians within 1 year of the initial admission. In this study, 95.32% of the patients responded to the phone call at the 1‐year follow‐up.

Statistical Analysis

Data were initially analyzed for normality of distribution by using the Kolmogorov–Smirnov test. Continuous variables were presented as medians and interquartile range (median, interquartile range), and categorical variables were presented as numbers and percentages (%). For displaying the information on nontraditional lipid parameters and long‐term prognosis in more detail, mortality, and readmission grouping was used for baseline characteristics classification and comparison. The Mann–Whitney U test was applied for comparisons involving continuous variables, while the chi‐square test or Fisher’s exact test was employed to compare the groups involving categorical variables.

Lipid parameters associated with the outcome were screened by random forest analysis to rank the influencing factors. Parameters that ranked in the top 10 for the importance values of both the primary and secondary outcomes were selected for further study. To display the information on nontraditional lipid parameters and long‐term prognosis in more detail, the grouping of nontraditional lipid parameters was used for classification and comparison of prognosis outcomes. Kaplan–Meier survival curves and cloud‐rain maps were used to initially explore the relationship between the nontraditional lipid parameters and prognosis outcomes. Based on logistic and Cox regression models, associations between nontraditional lipid parameters and long‐term prognosis outcomes were measured by odds ratios (ORs), hazard ratios (HRs), and 95% CIs. Before conducting the Cox proportional hazards model analysis, we verified the key model assumptions. The proportional hazards assumption was assessed using Schoenfeld residuals for each covariate and for the global model, with all P values >0.05, indicating that the assumption was satisfied. Martingale and deviance residuals were examined to assess the presence of influential observations and the linearity between covariates and the log hazard. No significant nonlinearity or influential outliers were detected. These findings confirm the appropriateness and robustness of the Cox model for our analysis. Factors known to be associated with stroke prognosis outcomes were included in the analyses as confounders, which included demographics (age, sex), National Institutes of Health Stroke Scale score at admission, smoking, drinking, medical history (hypertension, diabetes, previous stroke, CAD, AF), and therapy history data (lipid‐lowering, antihypertensive, and hypoglycemic therapy). In multivariate regression analysis, we adjusted for covariates with a P value <0.05 in multivariate regression analysis: age, sex, National Institutes of Health Stroke Scale score at admission, diabetes, and AF. We also explored the relationship between 4 traditional lipid parameters and long‐term prognosis outcomes to compare the clinical predictive value of traditional and nontraditional lipid parameters. Restricted cubic splines with 5 knots (at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles) were also plotted to demonstrate the correlation between nontraditional lipid parameters and prognosis outcomes.

PCA is a widely used exploratory tool in data analysis, helping to reveal dominant patterns of variability within a population and to represent data in a lower‐dimensional space. ²⁷ Kaiser–Meyer–Olkin value and Bartlett sphericity test were used to evaluate the validity of PCA analysis for the included nontraditional lipid parameters. The weight of each nontraditional lipid parameter was calculated using the variance interpretation rate, and the results were shown in a load diagram. Dimensional reduction of the nontraditional lipid parameters was undertaken using PCA, and the PCA score of nontraditional lipid parameters was reconstructed. The correlation between the PCA score and the outcome was investigated in the regression analysis. Meanwhile, a web‐based dynamic nomogram with nontraditional lipid PCA score was constructed to predict the risk of death in patients with cerebral infarction.

A 2‐tailed P<0.05 was regarded as statistically significant. All statistical analyses were performed by SPSS Statistics 26.0 software (SPSS Inc, Chicago, IL), R version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Baseline Characteristics

In this study, a total of 1858 patients with AIS were enrolled and their baseline characteristics divided according to all‐cause mortality are shown in Table 1. The median age of the included patients was 68.00 (59.00–76.00) years, and 1220 (65.66%) were men. The most common complication was hypertension (n=1236 [66.52%]), followed by diabetes (n=520 [27.99%]) and previous stroke (n=254 [13.67%]). The median AIP for all included participants was 0.16 (−0.02 to 0.33), the median AC was 4.25 (3.39–5.29), the median CRI‐I was 3.66 (2.97–4.43), the median CRI‐II was 2.62 (2.01–3.40), the median LCI was 17.55 (9.73–31.39), the median non‐HDL‐C was 3.60 (2.94–4.36), and the median remnant cholesterol was 0.88 (0.63–1.15). The nonsurvivors tended to be older (67.00 [58.00–76.00] versus 81.00 [75.00–85.50], P<0.001), had higher National Institutes of Health Stroke Scale scores on admission (4.00 [2.00–6.00] versus 9.00 [5.00–15.00], P<0.001), and had a higher proportion of AF (83 [4.69%] versus 14 [16.09%], P<0.001). In traditional lipid parameters, only triglyceride was different between the 2 groups of patients (1.46 [1.06–2.01] versus 1.10 [0.94–1.52], P<0.001). In the nontraditional lipids, the nonsurvivors had lower levels of AIP (0.16 [−0.01 to 0.34] versus 0.03 [−0.10 to 0.19], P<0.001), CRI‐II (2.63 [2.02–3.41] versus 2.21 [1.89–3.12], P=0.031), LCI (17.86 [9.88–31.62] versus 11.43 [6.27–20.56], P<0.001), and non‐HDL‐C (3.63 [2.95–4.36] versus 3.24 [2.70–4.14], P=0.027). The baseline characteristics of patients with AIS divided according to 1‐year all‐cause hospital readmission are shown in Table S1. The patients with hospital readmission at 1 year also had lower levels of CRI‐II (2.64 [2.04–3.41] versus 2.40 [1.89–3.37], P=0.018), LCI (17.74 [9.97–31.56] versus 15.80 [8.30–29.91], P=0.048), and non‐HDL‐C (3.64 [2.95–4.36] versus 3.38 [2.72–4.28], P=0.034). Figure S2 shows the distribution of major causes of readmissions in patients with AIS within 1 year. Patients with AIS were readmitted within 1 year mainly for cerebrovascular (35.2%) and cardiovascular events (18.6%). Among the cerebrovascular events, ischemic events accounted for 89.13%, and hemorrhagic events accounted for 10.87%.

Table 1.

Baseline Characteristics of Subjects Divided by All‐Cause Mortality at 1 Year

Characteristics	Total (n=1858)	Survivor (n=1771)	All‐cause mortality (n=87)	P value
Demographic characteristics
Age, y	68.00 (59.00–76.00)	67.00 (58.00–76.00)	81.00 (75.00–85.50)	<0.001
Sex, male, n (%)	1220 (65.66)	1171 (66.12)	49 (56.32)	0.060
Smoking, n (%)	737 (39.67)	709 (40.03)	28 (32.18)	0.144
Drinking, n (%)	623 (33.53)	601 (33.94)	22 (25.29)	0.095
Clinical status
NIHSS score at admission	4.00 (2.00–7.00)	4.00 (2.00–6.00)	9.00 (5.00–15.00)	<0.001
Hospital length of stay	10.00 (7.00–13.00)	10.00 (7.00–13.00)	10.00 (5.00–14.00)	0.705
Medical history
Hypertension, n (%)	1236 (66.52)	1179 (66.57)	57 (65.52)	0.839
Diabetes, n (%)	520 (27.99)	492 (27.78)	28 (32.18)	0.372
Previous stroke, n (%)	254 (13.67)	241 (13.61)	13 (14.94)	0.724
Coronary artery disease, n (%)	42 (2.26)	38 (2.15)	4 (4.60)	0.133
Atrial fibrillation, n (%)	97 (5.22)	83 (4.69)	14 (16.09)	<0.001
Therapy
Lipid‐lowering, n (%)	68 (3.66)	62 (3.50)	6 (6.90)	0.100
Antihypertensive, n (%)	468 (25.19)	449 (25.35)	19 (21.84)	0.461
Hypoglycemic, n (%)	251 (13.51)	239 (13.50)	12 (13.79)	0.937
Laboratory testing
Total cholesterol, mmol/L	4.66 (3.97–5.43)	4.67 (3.99–5.44)	4.30 (3.79–5.38)	0.055
Triglyceride, mmol/L	1.44 (1.05–1.99)	1.46 (1.06–2.01)	1.10 (0.94–1.52)	<0.001
HDL‐C, mmol/L	1.02 (0.86–1.19)	1.01 (0.86–1.19)	1.07 (0.90–1.22)	0.258
Low‐density lipoprotein cholesterol, mmol/L	2.68 (2.14–3.31)	2.68 (2.15–3.32)	2.50 (1.99–3.16)	0.114
Atherogenic index of plasma	0.16 (−0.02–0.33)	0.16 (−0.01–0.34)	0.03 (−0.10–0.19)	<0.001
Atherogenic coefficient	4.25 (3.39–5.29)	4.28 (3.40–5.30)	3.86 (3.36–5.00)	0.087
CRI‐I	3.66 (2.97–4.43)	3.68 (2.99–4.43)	3.27 (2.79–4.38)	0.057
CRI‐II	2.62 (2.01–3.40)	2.63 (2.02–3.41)	2.21 (1.89–3.12)	0.031
Lipid comprehensive index	17.55 (9.73–31.39)	17.86 (9.88–31.62)	11.43 (6.27–20.56)	<0.001
Non‐HDL‐C	3.60 (2.94–4.36)	3.63 (2.95–4.36)	3.24 (2.70–4.14)	0.027
Remnant cholesterol	0.88 (0.63–1.15)	0.89 (0.63–1.15)	0.81 (0.62–1.06)	0.117

CRI‐I indicates Castelli’s index‐I; and HDL‐C, high‐density lipoprotein cholesterol.

Nontraditional Lipid Parameter Screening by Random Forest

We conducted random forest analysis in three prognosis outcomes: all‐cause mortality, all‐cause hospital readmission, and prolonged LOS. The importance values of the top 10 variables of each outcome are shown in Table S2. Results indicated that AIP, AC, CRI‐II, and LCI consistently ranked among the top 10 variables across all outcomes, identifying this overlapping subset of nontraditional lipid parameters as significant prognostic markers for AIS. We found the low importance of traditional lipid parameters or other confounding factors in long‐term prognostic outcomes by random forest analysis.

Association Between Nontraditional Lipid Parameters and Long‐Term Prognosis

Differences in prognosis outcomes based on tertiles of nontraditional lipid parameters are shown in Figure 1. The red lines in the figure represent the lines of the median length of hospital stay in terciles groups. The Kaplan–Meier survival curves showed that patients with lower levels of CRI‐II had a higher risk of death and hospital readmission within 1 year. However, patients with lower AIP and LCI levels were associated only with a higher risk of death within 1 year (log‐rank P<0.05; Figure 1A, 1C, 1D, and 1G). There was no statistically significant association between AIP and AC levels and 1‐year readmission and LCI and AC levels and 1‐year all‐cause mortality (log‐rank P>0.05; Figure 1B, 1E, 1F and 1H). The cloud‐rain map showed that patients with higher AC and CRI‐II levels had longer hospital stays (P<0.05; Figure 1J and 1K). However, no obvious differences were detected in hospital stays among patients with different levels of AIP and LCI (P>0.05; Figure 1I and 1L).

Figure 1. Differences in outcomes based on nontraditional lipid parameters.

The Kaplan–Meier survival curves between nontraditional lipid parameters (A: AIP, B: AC, C: CRI‐II, D: LCI) and 1‐year mortality. The Kaplan–Meier survival curves between nontraditional lipid parameters (E: AIP, F: AC, G: CRI‐II, H: LCI) and 1‐year hospital readmission. The cloud‐rain map shows the level of the length of hospital stay under nontraditional lipid parameters (I: AIP, J: AC, K: CRI‐II; L: LCI) groups. AC indicates atherogenic coefficient; AIP, atherogenic index of plasma; CRI‐II, Castelli’s index‐II; LCI, lipoprotein combine index; and LOS, length of hospital stay.

The univariate logistic regression and Cox regression results for the risk of poor functional outcome in patients with AIS and variables with analysis that may influence prognosis were obtained from clinicians’ suggestions and clinical experience (Table S3). Table 2 and Table S4 showed the prognosis outcomes after 1 year across tertiles of 4 nontraditional lipid parameters. After adjusting, we found that compared with the first tertile of CRI‐II and AIP, the highest tertile of CRI‐II and AIP were associated with a reduced risk of all‐cause mortality (HR, 0.571 [95% CI, 0.342–0.954]; HR, 0.429 [95% CI, 0.226–0.815]), and the second tertile of LCI was associated with an increased risk of all‐cause mortality (HR, 1.899 [95% CI, 1.062–3.396]). The second tertile of CRI‐II was associated with an increased risk of hospital readmission (HR, 1.364 [95% CI, 1.004–1.852]). Compared with the first tertile, the highest terrile of AC was associated with a reduced risk of prolonged LOS (OR, 0.696 [95% CI, 0.550–0.881]), and the highest terrile of CRI‐II was associated with an increased risk of prolonged LOS (OR, 1.673 [95% CI, 1.322–2.166]). In addition, CRI‐II was associated with a higher risk of functional outcome improvement (OR, 1.138 [95% CI, 1.004–1.290]), and there was no statistical significance between the other 3 nontraditional lipid parameters and functional outcome improvement. All nontraditional lipid parameters and stroke recurrence were not statistically significant. Furthermore, in Figure S3, restricted cubic splines were used to reveal the risk of all‐cause mortality increased nonlinearly with LCI levels (P for nonlinearity=0.019, P for overall=0.041). The risk of hospital readmission at 1 year also increased nonlinearly with CRI‐II levels (P for nonlinearity=0.047, P for overall=0.024). Still, the risk of prolonged LOS increased linearly with CRI‐II levels (P for nonlinearity=0.167, P for overall <0.001).

Table 2.

Multivariate Regression Analysis for Nontraditional Lipid Parameters With Outcomes at 1 Year

Outcomes	Groups	AIP		AC		CRI‐II		LCI
		OR/HR (95% CI)	P value	OR/HR (95% CI)	P value	OR/HR (95% CI)	P value	OR/HR (95% CI)	P value
All‐cause mortality*	Continuous	0.562 (0.242–1.305)	0.180	0.912 (0.781–1.066)	0.248	0.902 (0.736–1.105)	0.318	0.995 (0.984–1.006)	0.379
	T1	Ref.		Ref.		Ref.		Ref.
	T2	0.887 (0.555–1.418)	0.616	0.697 (0.419–1.158)	0.163	0.477 (0.281–0.808)	0.006	1.899 (1.062–3.396)	0.031
	T3	0.429 (0.226–0.815)	0.010	0.652 (0.386–1.101)	0.109	0.571 (0.342–0.954)	0.032	1.401 (0.752–2.611)	0.288
	P for trend		0.013		0.096		0.018		0.025
Hospital readmissions*	Continuous	0.978 (0.593–1.612)	0.930	1.012 (0.930–1.102)	0.774	0.894 (0.787–1.014)	0.082	0.999 (0.994–1.004)	0.619
	T1	Ref.		Ref.		Ref.		Ref.
	T2	0.823 (0.596–1.136)	0.236	0.880 (0.650–1.220)	0.469	1.364 (1.004–1.852)	0.047	0.809 (0.594–1.103)	0.181
	T3	1.018 (0.744–1.394)	0.910	0.995 (0.731–1.355)	0.977	0.890 (0.636–1.245)	0.497	0.810 (0.589–1.114)	0.195
	P for trend		0.907		0.967		0.038		0.183
Prolonged length of hospital stay†	Continuous	1.051 (0.722–1.530)	0.794	0.928 (0.871–0.989)	0.022	1.204 (1.097–1.321)	<0.001	1.001 (0.998–1.004)	0.552
	T1	Ref.		Ref.		Ref.		Ref.
	T2	1.001 (0.791–1.267)	0.995	0.699 (0.553–0.884)	0.003	1.286 (1.021–1.621)	0.032	0.982 (0.778–1.240)	0.878
	T3	0.979 (0.966–1.248)	0.865	0.696 (0.550–0.881)	0.003	1.673 (1.322–2.166)	<0.001	1.130 (0.890–1.434)	0.315
	P for trend		0.864		0.003		0.106		0.314

AC indicates atherogenic coefficient; AIP, atherogenic index of plasma; CRI‐II, Castelli’s index‐II; HR, hazard ratio; LCI, lipoprotein combine index; and OR, odds ratio.

^*Adjusted for age, sex, NIHSS score at admission, atrial fibrillation.

^† Adjusted for age, sex, NIHSS score at admission, diabetes.

Predictive Values of Traditional Lipid Parameters

We explored the correlation between traditional lipid parameters and prognosis by multifactor logistic regression in Table S5. After adjusting, we observed that only the highest tertile of TC was associated with a reduced risk of all‐cause mortality compared with the first tertile of triglyceride (HR, 0.524 [95% CI, 0.310–0.885]). Only the highest tertile of LDL‐C was associated with a reduced risk of all‐cause hospital readmission compared with the first tertile of LDL‐C (HR, 0.626 [95% CI, 0.452–0.868]). Elevated HDL‐C (OR, 0.586 [95% CI, 0.402–0.855]) and LDL‐C (OR, 1.160 [95% CI, 1.042–1.291]) levels were associated with a lower and higher risk of prolonged LOS, respectively. The highest tertiles of LDL‐C was associated with a reduced risk of stroke recurrence (HR, 0.593 [95% CI, 0.355–0.992]) and a greater likelihood of functional outcome improvement (OR, 1.449 [95% CI, 1.033–2.034]) compared with the first tertiles of LDL‐C. Other parameters did not show statistically significant associations.

Nontraditional Lipid Parameter PCA and Dimension Reduction

The Kaiser–Meyer–Olkin value of the nontraditional lipid parameters was 0.679, >0.6, which met the prerequisite requirements of PCA, meaning that the data could be used for PCA research. The data passed the Bartlett sphericity test (P<0.001), indicating that the research data were suitable for PCA.

The number of PCs and the weights assigned to each PC of PCA were shown in Table S6. When the PC changes from 2 to 3, the variance interpretation rate changes less (68.55%–82.94%, 93.77%–100%). That is, when the trend of the gravel map suddenly changes from steep to stable, the number of PCs corresponding to steep to stable is the number of PCs for reference extraction in Figure 2A. Component loadings for the nontraditional lipid parameters are shown in Figure 2B. Systemic components 1 and 2 show the different effects of the individual parameters in Figure 2C. The composite score is the weight of the rate of variance interpretation and the component score after the cumulative calculation. The weight of each nontraditional lipid parameter in the PCA score is shown in Figure 2D and Table S7. We found that LCI (16.49%), AIP (16.44%), and AC (16.27%) were the 3 most representative parameters in PCA analysis. Interestingly, HDL‐C was present in the calculation formulas for all 3 parameters, followed by TC and triglyceride, which appear in 2 nontraditional lipid parameters.

Figure 2. PCA of baseline traditional and nontraditional lipid parameters.

Component loadings for the traditional and nontraditional lipid parameters are shown to produce “lipid components” (A). The eigenvalues in the scree plot (B). The biplots of systemic components 1 and 2 show the different effects of the individual parameters (C). The weight of each nontraditional lipid parameter in the PCA score (D). AC indicates atherogenic coefficient; AIP, atherogenic index of plasma; CRI‐I, Castelli’s index‐I; CRI‐II, Castelli’s index‐II; HDL‐C, high‐density lipoprotein cholesterol; LCI, lipoprotein combine index; PCA, principal components analysis; and RC, remnant cholesterol.

Association Between PCA Score and Long‐Term Prognosis

The PCA score calculation formula of nontraditional lipid parameters can be obtained through PCA as follows: PCA score=36.27% AIP+35.89% AC+24.21% CRI‐I+25.40% CRI‐II+36.36% LCI+34.61% remnant cholesterol +27.86% non‐HDL‐C. The PCA score was divided into higher and lower groups based on its median of −0.130. Table 3 and Table S8 showed the associations of 1‐year prognostic outcomes with exposure to nontraditional lipid PCA scores. Higher PCA scores were associated with lower 1‐year all‐cause mortality risk, regardless of adjustment for confounders (HR, 0.498 [95% CI, 0.313–0.791]). However, no statistical significance was observed between PCA score and 1‐year readmission, stroke recurrence, functional outcome improvement, and prolonged LOS.

Table 3.

Associations of 1‐Year Outcome With Exposure to Nontraditional Lipid Score (PCA scores>−0.130)

Outcomes	Unadjusted		Adjusted
	OR/HR (95% CI)	P value	OR/HR (95% CI)	P value
All‐cause mortality*	0.442 (0.281–0.696)	<0.001	0.498 (0.313–0.791)	0.003
Hospital readmissions*	0.809 (0.627–1.045)	0.105	0.857 (0.661–1.111)	0.244
Prolonged length of hospital stay†	0.890 (0.739–1.072)	0.219	0.893 (0.738–1.082)	0.249

HR indicates hazard ratio; NIHSS, National Institute of Health Stroke Scale; OR, odds ratio; and PCA, principal component analysis.

^*Adjusted for age, sex, NIHSS score at admission, atrial fibrillation.

^† Adjusted for age, sex, NIHSS score at admission, diabetes.

Dynamic Nomogram for Predicting All‐Cause Mortality Risk

Additionally, a web‐based dynamic nomogram was developed to facilitate the clinical application of this predictive model. By entering the corresponding values for independent prognostic factors, the probability of all‐cause mortality in patients with AIS can be rapidly predicted. This dynamic nomogram is freely available at the following website: https://strokenontraditionalipid.shinyapps.io/dynnomapp/. Figure 3A shows the input interface for the nomogram, where the user can personally enter the required variables. As a demonstration, Figures 3B and 3C show the survival curves, predicted probabilities (95% CI) for the risk of all‐cause mortality at 1 year for different patients.

Figure 3. A web‐based dynamic nomogram for predicting all‐cause mortality at 1 year.

The nomogram incorporated a panel of independent prognostic factors for the prediction of all‐cause mortality at 1 year (A). Graphical summaries showed survival curves (B) and predicted probabilities and 95% CI (C) for 1‐year all‐cause mortality risk predicted by dynamic nomogram for different patients. AF indicates atrial fibrillation; NIHSS, National Institutes of Health Stroke Scale; and PCA, principal component analysis.

Table S9 provided specific details and predicted probabilities for the 7 patients in the demonstration.

DISCUSSION

This study is the first to focus on the impact of nontraditional lipid parameter levels on long‐term prognostic outcomes in patients with AIS. Notably, we found that higher CRI‐II levels were associated with a lower risk of all‐cause death at 1 year and a longer length of hospital stay. This suggests that CRI‐II may be a relatively comprehensive predictor of AIS. In general, the predictive efficacy of nontraditional lipid parameters for prediabetes surpassed that of traditional lipid parameters. After PCA analysis, our constructed composite PCA score for lipids was associated with all‐cause mortality at 1 year. Besides, LCI, AIP, and AC were the 3 most representative parameters in PCA analysis, highlighting their significance. Additionally, HDL‐C was included in the formulas for all 3 parameters, and TC and triglyceride appeared in 2 nontraditional lipid parameters. This suggests the crucial role of HDL‐C, TC, and triglyceride, especially HDL‐C, in lipid metabolism. To sum up, our study underscores that nontraditional lipid parameters could serve as alternative markers of traditional lipid parameters and in predicting poor prognostic outcomes of patients with AIS in primary clinical settings. This may provide physicians with additional prognostic information to help formulate treatment and rehabilitation for patients with IS.

Potential explanations have been proposed regarding the mechanisms underlying the predictive value of AIP and other nontraditional lipid parameters. HDL‐C has long been known to play a crucial role in cholesterol clearance. ²⁸ Previous studies have demonstrated that an elevated triglyceride/HDL‐C ratio, reflected by AIP, correlates with high concentrations of proatherogenic lipoprotein subclasses and the presence of small dense LDL particles. ²⁹ Moreover, AIP is strongly associated with insulin resistance and metabolic syndrome, both of which impair lipid metabolism and promote systemic inflammation. ³⁰ Hyperinsulinemia related to insulin resistance promotes the formation of small dense LDL, reduces HDL‐mediated cholesterol efflux, and triggers the release of inflammatory cytokines such as TNF‐α (tumor necrosis factor alpha), IL‐6 (interleukin‐6), and CRP (C‐reactive protein). ³¹ These processes accelerate endothelial dysfunction and early plaque formation, linking AIP mechanistically to atherosclerotic stroke. ³²

AC, a combination of non‐HDL‐C and HDL‐C, captures the atherogenic burden posed by multiple apolipoprotein B‐containing lipoproteins—namely, LDL, VLDL, intermediate‐density lipoprotein, and lipoprotein(a). Elevated non‐HDL‐C levels reflect the cumulative effect of these atherogenic particles, which contribute to foam cell formation, immune activation, and ultimately atherosclerotic progression. Intracranial plaques and stenoses are highly prevalent in fatal stroke cases, ³³ suggesting that high non‐HDL‐C may worsen AIS prognosis through mechanisms involving intracranial atherosclerosis and impaired cerebral perfusion.

LCI, as an integrative parameter derived from 4 conventional lipid markers (TC, triglyceride, LDL‐C, and HDL‐C), may reflect underlying lipid metabolism imbalances associated with chronic inflammation and plaque instability. It has been proposed that inflammatory conditions accelerate the crystallization of cholesterol within macrophages, leading to necrotic core formation, plaque rupture, and increased thrombogenicity—events highly relevant in the pathogenesis of AIS. ³⁴

When it comes to CRI‐II, calculated as LDL‐C/HDL‐C, the role of LDL‐C in lipid transportation and plaque development becomes evident. High LDL‐C levels contribute to LDL oxidation and foam cell formation, whereas low HDL‐C compromises reverse cholesterol transport, reducing cholesterol clearance from arterial walls. ³⁵ These mechanisms promote atherosclerotic plaque development and vascular narrowing, especially through intracranial atherosclerotic stenosis, which is a recognized cause of IS. ³⁶ As previously reported, CRI‐II is positively associated with stroke risk. Interestingly, in terms of stroke prognosis, several studies have found that CRI‐II may behave as a protective factor. ¹⁴ , ³⁷ , ³⁸ Our study similarly observed that CRI‐II was negatively associated with 1‐year mortality after AIS (Figure 2). This apparent paradox may be explained by a U‐shaped relationship between LDL‐C and poststroke mortality. A large‐scale cohort study involving 800 000 individuals demonstrated that patients with LDL‐C levels <1.00 mmol/L had a 2.22‐fold increased mortality compared with those with LDL‐C between 2.50 and 2.99 mmol/L— even higher than in groups with elevated LDL‐C levels. ³⁹

Emerging evidence also suggests that lipid metabolism may be involved in poststroke neural repair. For instance, secreted phospholipase A2G2E in the brain produces eicosapentaenoic acid and its metabolites, such as 15‐hydroxy‐eicosapentaenoic acid, which activate PAD4 (peptidyl arginine deiminase 4), a mediator of neural regeneration. ⁴⁰ This highlights the dual role of lipids not only in stroke risk but also in recovery processes.

Beyond 30 days, several studies of AIS have shown that the 1‐year readmission rate is maintained at around 30%. ⁴¹ , ⁴² This is quite different from the 12.76% readmission rate in our study. This discrepancy may be attributable to our exclusion of patients with severe coexisting conditions (eg, cancer, liver or kidney failure), who typically have a poorer prognosis and higher readmission risk. At the same time, some patients’ visits to other hospitals and hospitalizations cannot be ruled out. Compared with all‐cause death and readmission at 1 year, prolonged LOS is not an indicator of long‐term prognosis of patients, with AIS but an important evaluation criterion of medical level. Prolonged LOS is associated with comorbidities or hospital‐acquired infection. ⁴³ Previous studies have demonstrated that low cholesterol levels are a risk factor for sepsis. ⁴⁴ In the setting of infection, cholesterol levels may drop dramatically because of decreases in LDL‐C and HDL‐C. Additionally, both LDL‐C and HDL‐C play a proven role in the clearance of bacterial toxins, lipopolysaccharide from Gram‐negative bacteria, and lipoteichoic acid from Gram‐positive bacteria. ⁴⁵ This indicates that lipid abnormalities can prolong the LOS and bring unnecessary economic losses by mediating infection. Dyslipidemia can also cause coronary heart disease, fatty liver, diabetes, and other complications, which may aggravate the severity of AIS, which increases the difficulty of treatment and the length of hospital stay of patients with AIS. ¹¹ , ⁴⁶ , ⁴⁷

Indeed, dyslipidemia contributes to inflammation, atherosclerosis, and lipid toxicity, all culminating in AIS. Extensive research has probed the relationship between traditional lipid parameters—HDL‐C, TC, LDL‐C, and triglyceride—and stroke. LDL‐C is generally regarded as a primary contributor to cardiovascular disease, with a U‐shaped association observed between LDL‐C levels and mortality risk in AIS. ³⁹ Adults with both elevated triglyceride and reduced HDL‐C are at heightened risk for incident stroke. ⁵ However, research on nontraditional lipid parameters remains nascent. AIP level and cumulative exposure levels were associated with new strokes, poor functional outcomes, and early neurological deterioration development in patients with AIS. ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ CRI‐II was identified as a potential predictor of carotid plaque vulnerability and short‐term stroke prognosis at 3 months in AIS. ⁵² , ⁵³ Elevated non‐HDL‐C/HDL‐C ratios significantly increased the 1‐year risk of recurrent stroke in older patients. ⁵⁴ Moreover, there are few studies on LCI in the prognosis of AIS. Most studies focus solely on individual lipid parameters, lacking comprehensive evaluation and comparison. They also assessed the outcome only for a short period after the onset of AIS. Addressing these gaps, the current study comprehensively evaluated 11 lipid parameters concerning AIS prognosis. Our findings corroborate existing research, further underscoring the heightened predictive value of nontraditional lipid parameters, notably CRI‐II. The advantage of this study lies in its pioneering exploration of the relationship between nontraditional lipid parameters and AIS prognosis, along with a comprehensive lipid profile assessment via a PCA score. Additionally, our study unveiled a different trend. PCA analysis showed that the top 3 most important lipid parameters (AIP=lg [triglyceride/HDL‐C]; AC=[TC−HDL‐C]/HDL‐C; LCI=TC×triglyceride×LDL‐C/HDL‐C) among all nontraditional lipid parameters were mainly calculated by HDL‐C, triglyceride and TC (Figure 2). These findings underscore the importance for clinicians to monitor lipid levels in patients with AIS upon admission, with particular attention to nontraditional lipid parameters, especially CRI‐II. At the same time, for better clinical application, we developed an interactive web‐based nomogram to facilitate real‐time prognostic evaluation. Following standardization of patients’ nontraditional lipid parameters, clinicians can calculate PCA scores through our PCA algorithm (PCA score=36.27% AIP+35.89% AC+24.21% CRI‐I+25.40% CRI‐II+36.36% LCI+34.61% remnant cholesterol +27.86% non‐HDL‐C). This predictive tool enables mortality risk stratification by integrating multiple parameters via simple selection on the nomogram interface, providing instantaneous 1‐year mortality predictions. The visual quantification of composite risk factors through this platform offers clinicians an objective framework for risk stratification, ultimately informing personalized therapeutic decision‐making.

There are still some limitations in our study. First, the retrospective design of this study limited the ability to validate the causality between nontraditional lipid parameters and prognosis. Second, the study population was from 1 Chinese region, which may limit the generalization of outcomes. Also, this study was limited to patients with AIS without IVT, so we cannot know whether nontraditional lipid parameters have different prognostic associations in patients with AIS with IVT and without IVT. Third, due to the observational nature of the study, residual confounding factors, albeit adjusted for significant confounding variables, might persist, requiring cautious interpretation of causal relationships. Fourth, although apolipoprotein A1, apolipoprotein B, and lipoprotein(a) are also routine items for clinical lipid detection, we lack data on patients’ lipoprotein levels. This may have implications for the overall analysis of blood lipids in patients with AIS. Moreover, the lack of repeated lipid variable measurements precludes the exploration of lipid parameter fluctuations over time. Lastly, given the differences in regions, ethnicities, and dietary habits, caution should be exercised when generalizing our study findings. Due to data limitations, we are currently unable to obtain data from populations of other ethnicities and regions. Therefore, we hope that future studies can be conducted in diverse ethnic populations to verify the generalizability of our conclusions, and that prospective cohort studies can be carried out to further establish the causal relationship. This also provides a new perspective and foundation for future research on the association between nontraditional lipid parameters and prognosis.

CONCLUSIONS

Higher CRI‐II levels were associated with a lower risk of all‐cause mortality at 1 year and a longer LOS. This suggests that CRI‐II may be a relatively comprehensive predictor of AIS. Constructed composite PCA score for nontraditional lipids was associated with all‐cause mortality at 1 year, potentially providing physicians with additional prognostic information to help formulate treatment and rehabilitation for patients with IS.

Sources of Funding

This work was supported by the National Innovation and Entrepreneurship Training Program for College Students (No. 202410343059) and the Zhejiang Xinmiao College Students Innovation and Entrepreneurship Training Program, China (Grant number: 2025R413A039), both awarded to Wei Xie, and by the Wenzhou Municipal Sci‐Tech Bureau Program (No. Y20210585), awarded to Suwen Huang.

Disclosures

The authors declare that they have no competing interests.

Supporting information

Tables S1–S9

Figures S1–S3