Apolipoprotein E Plasma Concentrations Are Predictive of Recurrent Strokes: Insights From the SPARCL Trial

Abstract

Background

In the SPARCL (Stroke Prevention by Aggressive Reduction in Cholesterol Levels) trial, atorvastatin (80 mg per day) was compared with placebo in patients with recent stroke or transient ischemic attack and no known coronary artery disease. Given the central role of apoE (apolipoprotein E) in lipoprotein metabolism and in the central nervous system, we assessed the contribution of apoE to subsequent cerebrovascular and cardiovascular events in this trial.

Methods and Results

ApoE concentrations and major isoforms (E2/E3/E4) were determined by liquid‐chromatography high resolution mass‐spectrometry in plasma samples collected at baseline from 4348 SPARCL participants. Patients in the lowest quartile were compared with those in the highest quartiles of apoE concentrations. Multivariable‐adjusted hazard‐ratios (HR) and 95% CIs were calculated using Cox proportional hazards regression models. We found a significant association between low apoE concentrations and the risk of recurrent strokes or cerebrovascular events (HR, 1.401 [95% CI, 1.154–1.701], P<0.001 and 1.467 [95% CI, 1.260–1.708], P<0.001) driven by a higher incidence of ischemic strokes and transient ischemic attacks in the entire cohort as well as separately in each treatment arm of SPARCL. In contrast, apoE concentrations did not significantly associate with the incidence of hemorrhagic strokes in SPARCL. We also found a significant association between reduced apoE concentrations and the risk of subsequent coronary events (HR, 1.373 ([95% CI, 1.064–1.772], P=0.015) in the entire cohort that was, however, significant only in the placebo arm of SPARCL.

Conclusions

Low apoE concentrations are predictive of recurring cerebrovascular events in patients with a history of stroke or transient ischemic attack.

Registration

URL: https://www.clinicaltrials.gov; Identifier: NTC00147602.

Nonstandard Abbreviations and Acronyms

SPARCL

Stroke Prevention by Aggressive Reduction in Cholesterol Levels

Clinical Perspective.

What Is New?

• Patients in secondary prevention of stroke or transient ischemic attack are at extreme risk of having a recurrent cerebrovascular event despite adequate treatment, in particular with high doses of statins; specific biomarkers associated with this risk are lacking.

• We here demonstrate that patients with low plasma concentrations of apolipoprotein E are at increased risk of recurrent ischemic strokes or transient ischemic attack.

What Are the Clinical Implications?

• Apolipoprotein E appears to be a more specific predictor of stroke recurrence than any of the circulating biomarkers investigated to date, underpinning the clinical potential of measuring apolipoprotein E plasma levels in patients with recent stroke or transient ischemic attack.

Individuals who had a stroke or a transient ischemic attack (TIA) are at extremely high risk of having a subsequent stroke. ¹ The identification of patients at risk of recurrent events relies on the assessment of traditional cardiovascular risk factors, but their ability to specifically predict cerebrovascular events remains disappointingly low. ² The SPARCL (Stroke Prevention by Aggressive Reduction in Cholesterol Levels) trial, a study conducted in 4731 participants with recent stroke or TIA, showed that 80 mg atorvastatin per day substantially reduced the risk of subsequent stroke. ³ Circulating biomarkers selected for their role in inflammation or oxidative stress were found associated with cerebrovascular outcomes in SPARCL, but these biomarkers had been selected based on their known association with coronary events ⁴ and thus lacked specificity for stroke. Thus, elevated plasma levels of oxidized phospholipids on apoB100 (apolipoproteinB100) containing lipoproteins equally predicted recurrent strokes and incident coronary events in SPARCL. ⁵ Another study recently showed that apoB/apoA‐I ratio, measured during the acute phase of the first stroke, positively correlated with the risk of stroke recurrence within 1 year. ⁶ In addition, lipoprotein (a) plasma concentrations did predict incident coronary events in the atorvastatin arm of SPARCL but not cerebrovascular events, including fatal or nonfatal strokes. ⁷ Specific biomarkers therefore remain to be identified to determine among stroke or TIA survivors which patients will have another cerebrovascular event.

As a ligand for low‐density lipoprotein (LDL) receptor family of proteins, apoE promotes the clearance of triglyceride‐rich lipoprotein remnants from the circulation and thereby plays a key role in cardiovascular health. ⁸ ApoE is the only apolipoprotein widely expressed in the central nervous system, and astrocyte‐derived apoE is crucial for cerebral metabolism by shuttling lipids between brain cells. There are 3 common apoE isoforms in humans, E2, E3, and E4, that differ by the presence of distinct amino acids at position 112 and 158. The approximate prevalence for E3/E3 is 58%, followed by E3/E4 (22%), E2/E3 (14%), E4/E4 (3%), E2/E4 (2%), and E2/E2 (1%). ⁸ Given that apoE2 has a weak affinity for LDL receptors, E2/E2 homozygotes can develop dysbetalipoproteinemia, a severe disorder characterized by concomitant elevations in plasma cholesterol and triglycerides. In contrast, E4/E4 homozygotes have a normal lipid profile but are at high risk of developing Alzheimer's disease. ⁹ , ¹⁰

Given the aforementioned roles of apoE in cardiovascular and brain health as well as recent evidence showing that elevated apoE concentrations associate with cardiovascular mortality whereas low levels associate with dementia‐related mortality in the population, ¹¹ we investigated whether apoE could constitute a specific biomarker of recurrent cerebrovascular events. We therefore explored this possibility by measuring the concentrations of apoE and determining the isoforms of apoE of each participant of the SPARCL trial.

METHODS

The data that support the findings of this study are available from the corresponding author upon reasonable request. All patients included in the present study provided written informed consent in agreement with the ethical guidelines of the 1975 Declaration of Helsinki under protocols approved by the research ethics committee or institutional review board of each participating SPARCL center.

Study Population

Details of the SPARCL study protocol, efficacy, and safety outcomes have been published previously. ³ , ¹² Briefly, 4731 individuals aged ≥18 years with no known coronary heart disease and LDL‐cholesterol (LDL‐C) plasma concentrations between 100 and 190 mg/dL (2.6 and 4.9 mmol/L) were enrolled 1 to 6 months after a noncardioembolic stroke or TIA. Patients were clinically followed for a double‐blind period of 53 (24–60) months. The primary outcome was time from randomization to a first nonfatal or fatal incident stroke. ³ Secondary outcomes included time to cerebrovascular events (defined as the occurrence of any ischemic stroke, hemorrhagic stroke, TIA, or carotid revascularization), time to coronary events (defined as the occurrence of any myocardial infarction, resuscitated cardiac arrest and other acute coronary events, coronary revascularization, unstable angina, and angina/ischemia requiring hospitalization), and time to peripheral events (defined as revascularization and peripheral artery disease). ³ , ¹² Of a total of 4731 SPARCL participants, 4348 (91.9%) had plasma collected at baseline (ie, at time of enrolment) available in sufficient quantities to simultaneously determine apoE isoforms and plasma concentrations by liquid‐chromatography high‐resolution mass spectrometry. Of these 4348 patients, 2180 (50.1%) were randomized to atorvastatin and 2168 (49.9%) to placebo. Presence of carotid stenosis was documented for 922 of the 4348 patients (21.2%). ¹³

ApoE Isoforms and Plasma Concentrations

Plasma samples were prepared using a fully validated method ¹⁴ involving trypsin proteolysis and subsequent analysis of proteotypic peptides on a Dionex Ultimate 3000 UHPLC System interfaced to Q Exactive Plus Orbitrap mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA) at DéTROI laboratory. The LGPLVEQGR peptide is unique in all apoE isoforms and was used to measure apoE plasma concentrations. For the determination of apoE isoforms, we assessed the presence of the following signature peptides CLAVYQAGAR, LAVYQAGAR, LGADMEDVR, and LGADMEDVCGR, as previously described. ¹⁴

Statistical Analysis

Statistical analyses were performed using R version 4.2.2 for Mac (Foundation for Statistical Computing, Vienna, Austria). Continuous variables are described as medians (interquartile ranges), and binary variables are expressed as counts and percentages. Qualitative and binary variables were compared using Fisher and Kruskal–Wallis tests, respectively. We used Kaplan–Meier curves to generate outcomes cumulative incidence during follow‐up. We fitted univariate and multivariate Cox proportional hazards models to compute hazard ratios (HRs [95% CI]) analyzing the relationship between outcomes and apoE plasma concentrations in the entire cohort. ApoE concentrations were dichotomized based on the first quartile (cutoff value: 3.62 mg/dL). Given that 80 mg atorvastatin starkly reduced the incidence of cerebrovascular, coronary, and peripheral vascular events in SPARCL, we systemically performed these analyses in the entire cohort as well as separately in the group randomized to atorvastatin and in the group randomized to placebo. Cox proportional hazards regression models were fitted with no adjustment (Model 1), after adjustment for age, sex, type of entry event (stroke or TIA), and time since entry event (in days as a continuous variable), as in all previous SPARCL analyses ³ , ¹² , ¹⁵ (Model 2), and additionally for all other potential confounders: body mass index, LDL‐C, high‐density lipoprotein cholesterol, triglycerides, arterial hypertension, smoking status, and diabetes at baseline (Model 3). In addition, we performed sensitivity analyses using Model 3 in the subgroup of patients homozygous for the apoE3 isoform (n=2982), as well as in the entire cohort by adding treatment arm as an adjustment variable to Model 3 (Model 4). Proportional‐hazards assumptions were tested graphically using scaled Schoenfeld residuals. Because similar results were obtained before and after all adjustments indicated, only HR (95% CI) and their P values determined with Model 3 are displayed. We also used inverse probability treatment weighting (IPTW) to adjust for differences in baseline covariates between apoE concentrations groups (cutoff value: 3.62 mg/dL). Potential confounders included in the logistic regression model used to estimate propensity scores included the following covariates at baseline: age, sex, smoking status, hypertension, diabetes, body mass index, high‐density lipoprotein cholesterol, LDL‐C, triglycerides, presence of carotid stenosis, type of entry event (ischemic stroke, hemorrhagic stroke, or TIA), time since entry event (in days), and the administration of placebo or atorvastatin. IPTW‐adjusted HR (95% CI) for all subsequent cardiovascular events were estimated using Cox proportional hazards regression models. Balances between groups were assessed by the standardized differences of all baseline covariates, using a threshold of 0.1 to indicate an imbalance. All standardized differences in all models were <0.1 (not shown). P values < 0.05 were considered statistically significant and all tests were 2 sided. As all analyses were considered exploratory, no correction for multiplicity was performed.

RESULTS

We measured apoE concentrations and major isoforms in plasma samples collected at baseline from 4348 participants of the SPARCL study. Their baseline demographic, clinical, and biological characteristics were undistinguishable from those of the 383 SPARCL participants for whom we did not have enough plasma left to measure apoE. The demographic, clinical, and biological parameters, in particular apoE concentrations and isoforms distribution, were indistinguishable between groups of patients later randomized either to atorvastatin (n=2180) or placebo (n=2168) (Table 1).

Table 1.

Baseline Characteristics and Cardiovascular End Points of the SPARCL Study Population

	All	Placebo	Atorvastatin 80 mg
	(n=4348)	(n=2168)	(n=2180)
Demographic
Age, y	64 [55–72]	64 [55–72]	65 [56–72]
Male sex, n (%)	2605 (59.9)	1284 (59.2)	1321 (60.6)
Body mass index (kg/m2)	26.9 [24.5–29.7]	26.8 [24.4–29.7]	27.4 [24.5–29.9]
Risk factors
Hypertension, n (%)	2681 (61.7)	1328 (61.3)	1353 (62.1)
Diabetes, n (%)	735 (16.9)	371 (17.1)	364 (16.7)
Current smoking, n (%)	826 (19.0)	418 (19.3)	408 (18.7)
Time since entry event, d	75 [45–120]	73 [44–118]	76 [45–122]
Type of entry event, n (%)
Stroke	3015 (69.3)	1488 (68.6)	1527 (70.0)
Transient ischemic attack	1333 (30.7)	680 (31.4)	653 (30.0)
Plasma lipids*
Total cholesterol, mg/dL	211 [192–230]	211 [191–230]	210 [192–230]
Low‐density lipoprotein cholesterol, mg/dL	132 [116–149]	132 [117–149]	132 [116–148]
High‐density lipoprotein cholesterol, mg/dL	48 [40–57]	48 [40–57]	48 [40–57]
Triglycerides, mg/dL	128 [96–170]	128 [97–170]	128 [95–170]
Apolipoproteins
ApoA‐I, mg/dL	146 [130–165]	146 [130–165]	145 [130–164]
ApoB100, mg/dL	133 [118–148]	133 [118–149]	132 [119–147]
ApoE, mg/dL	4.6 [3.6–5.8]	4.5 [3.6–5.8]	4.7 [3.7–5.8]
ApoE genotypes
E2/E2	9 (0.2)	4 (0.2)	5 (0.2)
E2/E3	401 (9.2)	196 (9.0)	205 (9.4)
E2/E4	74 (1.7)	33 (1.5)	41 (1.9)
E3/E3	2982 (68.6)	1489 (68.7)	1493 (68.5)
E3/E4	812 (18.7)	402 (18.5)	410 (18.8)
E4/E4	70 (1.6)	44 (2.0)	26 (1.2)
Coronary events	329 (7.6)	202 (9.3)	127 (5.8)
Myocardial infarction	140 (3.2)	85 (3.9)	55 (2.5)
Coronary revascularization	148 (3.4)	99 (4.6)	49 (2.2)
Coronary artery bypass graft	20 (0.5)	9 (0.4)	11 (0.5)
Percutaneous coronary intervention	131 (3.0)	93 (4.3)	38 (1.7)
Unstable angina	65 (1.5)	43 (2.0)	22 (1.0)
Angina requiring hospitalization	115 (2.6)	72 (3.3)	43 (2.0)
Cerebrovascular events	889 (20.4)	490 (22.6)	399 (18.3)
Stroke	564 (13.0)	301 (13.9)	263 (12.1)
Ischemic stroke	485 (11.2)	267 (12.3)	218 (10.0)
Hemorrhagic stroke	81 (1.9)	30 (1.4)	51 (2.3)
Transient ischemic attack	378 (8.7)	213 (9.8)	165 (7.6)
Carotid revascularization	69 (1.6)	41 (1.9)	28 (1.3)
Peripheral events	47 (1.1)	30 (1.4)	17 (0.8)

Values displayed as median [interquartile range] for continuous variables and as percentages for categorical variables.

^*Total cholesterol, low‐density lipoprotein cholesterol, high‐density lipoprotein cholesterol, and triglycerides.

^† xxx.

SPARCL indicates Stroke Prevention by Aggressive Reduction in Cholesterol Levels.

The median concentration of apoE in the entire cohort was 4.6 (3.6–5.8) mg/dL, in line with previous reports. ¹¹ The majority of participants were E3/E3 carriers (68.6%), followed by E3/E4 (18.7%), E2/E3 (9.2%), E2/E4 (1.7%), E4/E4 (1.6%), and E2/E2 (0.2%). Compared with the E3/E3 reference group, carriers of at least 1 apoE2 or 1 apoE4 isoform had respectively higher and lower apoE plasma concentrations (Figure S1). We stratified the whole SPARCL population and each treatment arm by quartiles of apoE plasma concentrations measured at baseline. Patients from the lowest apoE quartile (≤3.62 mg/dL) were more likely to be men, had significantly reduced body mass index and time since entry event. Their plasma lipids (total cholesterol, triglycerides, high‐density lipoprotein cholesterol, and LDL‐C), and apoA1 and apoB plasma levels were also reduced compared with patients from the highest apoE quartiles (>3.62 mg/dL) (Table S1).

We first evaluated the association between apoE plasma concentration and a first nonfatal or fatal subsequent stroke, the primary end point of SPARCL. ³ , ¹² Over a median follow‐up period of 53 (24–60) months, 564 strokes occurred in the entire study population (prevalence 13%), including 263 in the atorvastatin group (12.1%) and 301 in the placebo group (13.9%) (Tables 1 and 2). Overall, 889 cerebrovascular events occurred in the entire population (prevalence 20.4%) during follow‐up, including 399 in the atorvastatin group (18.3%) and 490 in the placebo group (22.6%) (Tables 1 and 2). Reduced apoE concentrations (≤3.62 mg/dL) were independently and significantly associated with an increased risk of subsequent stroke (HR, 1.401 [95% CI, 1.154–1.701], P<0.001) and cerebrovascular event (HR, 1.467 [95% CI, 1.260–1.708], P<0.001). Similar results were found separately in both the atorvastatin and placebo arms of SPARCL. Propensity scores analyses yielded similar IPTW‐adjusted HR for cerebrovascular events in the entire SPARCL cohort (Table S2). Kaplan–Meier curves established on the basis of the first versus the top 3 quartiles of apoE plasma concentration illustrate the significant associations between reduced apoE and cerebrovascular events in the entire cohort as well as in both treatment arms of SPARCL (Figure 1). These associations were driven by a significantly higher incidence of ischemic strokes, TIA, and carotid revascularization procedures in the quartile of patients with the lowest apoE plasma concentrations. Similarly, within the apoE3/E3 subgroup of patients, we found that reduced apoE concentrations were positively and significantly associated with an increased risk of subsequent stroke (HR, 1.559 [95% CI, 1.237–1.966], P<0.001) and cerebrovascular events (HR, 1.601 [95% CI, 1.331–1.924], P<0.001) in both treatment arms (Table S3 and Figure S2). Likewise, in the subgroup of patients with carotid stenosis at baseline, reduced apoE concentrations were positively and significantly associated with an increased risk of subsequent stroke (HR, 1.59 [95% CI, 1.07–2.36], P=0.022) and cerebrovascular events (HR, 1.65 [95% CI, 1.22–2.23], P=0.001) (Table S4). Of note the association of apoE concentrations with cerebrovascular outcomes was not dose dependent but rather displayed a threshold pattern (Figure S3).

Table 2.

Distribution of Coronary, Cerebrovascular and Peripheral Events by Quartiles of apoE Plasma Concentrations in SPARCL

	All patients			Atorvastatin			Placebo
	ApoE	(mg/dL)	P value	ApoE	(mg/dL)	P value	ApoE	(mg/dL)	P value
	>3.62 (Q2–Q4)	≤3.62 (Q1)		>3.62 (Q2–Q4)	≤3.62 (Q1)		>3.62 (Q2–Q4)	≤3.62 (Q1)
	(n=3261)	(n=1087)		(n=1654)	(n=526)		(n=1607)	(n=561)
Coronary events	236 (7.2)	93 (8.6)	0.175	92 (5.6)	35 (6.7)	0.412	144 (9.0)	58 (10.3)	0.378
Myocardial infarction	102 (3.1)	38 (3.5)	0.621	40 (2.4)	15 (2.9)	0.696	62 (3.9)	23 (4.1)	0.898
Coronary revascularization	103 (3.2)	45 (4.1)	0.148	37 (2.2)	12 (2.3)	1.000	66 (4.1)	33 (5.9)	0.106
Coronary artery bypass graft	16 (0.5)	4 (0.4)	0.795	10 (0.6)	1 (0.2)	0.414	6 (0.4)	3 (0.5)	0.896
Percutaneous coronary intervention	88 (2.7)	43 (4.0)	0.046.	27 (1.6)	11 (2.1)	0.612	61 (3.8)	32 (5.7)	0.072
Unstable angina	38 (1.2)	27 (2.5)	0.003	14 (0.8)	8 (1.5)	0.273	24 (1.5)	19 (3.4)	0.010
Angina	82 (2.5)	33 (3.0)	0.414	31 (1.9)	12 (2.3)	0.687	51 (3.2)	21 (3.7)	0.609
Cerebrovascular events	620 (19.0)	269 (24.7)	<0.001	288 (17.4)	111 (21.1)	0.066	332 (20.7)	158 (28.2)	<0.001
Stroke	392 (12.0)	172 (15.8)	0.001	189 (11.4)	74 (14.1)	0.124	203 (12.6)	98 (17.5)	0.005
Ischemic stroke	331 (10.2)	154 (14.2)	<0.001	158 (9.6)	60 (11.4)	0.251	173 (10.8)	94 (16.8)	<0.001
Hemorrhagic stroke	63 (1.9)	18 (1.7)	0.649	34 (2.1)	17 (3.2)	0.165	29 (1.8)	1 (0.2)	0.009
Transcient ischemic attack	262 (8.0)	116 (10.7)	0.009	121 (7.3)	44 (8.4)	0.487	141 (8.8)	72 (12.8)	0.007
Carotid revascularization	49 (1.5)	20 (1.8)	0.529	23 (1.4)	5 (1.0)	0.576	26 (1.6)	15 (2.7)	0.161
Peripheral events	54 (1.7)	23 (2.1)	0.389	20 (1.2)	8 (1.5)	0.742	34 (2.1)	15 (2.7)	0.548

Events are displayed as n (percentages). Values displayed as median [interquartile range] for continuous variables and as percentages for categorical variables.

^*xxx.

^† xxxx.

Figure 1. Cumulative incidence and hazard ratios of cerebrovascular events by levels of apoE plasma concentrations at baseline during 6 years following randomization.

Kaplan–Meier estimates for cerebrovascular events in all patients (A), in patients randomized to atorvastatin 80 mg (B), or placebo (C), with apoE concentrations below (Quartile 1) or above (Quartiles 2–4) 3.62 mg/dL at baseline. Insets show the same data on an enlarged y axis. D, Hazard ratios for composite cerebrovascular events and each major type of cerebrovascular event by levels of apoE concentrations at baseline in all patients and each treatment arms. Kaplan–Meier estimates and hazard ratios were adjusted for age, sex, type of entry event, time since entry event, body mass index, triglycerides, low‐density lipoprotein cholesterol, high‐density lipoprotein cholesterol, smoking, hypertension, and diabetes at baseline. apoE indicates apolipoprotein E; and HR, hazard ratio.

In contrast with all other subtypes of cerebrovascular events, apoE concentrations did not significantly associate with the incidence of hemorrhagic strokes in the entire cohort. There was, however, a negative association between reduced apoE concentrations and risk of hemorrhagic stroke in the placebo arm of SPARCL (HR, 0.095 [95% CI, 0.013–0.708], P=0.022) (Figure 1D). Among patients from the lowest apoE quartile, only 1 had a hemorrhagic stroke in the placebo group, whereas 17 did in the atorvastatin group during follow‐up (Table 2). Compared with the rest of the cohort, patients who had a hemorrhagic stroke during follow‐up were on average older (64 [55–72] and 70 [60–76] years old, P<0.001) and more likely to present with hemorrhagic stroke as qualifying entry event (0.6% versus 6.2%, P<0.001), respectively. None of the other baseline characteristics significantly vary between these groups except for the distribution of apoE isoforms: 23.3% of patients randomized to placebo who had a subsequent hemorrhagic stroke were carriers of an apoE2 isoform, compared with only 10.6% for those who did not (P=0.005). This difference was not seen in the atorvastatin group (9.8% versus 11.5%, respectively, P=0.88) (Figure S4).

We next evaluated the associations between apoE concentrations and all incident coronary events. During follow‐up, 329 coronary events occurred in the whole population (prevalence 7.6%), including 127 in the atorvastatin group (5.8%) and 202 in the placebo group (9.3%) (Tables 1 and 2). Reduced apoE concentrations (≤3.62 mg/dL) were positively and significantly associated with an increased risk of incident coronary events (HR, 1.373 [95% CI, 1.064–1.373], P=0.015). Similar results were found in both treatment arms of SPARCL, albeit not reaching statistical significance for the atorvastatin group (Figure 2). Propensity scores analyses yielded similar but nonsignificant IPTW‐adjusted HR for coronary events (Table S2). Kaplan–Meier curves established on the basis of the first versus the top 3 quartiles of apoE plasma concentration illustrate the significant associations between reduced apoE and coronary events in the entire study population as well as in the placebo arm (Figure 2). These associations were primarily driven by a significantly higher incidence of coronary revascularization procedures and unstable angina in the quartile of patients with the lowest apoE concentrations (Figure 2D). We performed a similar series of analyses in the subgroup of apoE3/E3 carriers and also found that reduced apoE concentrations were positively and significantly associated with an increased risk of incident coronary events in this entire subgroup (HR, 1.546 [95% CI, 1.144–2.089], P=0.005) as well as in the placebo arm of this subgroup (Table S3 and Figure S2). In contrast, in the subgroup of patients with carotid stenosis at baseline, reduced apoE concentrations were not significantly associated with an increased risk of incident coronary events (HR, 1.21 [95% CI, 0.73–2.01]), P=0.45) (Table S4).

Figure 2. Cumulative incidence and hazard ratios of coronary events by levels of apoE plasma concentrations at baseline during 6 years following randomization.

Kaplan–Meier estimates for cerebrovascular events in all patients (A), in patients randomized to atorvastatin 80 mg (B), or placebo (C) with ApoE concentrations below (Quartile 1) or above (Quartiles 2–4) 3.62 mg/dL at baseline. Insets show the same data on an enlarged y axis. D, Hazard ratios for composite coronary events and each major type of coronary event by levels of apoE concentrations at baseline in all patients and each treatment arms. Kaplan–Meier estimates and hazard ratios were adjusted for age, sex, type of entry event, time to entry event, body mass index, triglycerides, low‐density lipoprotein cholesterol, high‐density lipoprotein cholesterol, smoking, hypertension, and diabetes at baseline apoE indicates apolipoprotein E; CABG, coronary artery bypass graft; HR, hazard ratio; and PCI, percutaneous coronary intervention.

Lastly, during the follow‐up period, 47 peripheral vascular events occurred in the whole population (prevalence 1.1%), including 17 in the atorvastatin group (0.8%) and 30 in the placebo group (1.4%) (Tables 1 and 2). We did not find any significant association between apoE plasma concentrations and incident peripheral vascular events in the entire cohort (HR, 1.415 [95% CI, 0.840–2.382], P=0.192) as well as separately in each arm of SPARCL (Figure S5). Similar nonsignificant observations were made in the subgroup of apoE3/E3 carriers as well as in the subgroup of patients with carotid stenosis at baseline (Tables S3 and S4).

DISCUSSION

The SPARCL trial demonstrates that intense lowering of LDL with atorvastatin reduces the risk of subsequent strokes by 16% in patients with recent stroke or TIA and no known coronary heart disease. ³ The cerebrovascular protection conferred by atorvastatin in this trial reflected a 22% reduction in the prevalence of ischemic strokes but there was a 66% increased incidence of the much rarer hemorrhagic stroke type. In the present study, we have investigated the value of apoE as a predictive biomarker of subsequent cerebrovascular events in SPARCL. We found a significant association between low apoE concentrations and the risk of recurrent cerebrovascular events in both treatment arms of the trial. This association was primarily driven by the occurrence of ischemic strokes and TIA. We also found an association between reduced apoE concentrations and the risk of incident coronary events in SPARCL, albeit of a lower magnitude than that observed with the risk of subsequent strokes.

To our knowledge, this is the first report showing an association between apoE concentrations and the risk of recurrent cerebrovascular events in secondary prevention of stroke or TIA. One study that compared stroke patients with and without intracranial artery stenosis showed, in line with our results, that among all major lipoprotein biomarkers, reduced apoE concentrations associated with the presence of intracranial artery stenosis, an important factor of stroke recurrence. ¹⁶ However, in primary prevention settings, discrepant observations have been reported. Thus, in a cohort of 552 senior individuals, apoE levels positively associated with a higher risk of stroke independently of apoE isoforms. ¹⁷ In a cohort of 563 middle‐aged individuals, carriers of a genetic variant in the APOE gene promoter known to increase circulating apoE concentrations were also at significantly higher risk of ischemic stroke than noncarriers. ¹⁸ Likewise, elevated apoE concentrations were found associated with cardiovascular mortality in the Danish population, ¹¹ but unlike our exploratory analysis performed in secondary prevention, the Danish study did not specifically explore cerebrovascular events/mortality.

Several studies have shown that the apoE4 isoform confers an increased risk of ischemic stroke. ¹⁹ , ²⁰ , ²¹ Our sensitivity analysis conducted in the large subgroup of apoE3/E3 patients may appear suggestive of a direct effect of apoE concentrations independent of isoforms on the risk of stroke recurrence, but given that the apoE2 and E4 isoforms respectively increase and lower apoE concentrations, ²² our sensitivity analysis is insufficient to state that apoE isoforms are not important modifying factors. Mechanistically, the predisposition to ischemic stroke of patients with low apoE could stem from an impaired clearance of triglyceride rich lipoproteins fueling the atherosclerotic process, ²³ , ²⁴ but this is highly unlikely given that circulating lipids and apoB are lower in patients with reduced apoE levels in SPARCL. Alternatively, this could stem from inflammatory processes, given that apoE is a direct check point inhibitor of unresolvable inflammation. ²⁵

The significant increase in hemorrhagic strokes observed in the atorvastatin arm of SPARCL remains a matter of considerable debate, ³ even if aggressive LDL lowering has not proven causal of hemorrhagic strokes in subsequent long‐term trials. ²⁶ , ²⁷ , ²⁸ Our observation that low concentrations of apoE associate with a reduced risk of hemorrhagic stroke in the placebo but not in the atorvastatin arm of SPARCL merits to be investigated further in other trials, because the number of patients who had a hemorrhagic stroke in SPARCL is probably too small to draw any definitive conclusion in that matter. Another limitation is that we did not take into account the putative influence of antithrombotic treatment on any of the outcomes in our analyses. At this stage, we may only speculate on the potential mechanisms underpinning this observation, based on studies showing an increased permeability of the blood–brain barrier in apoE‐deficient and apoE4‐knock‐in mice. ²⁹ , ³⁰ , ³¹ , ³² Nevertheless, if this observation proves meaningful, measuring apoE concentrations will be an important factor to consider before prescribing high‐intensity statins to survivors of stroke.

We also found that reduced apoE concentrations associated with an increased risk of coronary events in SPARCL discrepant with large populations studies showing that apoE levels positively associate with ischemic heart disease and cardiovascular mortality. ¹¹ , ³³ The association between apoE and coronary events was primarily driven by unstable angina but not by myocardial infarction or revascularization procedures and was significant only in 1 treatment arm of SPARCL. These associations were nonsignificant in IPTW analyses for coronary events. Together these findings indicate that reduced apoE concentration is a more specific biomarker of recurrent cerebrovascular events than it is of subsequent coronary events in survivors of stroke/TIA. Considering that the number of peripheral events was extremely low in SPARCL, ¹² it is not surprising that only a nonsignificant but similar correlation between low apoE and subsequent peripheral vascular events was seen in this trial.

The present study has inherent limitations. All analyses were performed using a stratification of apoE concentration by quartiles and not as a continuous variable. Our results are also indicative of a threshold effect of apoE concentrations (ie, <3.62mg/dL) on cerebrovascular risk. Nevertheless, when comparing separately each quartile of apoE concentration to the first, we observed that patients with higher apoE concentrations developed significantly fewer cerebrovascular events and strokes, and to a lesser extent fewer coronary events (not shown). Furthermore, the present study is observational and as such does not prove causation, so we can only speculate on the pathophysiological mechanisms underlying the observed associations. Another limitation is that apoE plasma concentrations may not accurately reflect the levels nor the potential protective function of apoE produced locally in the brain, in particular during the acute phase of stroke. The dual role of apoE on both lipid metabolism and neurological processes, makes it difficult to specifically tease out the specific role of apoE in that respect. We cannot definitely rule out the possibility that unmeasured confounding factors might lead to potential misinterpretation of apoE role as a biomarker of recurrent strokes. Lastly, apoE concentration was measured only at baseline and its modulation over time, in particular in the atorvastatin arm of SPARCL, is unknown. Neither do we know how apoE concentrations may vary during the time course of the trial in function of other drug treatments or lifestyle interventions.

Conclusions

Despite these limitations, our study shows for the first time that reduced apoE levels are associated with recurrent cerebrovascular events in patients with a history of stroke of TIA. These findings provide evidence for a specificity of apoE as a biomarker of recurrent cerebrovascular events and underpin the clinical potential of measuring apoE concentrations in patients with stroke. It would now be extremely valuable to address the predictive power of adding apoE concentrations measurement as a prognostic biomarker of recurrent strokes and to compare its value to existing risk assessment tools.

Sources of Funding

Antonio Gallo received a postdoctoral fellowship SPF202004011793 from La Fondation pour la Recherche Médicale (Paris, France). This work was supported by a project grant ANR‐20‐CE14‐0009 from L'Agence Nationale de la Recherche (Paris, France).

Disclosures

Gilles Lambert reports consulting fees from Amgen. Michael J. Swietek is an employee of Pfizer Inc. AG reports consultancy fees and honoraria from Akcea Therapeutics, Amgen, Mylan, Novartis, Sanofi, Servier, Ultragenyx, Unilever, and MSD. The remaining authors have no disclosures to report.

Supporting information

Tables S1‐S4Figures S1‐S5