Introduction

Despite advances in antiplatelets, lipid-lowering therapies, and advanced intervention techniques, atherosclerotic cardiovascular diseases (ASCVD) remain one of the leading causes of death worldwide1,2,3,4. The effect of aspirin on reducing ASCVD in secondary prevention settings is well established; however, its role as a primary preventive measure remains uncertain5.

Current guidelines recommend aspirin in individuals at high risk of future ASCVD in the primary prevention setting, yet the evidence remains unclear mainly due to its significant bleeding risk6,7,8. The recent ASPREE (In Aspirin in Reducing Events in Elderly)9, ASCEND (A Study of Cardiovascular Events in Diabetes)10, and the ARRIVE (Aspirin to Reduce Risk of Initial Vascular Events) trial investigated the overall benefit of aspirin in individuals without ASCVDs. However, all three trials failed to demonstrate benefit of aspirin due to increased bleeding events or failure to prove clinical benefit. Due to these reasons, there is uncertainty and discrepancy in aspirin recommendations for primary prevention across different guidelines due to the lack of clear evidence in the randomized controlled trial6,7,8,11. Aspirin is only recommended for high or very high-risk diabetes mellitus (DM) in the primary prevention settings in current European DM7 and cardiovascular (CV) disease prevention guidelines8. Additionally, the US guidelines no longer recommend routine aspirin prescription for primary ASCVD prevention, but suggest its use in adults aged 40–70 years who are at higher ASCVD risk but not at increased bleeding risk11,12. Given the bleeding risks associated with aspirin, an alternative agent with a more favorable bleeding profile may be considered for a safer option in primary prevention settings.

Sarpogrelate is a selective 5-hydroxytryptamine (5-HT)-2A receptor antagonist that inhibits thrombus formation, platelet aggregation, vasoconstriction, and vascular smooth muscle cell proliferation mediated by the 5-HT-2A receptor. Sarpogrelate has been used for the treatment of patients with peripheral arterial disease (PAD) in East Asia for nearly 20 years13,14,15,16,17. Sarpogrelate was studied in a small-sized randomized controlled trial for patients with recent stroke, the S-ACCESS (Sarpogrelate-aspirin comparative clinical study for efficacy and safety in secondary prevention of cerebral infarction) trial, to compare its effect with aspirin for secondary prevention13. The primary endpoint was recurrence of stroke. However, sarpogrelate was not able to satisfy the noninferior criteria compared to aspirin for the primary endpoint (hazard ratio [HR] = 1.25; 95% confidence interval [CI] 0.89 to 1.77; P = 0.19), although bleeding events were significantly lower (11.9% vs. 17.3%, P < 0.01). Interestingly, a post-hoc analysis suggested that sarpogrelate might have more CV benefits in diabetic patients compared to aspirin14.

Given this context, we hypothesized that sarpogrelate may be a viable alternative to minimize the bleeding risk associated with aspirin in DM patients with high risk. We thus analyzed real-world data from the Korean National Health Insurance System (NHIS) National Sample Cohort (NHIS-NSC) to assess the efficacy and safety of sarpogrelate versus aspirin in high-/very-high-risk DM patients in a primary prevention setting.

Results

Baseline characteristics

From 1 million individuals of the NHIS-NSC, we identified 133,534 individuals diagnosed with DM during 2010–2019 (Fig. 1). The database (NHIS-2024-2-039) used in our study was named as the SHIELD (Sarpogrelate Health Intervention for Early Lowering of cardiovascular risk in Diabetes) cohort. Of those, 29,414 took either sarpogrelate or aspirin as a single-antiplatelet therapy. We then identified 13,394 subjects who had taken sarpogrelate or aspirin for primary prevention after selecting those without an exclusion criterion. We further selected 10,778 patients who were eligible of high or very high-risk categories for DM, among which 9550 and 1228 were on aspirin and sarpogrelate, respectively. At baseline, the mean age was 62 years and body mass index (BMI)f was 25.7 kg/m2, while 87% had hypertension (Table 1). The mean follow-up duration was 77 ± 40 months. The prevalence of diabetic retinopathy, neuropathy, and nephropathy was 25.3, 28.9, and 15.0%. Most patients were classified as very high-risk DM (98.8%). There were statistically significant differences between the aspirin and sarpogrelate at baseline of the pre-matched analysis (Table 1). One to one propensity score matching (PSM) was conducted to balance the discrepancies between groups. After PSM, 1,228 matched subjects remained in each group. There were no differences between groups, except for a significantly longer duration of DM at baseline in the sarpogrelate group (65 vs. 74 months; p < 0.001) (Table 1).

Fig. 1
figure 1

Flow diagram for selection of study participants. NHIS-NSC, National Health Insurance System – National Sample Cohort; DM, diabetes mellitus; SHIELD, Sarpogrelate Health Intervention for Early Lowering of cardiovascular risk in Diabetes; GI, gastrointestinal.

Full size image
Table 1 Baseline characteristics.
Full size table

Most patients were assigned to the very-high-risk group. As indicated in Supplemental Table 1, only 18.8% of patients in the post-PSM cohort had diabetes of 10 years or more, which is key criteria for high-risk. In contrast, CV risk factors such as hypertension (75.7%), dyslipidemia (85.9%), older age (85.3%), and elevated BMI were prevalent in 70–80% of the cohort, with 55.9% exhibiting target organ damage. These results imply that, irrespective of the diabetes duration criterion for high-risk classification, most patients met the CV risk factor thresholds for very-high-risk diabetes, leading to their classification in the very-high-risk group7,8.

Outcome analysis

The outcome analyses are shown in Table 2. In the pre-matched analysis, there was no difference in net adverse clinical events (NACE) (11.5% vs. 17.6%; HR: 1.02, 95% CI 0.86–1.21) and other secondary endpoints except for overall bleeding. Sarpogrelate was associated with a 58% reduced risk (0.7% vs. 1.7%; HR: 0.42, 95% CI 0.22–0.83) in overall bleeding in the pre-matched analysis.

Table 2 Pre-matched and post-matched events by antiplatelet therapy.
Full size table

After matching, however, sarpogrelate was linked with a significantly lower incidence of NACE (11.5% vs. 18.4%; HR: 0.71, 95% CI 0.57–0.88) and efficacy endpoint (10.8% vs. 16.5%; HR: 0.75, 95% CI 0.60–0.94) (Fig. 2A–B). The improvement in NACE was mainly driven by 62% reduction in overall bleeding (0.7% vs. 2.0%; HR: 0.38, 95% CI 0.17–0.81) (Table 2).

Fig. 2
figure 2

Kaplan–Meier survival plots of event free survival for post-matched analyses. Post-matched analysis of aspirin versus sarpogrelate on NACE (A), efficacy endpoint (B), GI bleeding (C), and ICH (D). Matched, propensity score matched; NACE, net adverse clinical events; GI, gastrointestinal; ICH, intracranial hemorrhage.

Full size image

Outcomes of the secondary endpoints are shown in Table 2. Among the components of the efficacy endpoint, there was borderline significant reduction in all-cause death but no difference in myocardial infarction (MI) or stroke. As for the constituents with overall bleeding, post-PSM analysis indicated a lower incidence of gastrointestinal (GI) bleeding (0.7% vs. 2.5%; HR: 0.36, 95% CI 0.17–0.77) and intracranial hemorrhage (ICH) (0.2% vs. 1.1%; HR: 0.28, 95% CI 0.08–1.00) in the sarpogrelate group (Fig. 2C–D). Furthermore the sarpogrelate group demonstrated a significantly lower incidence of hospitalization due to bleeding (2.9% vs. 0.7%; HR: 0.29, 95% CI 0.13–0.62). Fatal bleeding events were also lower in the sarpogrelate group (0.4% vs. 0.1%), although the difference did not reach statistical significance (Table 2). There was also 35% reduction in transfusions (12.5% vs. 8.5%, Odds ratio (OR): 0.65, 95% CI 0.53–0.80) and 79% reduction in upper-GI hemostatic procedures (0.8% vs. 0.2%, OR: 0.21, 95% CI 0.05–0.85) in the sarpogrelate group from the total cohort (Supplemental Table 2).

We further conducted a sensitivity analysis. First, we incorporated the duration of diabetes into the PSM process to address differences in DM duration, the findings were nearly identical to the original analysis (Supplementary Table 3). In another sensitivity analysis, we excluded 444 individuals (4.1% of the SHIELD cohort) who were diagnosed with diabetes solely based on fasting blood glucose because this method alone is less accurate and may include individuals who do not truly have diabetes18. After this exclusion, the results remained consistent with the original findings (Supplemental Table 4). Sarpogrelate was associated with a 23% reduction in NACE (16.9% vs. 11.6%; HR: 0.77, 95% CI 0.62–0.96), primarily driven by a 73% reduction in overall bleeding (3.5% vs. 0.8%; HR: 0.27, 95% CI 0.13–0.56). There were no significant differences in the efficacy endpoint (15.2% vs. 11.0%; HR: 0.81, 95% CI 0.65–1.03), all-cause mortality (14.2% vs. 10.9%; HR: 0.88, 95% CI 0.69–1.11), or MI (0.4% vs. 0.2%; HR: 0.38, 95% CI 0.07–2.14). However, stroke was reduced by 72% (1.4% vs. 0.3%; HR: 0.28, 95% CI 0.09–0.83). GI bleeding decreased by 65% (2.1% vs. 0.6%; HR: 0.35, 95% CI 0.15–0.80), and ICH decreased by 85% (1.4% vs. 0.2%; HR: 0.15, 95% CI 0.03–0.68). Importantly, the significant reduction in overall bleeding was observed regardless of the type of sensitivity analysis.

Subgroup analysis

The subgroup analysis of the post-PSM cohort is shown in Fig. 3. Among individuals aged 65 or older or DM duration of 5 years or more, sarpogrelate had a significantly lower incidence of NACE (Pinteraction < 0.05 for both). There was no significant interaction in other subgroups. However, sarpogrelate showed a trend toward a lower hazard of NACE in men compared to women (Pinteraction = 0.069) and individuals with DM retinopathy compared with those without (Pinteraction = 0.124).

Fig. 3
figure 3

Subgroup analysis for NACE. NACE, net adverse clinical events; BMI, body mass index; DM, diabetes mellitus; LDL, low-density lipoprotein; HRadj, adjusted hazard ratio; CI, confidence interval. * Adjusted for age, sex, insulin therapy, HTN, DM duration, income, smoking status, BMI, statin therapy, LDL, HDL, TG, creatinine, and hemoglobin.

Full size image

Discussions

In this study, we analyzed the net clinical benefit of aspirin versus sarpogrelate in high-/very-high-risk diabetic patients in the primary prevention setting. Sarpogrelate was associated with a 29% reduction in NACE, comprising a 25% reduction in the composite efficacy endpoint and a 62% improvement in overall bleeding compared to aspirin. These results suggest that sarpogrelate may offer significant clinical benefits over aspirin in a high-/very-high-risk diabetic population, primarily driven by the reduction of bleeding (Fig. 4).

Fig. 4
figure 4

A summary of study. SHIELD, Sarpogrelate Health Intervention for Early Lowering of cardiovascular risk in Diabetes; NHIS, national health insurance system; NACE, net adverse clinical events; GI, gastrointestinal; ICH, intracranial hemorrhage; ASCVD, atherosclerotic cardiovascular events.

Full size image

The role for aspirin in primary ASCVD prevention is unclear. In the ASPREE trial, 20,000 people aged 70 years or older with moderate CV risk without ASCVD (or 65 years and older for certain ethnic groups with higher CV risks), atrial fibrillation, dementia or disability were enrolled. The primary endpoint was disability-free survival, defined as survival free from dementia or persistent physical disability. However, aspirin was associated with a higher all-cause death compared to placebo. Additionally, aspirin was associated with a 38% higher risk of major bleeding compared to placebo (HR: 1.38; 95% CI, 1.18–1.62), including GI bleeding (aspirin vs. placebo: 2.1% vs. 1.3%) and ICH (0.5% vs. 0.4%) over a median follow-up of 4.7 years9. The ASCEND trial enrolled 15,000 moderate-to-high ASCVD risk participants with DM but no evident ASCVD. The primary efficacy endpoint was a composite of vascular events, including MI, stroke, transient ischemic attack, and vascular death. The primary safety endpoint was major bleeding events, defined as a composite of eye, GI, or other serious bleeding events. Aspirin was associated with a 12% reduction in vascular events, but this benefit was offset by a 29% increase in major bleeding10. Similarly, the ARRIVE trial, which included 12,546 participants with moderate CV risk but no previous ASCVD, found no significant reduction in CV events with aspirin use while reporting a higher incidence of GI bleeding (HR, 2.11; 95% CI, 1.36–3.28)19. A meta-analysis of 13 trials, including 164,225 participants without ASCVD, similarly found that while aspirin use was associated with a significant reduction in CV events (HR, 0.89; 95% CI, 0.84–0.94), it was also linked to a 43% increased risk of major bleeding (HR, 1.43; 95% CI, 1.30–1.56)5. From these data, we learned that while the anti-ischemic effect of aspirin is clear, preventing bleeding events is crucial to maintain the overall benefit of aspirin.

Aspirin is increasingly recognized for its significant bleeding risks and the adverse effects that stem from them. Studies have shown that aspirin-related GI bleeding is not only common but also associated with high mortality rates. Saloheimo et al. demonstrated that aspirin use for primary prevention (N = 268, mean age: 68 years; follow-up duration: 3 months) increased the risk of mortality due to ICH20. The study also found that aspirin use was associated with a significant increase in intracranial hematoma enlargement during the first week of post-ICH. Considering that severe bleeding, such as hemoglobin drop of more than 5 g/L had a comparable risk of mortality with recurrent MI in patients with coronary artery disease. Moreover, ICH had a 4.5-fold higher mortality risk compared to recurrent MI21.

Sarpogrelate appears to be associated with a lower risk of bleeding, particularly upper-GI bleeding, compared to aspirin in diabetic patients with high CV risk. Sarpogrelate selectively inhibits serotonin-induced platelet aggregation without significantly affecting other pathways involved in hemostasis22,23,24. By selectively blocking this pathway, sarpogrelate reduces the risk of thrombus formation while maintaining a lower risk of bleeding. Aspirin, on the other hand, irreversibly inhibits cyclooxygenase (COX) in platelets, thereby reducing thromboxane A2 production, a potent promoter of platelet aggregation25. This inhibition is more broad-spectrum and affects multiple pathways in platelet function, which can increase the risk of bleeding. Additionally, aspirin can cause direct irritation to the stomach lining gastric mucosa (by inhibiting COX-1 and COX-2 in the stomach), which can lead to an increased risk of upper-GI bleeding5,26. This may explain why sarpogrelate use was associated with a lower risk of upper-GI hemostasis, while no significant difference was observed for lower-GI procedures in our data (Supplemental Table 3). Furthermore, diabetic patients often experience enhanced platelet activation and endothelial dysfunction, which can exacerbate bleeding risks with aspirin but may be mitigated by the targeted mechanism by sarpogrelate27,28,29. Supporting this, the S-ACCESS trial demonstrated that sarpogrelate was associated with significantly fewer bleeding events compared to aspirin while maintaining comparable efficacy in preventing recurrent cerebral infarction13. These results align with our findings showing that ischemic events were similar while bleeding events were markedly reduced. Similarly, ex vivo studies have shown that sarpogrelate’s selective inhibition of serotonin achieves platelet aggregation reduction comparable to aspirin, with less impact on bleeding time30. Future studies are warranted to confirm these mechanisms and evaluate their clinical relevance in broader diabetic populations.

The subgroup analysis highlights the potential benefits of sarpogrelate over aspirin in specific high-risk populations. Among individuals aged 65 or older, as well as those with a DM duration of 5 years or more, sarpogrelate demonstrated a significantly lower incidence of NACE. These findings are particularly important because both older age and longer duration of diabetes are well-established risk factors for bleeding with antiplatelet therapy7,31,32,33. Studies have shown that elderly patients face an increased risk of GI bleeding and other complications when treated with antiplatelets, due to age-related factors such as declining renal function and increased vascular fragility34. Similarly, patients with longer diabetes duration are more prone to both thrombotic and bleeding events, as prolonged hyperglycemia causes vascular damage and further complicates antiplatelet management35,36. Therefore, in primary prevention settings, sarpogrelate may be considered as an alternative to aspirin in patients with high bleeding risk, such as older adults and those with a long history of diabetes.

This study has several limitations. First, as this study is a retrospective observational analysis, causal inferences should be made cautiously despite the use of propensity score matching. In addition, International Classification of Diseases 10th Revision (ICD-10) codes may misclassify or not fully detect previous ASCVD due to its complexity and variability of ASCVD presentations, underscoring the need for well-designed randomized controlled trials to confirm these findings. Second, the study was limited to Korean ethnicity and high- or very-high-risk diabetic population. The majority of patients were categorized as very-high-risk due to the presence of comorbid CV risk factors, even in those without a diabetes duration exceeding 10 years. Caution is needed when generalizing the results to all diabetic patients or extending these findings to other ethnicities. Third, as some secondary outcomes, such as all-cause mortality and stroke, did not achieve statistical significance, extended studies with a larger sample size is required to confirm these findings. Fourth, the study did not include a comparison between patients receiving antiplatelet therapy and those who were not on any medication. This absence of a non-medication group limits the ability to fully evaluate the spectrum of potential benefits and harms associated with sarpogrelate or aspirin use. The exclusion of a non-medication group was primarily due to challenges in defining a consistent index date and the heterogeneity in diabetes risk classification among non-medication users. Lastly, we were not able to include the Bleeding Academic Research Consortium (BARC) classification in our analysis because the Korean NHIS data does not provide hemoglobin levels during the bleeding event, which are necessary for BARC-based bleeding classification. This limitation may hinder the precise categorization of bleeding severity, particularly for non-fatal but clinically significant bleeding events. As an alternative analysis, we categorized bleeding events based on hospitalization due to bleeding or fatal bleeding.

In conclusion, the use of sarpogrelate over aspirin in high- and very high-risk diabetic patients without ASCVD was associated with 29% improvement in net clinical benefit, primarily driven by a 62% decrease in overall bleeding. While some secondary outcomes, such as all-cause mortality and stroke, did not reach statistical significance, the trends observed suggest that sarpogrelate could help reduce bleeding complications in certain patient groups.

Methods

Data source and study population

We examined the NHIS–NSC from the Korean NHIS, which covers all residents living in Korea37,38,39. The NHIS-NSC is consisted of a sample of 1 million Korean individuals who were enrolled in health insurance and received medical benefits in Korea. The NHIS–NSC represents approximately 2.2% of the general Korean population and covers the period from 2002 to 2019.

We targeted individuals with DM who were receiving either sarpogrelate or aspirin as single-antiplatelet therapy between 2010 and 2019. Individuals with DM were defined as those who met at least one of the following criteria: use of DM medications, or fasting blood glucose levels of 126 mg/dL or higher40,41,42.

To ensure the study targeted primary prevention settings, we excluded those with a history of coronary artery disease ICD-10 codes: I20-I22), procedure codes for those who had undergone percutaneous coronary intervention, stroke (hospitalization due to ICD-10 code: I63-I64 and computed tomography or magnetic resonance imaging of the brain performed within 90 days before and after admission] or PAD (ICD-10 code: I73.9) were excluded from the study. Additionally, subjects with missing demographic data were excluded.

The study was approved by the Institutional Review Board of Gachon University Gil Medical Center, Incheon, Korea (GFIRB2023-459). Written informed consent was waived by the Institutional Review Board of Gachon University Gil Medical Center, Incheon, Korea, as this is a retrospective study of de-identified administrative data. All methods were performed in accordance with the relevant guidelines and regulations.

Definition and risk assessment of diabetes

Among the selected subjects, only those classified as high-risk or very-high-risk for diabetes were included7,8. The high-risk group was defined as individuals diagnosed with DM for 10 years or more and having at least one of the following conditions: age ≥ 50, hypertension, dyslipidemia (ICD-10 code E78, taking statins or low-density lipoprotein (LDL) cholesterol ≥ 130 mg/dL), current smoker, or BMI ≥ 25. The very-high-risk group was defined as individuals with target organ damage or having at least three of the following conditions: age ≥ 50, hypertension, dyslipidemia, current smoker, or BMI ≥ 257,8. Target organ damage was defined as having diabetic retinopathy (ICD-10 codes H36.0, E10.3, E11.3, E12.3, E13.3, or E14.3)43, diabetic neuropathy (a diagnosis of DN [ICD-10 codes E10.4, E11.4, E12.4, E13.4, E14.4, G59.0, G63.2, and G99.0] or receipt of a prescription for glucose-lowering drugs [ATC code A10B or A10A] plus drugs for diabetic neuropathy [ATC codes A16AX01, D11AX02, N06AA10, N06AA09, N03AX12, N03AX16, and N06AX21])44, or diabetic nephropathy (ICD-10 codes ICD-10 codes E10.2, E11.2, E12.2, E13.2, E14.2, or N08.3)45. The use of insulin at baseline was defined as insulin administration for more than 3 months within the year prior to the initiation of aspirin or sarpogrelate.

Aspirin and sarpogrelate therapy

We selected patients who had been on aspirin (100 mg daily) or sarpogrelate (either 100 mg twice daily or 300 mg once daily) single-antiplatelet therapy for at least 90 days. To ensure all participants were consistently on therapy throughout follow-up, we collected records of the dates and durations of drug prescriptions. We considered full adherence when the total number of prescribed days was equal to or greater than the follow-up duration. The adherence rate during follow-up was determined by calculating the proportion of total prescribed days relative to the total number of follow-up days.

Definition of other comorbidities and laboratory data

Hypertension was defined by diagnosis (I10), anti-hypertensive medication administration, or blood pressure ≥ 140/90 mmHg recorded during health checkup46,47. Data on blood pressure, fasting blood glucose, BMI, LDL cholesterol, high-density lipoprotein cholesterol, triglyceride, creatinine, and hemoglobin were obtained from the national health examination conducted within the two years prior to the specific drug initiation. LDL cholesterol was calculated based on the Friedewald Eq. 48,49,50 For subjects with missing values from the national health examination data, mean values according to each DM risk group were imputed. Statin therapy, including statin-ezetimibe single-pill combination therapy, was defined as its use for more than 3 months prior to the initiation of the antiplatelet.

Clinical follow-up and definition of endpoints

The primary endpoint was NACE, defined as a composite of efficacy and overall bleeding. The efficacy endpoint was a composite of all-cause death, MI, and stroke. Overall bleeding was defined as composite of ICH, and GI bleeding. Other individual secondary endpoints were all-cause death, MI, stroke, GI bleeding, and ICH. These events were identified using the ICD-10 and Medical Care Procedure codes: MI (I21), stroke (hospitalization with I63-I64 and CT/MRI performed within a 90-day window of hospitalization), ICH (hospitalization with I60-I62 or clinic visit for I60-I62 and transfusion), and GI bleeding. GI bleeding was defined as endoscopic hemostasis, vascular embolization combined with specific ICD-10 codes or transfusion accompanied by specific ICD-10 codes and endoscopic examination (E7611, EZ937, E7660, E7670, E7680) within 30 days before or after the transfusion. The ICD-10 codes for GI bleeding are K26.0, K26.2, K26.4, K26.6, K27.0, K27.2, K27.4, K27.6, K28.0, K28.2, K28.4, K28.6, K29.0, K62.0, K92.1, and K92.2. Both groups were followed until the onset of the first NACE event, or the end of study (December 31, 2019). Individuals who experienced a NACE event within 90 days of the initial drug prescription were excluded to minimize reverse causality. Those who took sarpogrelate or aspirin for less than 30 days were also excluded.

Hospitalization due to bleeding was defined as ICH or GI bleeding events that occurred during hospitalization. Fatal bleeding was defined as cases where ICH or GI bleeding was documented among patients who died during hospitalization. Deaths due to bleeding that occurred outside of a hospital setting were excluded from this analysis because they could not be accurately captured using Korean NHIS.

Statistical analysis

PSM was performed in a 1:1 ratio (caliper width: 0.2) to compare the effect on NACE51,52,53. In the pre-PSM cohort, key differences between the two groups were included as covariates. Covariates for matching included sex, age, hypertension, diabetic retinopathy, diabetic neuropathy, diabetic nephropathy, atrial fibrillation, heart failure, statin therapy, fasting blood glucose, BMI, and duration of DM.

Chi-square tests, t-tests, and ANOVA were performed to compare the baseline characteristics based on the intake of antiplatelet therapy. A Cox proportional hazard regression model was used to estimate the HR and 95% CI while adjusting for covariates. The survival rates of NACE were calculated using the Kaplan–Meier method by the intake of specific drugs. Statistical significance was set at p < 0.05. All statistical analyses were performed using SAS Enterprise version 8.3 (SAS Institute, Cary, NC, USA) and R version 4.3.0 (R studio, PBC).