Abstract
Background:
Cerebral hemorrhage has a high early mortality rate, and platelet function may be related to patient outcomes. However, the prognostic role of platelet aggregation function remains unclear.
Objective:
To evaluate the associations of arachidonic acid (AA)-, adenosine diphosphate (ADP)-, and collagen (COL)-induced platelet aggregation with 30-day mortality in patients with cerebral hemorrhage, and to determine whether these indicators provided additional prognostic information beyond conventional clinical variables and disease severity scores.
Methods:
This retrospective observational study enrolled adult patients with spontaneous intracerebral hemorrhage or traumatic cerebral hemorrhage admitted to the neurological intensive care unit (NICU) between September 2022 and April 2026. Platelet aggregation was assessed early after admission using light transmission aggregometry (LTA). The primary outcome was all-cause mortality within 30 days. Principal component analysis (PCA) was used to generate an integrated AA + ADP + COL PCA score. Candidate covariates were screened using least absolute shrinkage and selection operator (LASSO) regression, Boruta, and random forest methods, and the selected relevant variables were incorporated into multivariable logistic regression models. Supportive analyses examined discrimination, calibration, decision curve, incremental information beyond APACHE II, overlap-weighted estimates, and subgroup consistency.
Results:
Among 335 included patients, 78 died within 30 days. In the fully adjusted models, higher platelet aggregation values were associated with a lower risk of 30-day mortality for AA (OR 0.978, 95% CI 0.964–0.992, p = 0.002), ADP (OR 0.960, 95% CI 0.942–0.978, p < 0.001), COL (OR 0.960, 95% CI 0.943–0.977, p < 0.001), and the AA + ADP + COL PCA score (OR 0.613, 95% CI 0.494–0.759, p < 0.001). In supplementary model assessment, the fully adjusted models demonstrated good discrimination, with areas under the receiver operating characteristic curves (AUCs) ranging from 0.839 to 0.857, acceptable bootstrap calibration, and positive clinical net benefit across a wide range of decision thresholds. In supportive analyses, adding platelet aggregation measures to an APACHE II-only model was associated with higher AUCs and favorable NRI/IDI results. Subgroup analyses showed generally consistent results across the pre-specified subgroups. The associations were directionally consistent in extended adjustment and overlap-weighted sensitivity analyses.
Conclusion:
Reduced AA-, ADP-, and COL-induced platelet aggregation was associated with higher 30-day mortality after multivariable adjustment in patients with cerebral hemorrhage. Platelet aggregation testing may provide supportive prognostic information, but external validation and imaging-based confirmation were needed.
1 Introduction
Cerebral hemorrhage, including spontaneous intracerebral hemorrhage and traumatic cerebral hemorrhage, is a neurological emergency with high early mortality and substantial disability. Population-based studies have reported high case fatality and poor functional outcomes after intracerebral hemorrhage (1). Prognosis is difficult to assess early because it depends on several factors, including neurological severity, hematoma volume, and clinical management (2–4).
Platelets are important for primary hemostasis after vascular injury. Platelet dysfunction has been reported after traumatic injury and may contribute to hemostatic disturbance (5). Inherited and acquired platelet function disorders can also impair platelet activation and aggregation (6). These findings suggest that platelet function may provide information beyond platelet count alone.
Light transmission aggregometry (LTA) is a standard method for assessing platelet aggregation. Born described adenosine diphosphate-induced platelet aggregation, which formed the basis for optical platelet aggregation testing (7). Current recommendations support standardized LTA for platelet function assessment (8, 9). Aggregation induced by arachidonic acid (AA), adenosine diphosphate (ADP), and collagen (COL) is commonly used to assess different aspects of platelet reactivity (10).
In cerebral hemorrhage, previous studies have mainly focused on prior antiplatelet use and reversal strategies. In the PATCH trial, platelet transfusion did not improve outcomes in patients with antiplatelet-associated spontaneous intracerebral hemorrhage and was potentially harmful (11). In contrast, fewer studies have evaluated whether directly measured platelet aggregation function is associated with prognosis. Therefore, this study evaluated the associations of AA-, ADP-, and COL-induced platelet aggregation, both separately and as an integrated AA + ADP + COL principal component analysis (PCA) score, with 30-day mortality in patients with cerebral hemorrhage.
2 Materials and methods
2.1 Study design and population
This retrospective observational study screened adult patients admitted to the neurological intensive care unit of our hospital between September 2022 and April 2026. All patients had cerebral hemorrhage confirmed by computed tomography (CT) or magnetic resonance imaging (MRI). The cohort comprised both spontaneous intracerebral hemorrhage and traumatic intracerebral hemorrhage. Traumatic cerebral hemorrhage was defined as a head Abbreviated Injury Scale (AIS) score ≥3, AIS scores ≤2 in all other body regions, and no evidence of circulatory shock or acidosis. Blood samples for AA-, ADP-, and COL-induced platelet aggregation testing were collected at admission together with samples for other laboratory tests. Sampling preceded any therapeutic intervention that could substantially affect platelet function. Complete baseline data and 30-day outcome information were also required.
Patients were excluded if they met any of the following criteria: (1) long-term use of antiplatelet or anticoagulant drugs before admission; (2) treatment affecting platelet function or coagulation before blood sampling; (3) hematologic disease, recent radiotherapy or chemotherapy for malignancy, end-stage renal disease, or severe liver dysfunction; (4) severe infection, shock, disseminated intravascular coagulation, or massive blood loss; (5) missing or uninterpretable platelet aggregation results; (6) unavailable 30-day outcome information or loss to follow-up; or (7) pregnancy or puerperium.
A total of 388 patients met the initial eligibility criteria and entered study screening. Patients who died before LTA testing could be completed were not included in this screening population. Their number was not recorded separately. Of the 388 screened patients, 53 were excluded. The reasons for exclusion were hematologic disease, including leukemia (n = 2); end-stage renal disease or uremia (n = 5); aspirin use for cerebral infarction (n = 16); antiplatelet or anticoagulant therapy for atrial fibrillation, coronary heart disease, or coronary stent implantation (n = 13); antiplatelet therapy after splenic surgery or for other reasons (n = 4); transfer to another hospital, discharge against medical advice, or incomplete data within 24 h after admission with unavailable 30-day outcome information (n = 11); severe lipemia interfering with platelet aggregation testing (n = 1); and insufficient sample volume for testing (n = 1). Finally, 335 patients were included in the analysis. Of these, 257 survived for 30 days and 78 had 30-day mortality (Figure 1). The cohort included 221 men and 114 women, with a mean age of 57.6 years and a median hospital stay of 19.0 days.
Figure 1
The study was reviewed and approved by the hospital ethics committee (approval number: KYLW2026-066-01). Personal identifiers were removed before analysis, and all analyses used anonymized data. The requirement for informed consent was waived by the ethics committee because of the retrospective design and anonymized data.
2.2 Data collection and outcome grouping
Patients were categorized into survivor and non-survivor groups according to all-cause mortality within 30 days after admission. Baseline covariates were collected from the medical records and grouped by variable type. Demographic variables included age and Gender. Clinical history and disease-related variables included Hypertension, Diabetes Mellitus (DM), Hemorrhage_cause, Surgery, GCS, and APACHE II score. Vital signs included temperature, pulse rate, respiratory rate, SBP, DBP, and MAP. Routine blood and platelet-related indices included WBC, RBC, Hb, HCT, PLT_count, L_count, N_count, PDW, MPV, P-LCR, and PLT_PCT. Biochemical, electrolyte, lipid, myocardial enzyme, and coagulation variables included tBiL, DBiL, ALT, AST, ALB, BUN, SCr, Ca, Mg, phosphorus, K, Na, Cl, TG, TC, HDL, LDL, CK, CKMB, TT, PT, APTT, and Fib.
2.3 Platelet aggregation measurements
Platelet aggregation was measured with an AG800 automatic platelet aggregometer (Shandong Tailixin Medical Technology Co., Ltd., China) according to the manufacturer’s instructions. The test was performed using light transmission aggregometry, which measures changes in optical transmittance after platelet stimulation by specific agonists (7–10). Venous blood was collected into sodium citrate tubes at a blood-to-anticoagulant ratio of 9:1. Platelet-rich plasma was prepared by centrifugation at 160 × g for 10 min, with care taken to avoid red blood cell contamination. When the instrument temperature reached 37 °C, platelet-rich plasma was added to cuvettes containing a stir bar. AA, ADP, and COL were tested separately. For each agonist, the AG800 system generated an aggregation curve and reported the maximal aggregation rate (%). Higher values represented greater platelet aggregation.
2.4 Statistical analysis
Statistical analyses were conducted in R 4.4.0. Continuous variables were reported as medians with interquartile ranges, and categorical variables as counts with percentages. Group differences were assessed with the Mann–Whitney U test, chi-square test, or Fisher’s exact test, as appropriate. Statistical significance was defined as a two-sided p < 0.05.
The outcome was 30-day mortality. The primary exposures were AA-, ADP-, and COL-induced platelet aggregation. Because these measures were correlated, principal component analysis was used to derive an integrated AA + ADP + COL PCA score (12). Candidate covariates were selected based on published evidence, clinical relevance, and data availability. Because the number of outcome events was limited, we avoided entering all candidate covariates into a single multivariable model. AA, ADP, and COL were not included in the covariate selection procedure because they were the main exposure variables. APACHE II was excluded because it overlapped with several candidate covariates and could cause duplicate adjustment. Candidate covariates were subsequently screened using LASSO regression, the Boruta algorithm, and random forest (13–15), and the final adjustment set was determined by integrating statistical selection results, clinical relevance, and the event-number constraint.
Multivariable logistic regression was used to evaluate the associations of AA, ADP, COL, and the AA + ADP + COL PCA score with 30-day mortality. In the fully adjusted primary models, discrimination was assessed using ROC curves and AUC, calibration was assessed using bootstrap calibration curves, and clinical net benefit was assessed using decision curve analysis (16).
To evaluate the robustness of the findings, an extended adjustment model was further constructed by adding Hypertension, DM, Surgery, Hemorrhage_cause, and Gender to the fully adjusted model. In addition, AA, ADP, COL, and the AA + ADP + COL PCA score were separately added to an APACHE II-only model to assess additional prognostic information beyond APACHE II. Model comparisons were performed using the DeLong test, net reclassification improvement (NRI), and integrated discrimination improvement (IDI) (17, 18).
As an additional sensitivity analysis, propensity score-based overlap weighting was used to balance baseline candidate covariates (19). The primary regression analyses retained platelet aggregation as continuous exposures. AA, ADP, and COL were dichotomized according to the clinical reference cutoff provided by the laboratory of our center, with values <52% defined as low aggregation status and values ≥52% defined as normal aggregation status. Because no established clinical cutoff was available for the AA + ADP + COL PCA score, the PCA score was dichotomized using the ROC/Youden-derived cutoff. We also examined cohort-derived Youden cutoffs for AA, ADP, and COL in exploratory sensitivity analyses. The propensity score model included the baseline candidate covariates described above. We estimated propensity scores and overlap weights separately for each exposure and cutoff definition. Covariate balance after weighting was evaluated using standardized mean differences, with SMD < 0.1 indicating good balance. Weighted logistic regression was then used to evaluate the association between low aggregation status and 30-day mortality in the overlap-weighted population.
Finally, we performed subgroup analyses by Hemorrhage_cause, Hypertension, DM, Gender, and Surgery, using interaction terms to assess effect modification. As a sensitivity analysis, we compared logistic regression and random forest classifiers using repeated cross-validation.
3 Results
3.1 Baseline characteristics
After excluding patients with missing 30-day mortality status, 335 patients were included in the analysis, including 257 survivors and 78 non-survivors. As shown in Table 1, compared with survivors, non-survivors were older, had lower GCS scores, and had higher APACHE II scores. The three platelet aggregation indicators were all lower in the 30-day mortality group than in the 30-day survival group: AA was 30.45 (9.03, 40.98)% versus 42.70 (16.30, 58.20)%, ADP was 42.55 (25.07, 54.82)% versus 56.00 (43.40, 63.90)%, and COL was 43.18 (26.28, 53.84)% versus 57.39 (45.60, 65.60)% (all p < 0.001). Several laboratory variables also differed between groups, including PLT_count, PT, SCr, ALB, BUN, and APACHE II score, supporting the need for covariate adjustment in subsequent models.
Table 1
| Variable | Overall | 30-day survival | 30-day mortality | p-value |
|---|---|---|---|---|
| Gender | ||||
| Female | 114 (34.0%) | 88 (34.2%) | 26 (33.3%) | 0.991 |
| Male | 221 (66.0%) | 169 (65.8%) | 52 (66.7%) | 0.991 |
| Age | 58.00 (46.00, 71.00) | 55.00 (45.00, 68.00) | 68.00 (55.00, 77.75) | <0.001 |
| GCS | 8.00 (5.00, 12.00) | 10.00 (7.00, 13.00) | 5.00 (4.00, 8.00) | <0.001 |
| Hypertension | ||||
| No | 153 (45.7%) | 121 (47.1%) | 32 (41.0%) | 0.418 |
| Yes | 182 (54.3%) | 136 (52.9%) | 46 (59.0%) | 0.418 |
| Diabetes | ||||
| No | 294 (87.8%) | 231 (89.9%) | 63 (80.8%) | 0.051 |
| Yes | 41 (12.2%) | 26 (10.1%) | 15 (19.2%) | 0.051 |
| T | 36.50 (36.50, 36.80) | 36.50 (36.50, 36.80) | 36.50 (36.50, 36.80) | 0.875 |
| P | 80.00 (71.50, 95.00) | 80.00 (71.00, 93.00) | 83.00 (73.25, 97.00) | 0.177 |
| R | 18.00 (16.00, 20.00) | 18.00 (16.00, 20.00) | 17.00 (15.00, 18.00) | 0.083 |
| SBP | 162.00 (140.00, 183.50) | 161.00 (141.00, 182.00) | 167.50 (133.25, 187.75) | 0.673 |
| DBP | 93.00 (82.00, 105.00) | 93.00 (83.00, 105.00) | 93.50 (78.25, 104.75) | 0.274 |
| MAP | 117.00 (103.00, 131.00) | 117.00 (103.33, 129.67) | 119.50 (97.17, 133.92) | 0.854 |
| AA | 38.10 (13.79, 55.75) | 42.70 (16.30, 58.20) | 30.45 (9.03, 40.98) | <0.001 |
| ADP | 53.20 (40.00, 62.55) | 56.00 (43.40, 63.90) | 42.55 (25.07, 54.82) | <0.001 |
| COL | 54.53 (40.91, 64.14) | 57.39 (45.60, 65.60) | 43.18 (26.28, 53.84) | <0.001 |
| WBC | 13.83 (10.17, 18.26) | 13.41 (10.08, 17.82) | 15.71 (11.18, 19.46) | 0.063 |
| RBC | 4.47 (4.03, 4.92) | 4.49 (4.10, 4.95) | 4.37 (3.64, 4.76) | 0.042 |
| Hb | 137.00 (123.00, 150.00) | 138.00 (126.00, 152.00) | 133.00 (113.25, 144.75) | 0.020 |
| HCT | 40.70 (36.90, 44.50) | 40.80 (37.50, 44.70) | 40.00 (34.65, 44.00) | 0.049 |
| PLT_count | 195.00 (156.00, 237.00) | 198.00 (168.00, 239.00) | 170.50 (134.50, 228.50) | 0.004 |
| L_count | 1.20 (0.76, 1.79) | 1.14 (0.76, 1.75) | 1.29 (0.70, 1.81) | 0.991 |
| N_count | 11.73 (8.41, 16.01) | 11.44 (8.27, 15.67) | 12.94 (9.56, 16.97) | 0.079 |
| PDW | 16.00 (15.05, 16.40) | 16.00 (14.30, 16.40) | 16.25 (15.83, 16.60) | 0.004 |
| MPV | 9.80 (8.95, 10.60) | 9.90 (9.10, 10.60) | 9.80 (8.83, 10.67) | 0.948 |
| P-LCR | 25.20 (19.40, 30.80) | 25.30 (19.60, 30.40) | 24.20 (19.12, 31.32) | 0.839 |
| PLT_PCT | 0.19 (0.16, 0.23) | 0.20 (0.17, 0.23) | 0.17 (0.14, 0.21) | 0.003 |
| TBil | 12.60 (8.90, 16.75) | 12.60 (9.00, 16.50) | 12.25 (8.75, 17.38) | 0.764 |
| DBil | 4.50 (3.10, 6.60) | 4.50 (3.10, 6.40) | 4.65 (3.52, 6.70) | 0.448 |
| ALT | 19.00 (14.00, 27.00) | 19.00 (14.00, 28.00) | 19.00 (13.25, 26.00) | 0.506 |
| ALB | 41.50 (37.80, 44.85) | 42.10 (38.60, 45.30) | 40.00 (36.25, 42.75) | <0.001 |
| AST | 28.00 (20.00, 38.00) | 26.00 (20.00, 38.00) | 30.50 (20.75, 39.00) | 0.107 |
| BUN | 5.61 (4.71, 6.73) | 5.53 (4.56, 6.41) | 6.41 (5.15, 7.71) | <0.001 |
| SCr | 63.00 (52.00, 78.00) | 61.00 (50.00, 74.00) | 74.50 (58.00, 91.00) | <0.001 |
| Ca | 2.12 (2.02, 2.22) | 2.14 (2.05, 2.22) | 2.09 (1.98, 2.19) | 0.018 |
| Mg | 0.74 (0.68, 0.78) | 0.74 (0.68, 0.79) | 0.73 (0.68, 0.77) | 0.288 |
| Phos | 0.95 (0.76, 1.10) | 0.94 (0.76, 1.08) | 0.99 (0.80, 1.21) | 0.103 |
| K | 3.51 (3.17, 3.81) | 3.53 (3.23, 3.82) | 3.46 (3.06, 3.79) | 0.298 |
| Na | 139.00 (137.00, 141.00) | 139.00 (137.00, 141.00) | 138.00 (136.00, 140.00) | 0.063 |
| Cl | 103.00 (100.00, 105.00) | 103.00 (101.00, 105.00) | 102.00 (100.00, 104.00) | 0.050 |
| TG | 1.11 (0.68, 1.73) | 1.12 (0.69, 1.75) | 0.99 (0.67, 1.66) | 0.735 |
| TC | 4.32 (3.59, 4.98) | 4.35 (3.63, 5.03) | 4.03 (3.32, 4.87) | 0.156 |
| HDL-C | 1.26 (1.02, 1.51) | 1.27 (1.03, 1.52) | 1.15 (0.95, 1.47) | 0.302 |
| LDL-C | 2.46 (1.75, 3.00) | 2.50 (1.77, 3.08) | 2.27 (1.62, 2.74) | 0.048 |
| CK | 144.00 (84.00, 279.50) | 140.00 (83.00, 284.00) | 154.00 (99.75, 272.50) | 0.211 |
| CKMB | 24.00 (17.00, 42.00) | 23.00 (17.00, 37.00) | 31.00 (18.50, 63.75) | 0.001 |
| TT | 15.90 (14.15, 18.10) | 15.80 (14.10, 17.80) | 16.65 (14.45, 19.28) | 0.023 |
| PT | 11.10 (10.60, 12.00) | 10.90 (10.50, 11.70) | 11.55 (10.80, 13.00) | <0.001 |
| APTT | 25.90 (23.90, 28.20) | 25.70 (23.80, 28.00) | 26.50 (24.20, 29.17) | 0.071 |
| Fib | 2.45 (1.90, 2.97) | 2.45 (1.94, 2.97) | 2.41 (1.61, 3.00) | 0.122 |
| Surgery | ||||
| No | 78 (23.3%) | 56 (21.8%) | 22 (28.2%) | 0.307 |
| Yes | 257 (76.7%) | 201 (78.2%) | 56 (71.8%) | 0.307 |
| Hemorrhage_cause | ||||
| Traumatic | 132 (39.4%) | 97 (37.7%) | 35 (44.9%) | 0.319 |
| Non-traumatic | 203 (60.6%) | 160 (62.3%) | 43 (55.1%) | 0.319 |
| APACHE II | 14.00 (10.00, 18.00) | 13.00 (9.00, 17.00) | 18.00 (15.00, 20.00) | <0.001 |
Baseline characteristics according to 30-day mortality status.
Continuous variables are presented as median (interquartile range), and categorical variables are presented as n (%). Patients were grouped according to all-cause mortality within 30 days after admission. AA, arachidonic acid-induced platelet aggregation; ADP, adenosine diphosphate-induced platelet aggregation; COL, collagen-induced platelet aggregation; GCS, Glasgow Coma Scale; Hypertension; DM, diabetes mellitus; T, body temperature; P, pulse rate; R, respiratory rate; SBP, systolic blood pressure; DBP, diastolic blood pressure; MAP, mean arterial pressure; WBC, white blood cell count; RBC, red blood cell count; Hb, hemoglobin; HCT, hematocrit; PLT_count, platelet count; L_count, lymphocyte count; N_count, neutrophil count; PDW, platelet distribution width; MPV, mean platelet volume; P-LCR, platelet larger cell ratio; PLT_PCT, plateletcrit; TBil, total bilirubin; DBil, direct bilirubin; ALT, alanine aminotransferase; ALB, albumin; AST, aspartate aminotransferase; BUN, blood urea nitrogen; SCr, serum creatinine; Ca, calcium; Mg, magnesium; Phos, phosphorus; K, potassium; Na, sodium; Cl, chloride; TG, triglyceride; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; CK, creatine kinase; CKMB, creatine kinase-MB; TT, thrombin time; PT, prothrombin time; APTT, activated partial thromboplastin time; Fib, fibrinogen; APACHE II, Acute Physiology and Chronic Health Evaluation II.
3.2 Covariate screening
Covariate screening was performed using LASSO regression, random forest, and Boruta algorithm, and the final adjustment set was determined by integrating statistical selection, overlap across methods, clinical relevance, and the number of outcome events. The LASSO coefficient path and cross-validation curve supported a parsimonious covariate structure, and the random forest and Boruta results were provided as Supplementary Figures S1–S3. As summarized in Table 2, LASSO selected GCS, PT, Age, PDW, SCr, and CKMB at lambda.1se, while random forest ranked GCS, Age, PT, BUN, SCr, and PLT_count among the leading predictors, and Boruta confirmed GCS, Age, PT, BUN, Fib, SCr, PLT_count, MAP, SBP, RBC, TT, and Na as important variables. Based on these findings, Age, GCS, PT, SCr, and PLT_count were retained as the final covariates for the fully adjusted model (Supplementary Table S1).
Table 2
| Selection method | Selected variables |
|---|---|
| LASSO lambda.1se | GCS; PT; Age; PDW; SCr; CKMB |
| Random forest top 15 | GCS; Age; PT; BUN; SCr; PLT_count; Phos; P; PDW; Fib; TT; P_LCR; CKMB; RBC; SBP |
| Boruta confirmed | GCS; Age; PT; BUN; Fib; SCr; PLT_count; MAP; SBP; RBC; TT; Na |
| Final covariates | GCS; Age; PT; SCr; PLT_count |
Covariate screening and final adjustment set.
Final covariates were Age, GCS, PT, SCr, and PLT_count. LASSO, least absolute shrinkage and selection operator; GCS, Glasgow Coma Scale; PT, prothrombin time; SCr, serum creatinine; PLT_count, platelet count.
3.3 PCA and primary associations
Because AA, ADP, and COL were correlated, PCA was used to construct an integrated platelet aggregation score. Figure 2 showed that AA, ADP, and COL were positively correlated, with the strongest correlation observed between ADP and COL (Spearman r = 0.813, p < 0.001), followed by AA and ADP (r = 0.624, p < 0.001) and AA and COL (r = 0.591, p < 0.001). In the PCA loading plot, all three variables loaded in the same positive direction on PC1, with PC1 loadings of 0.528 for AA, 0.606 for ADP, and 0.595 for COL. PC1 explained 78.8% of the total variance, while PC2 explained 15.8%. These findings supported the use of the AA + ADP + COL PCA score as an overall platelet aggregation index. The associations of AA, ADP, COL, and the PCA score with 30-day mortality were then evaluated using stepwise logistic regression models (Table 3). In the fully adjusted Model 3, higher AA, ADP, and COL were associated with lower odds of 30-day mortality, with ORs of 0.978 (95% CI, 0.964–0.992; p = 0.002), 0.960 (95% CI, 0.942–0.978; p < 0.001), and 0.960 (95% CI, 0.943–0.977; p < 0.001), respectively. The AA + ADP + COL PCA score was also inversely associated with 30-day mortality (OR, 0.613; 95% CI, 0.494–0.759; p < 0.001), indicating that better overall platelet aggregation function was associated with lower mortality risk.
Figure 2
Table 3
| Exposure | Crude OR (95% CI) | P | Model 1 OR (95% CI) | P | Model 2 OR (95% CI) | P | Model 3 OR (95% CI) | P |
|---|---|---|---|---|---|---|---|---|
| AA | 0.980 (0.969, 0.991) | <0.001 | 0.981 (0.970, 0.993) | 0.002 | 0.977 (0.964, 0.991) | <0.001 | 0.978 (0.964, 0.992) | 0.002 |
| ADP | 0.958 (0.943, 0.973) | <0.001 | 0.961 (0.945, 0.976) | <0.001 | 0.959 (0.942, 0.976) | <0.001 | 0.960 (0.942, 0.978) | <0.001 |
| COL | 0.961 (0.947, 0.975) | <0.001 | 0.963 (0.949, 0.978) | <0.001 | 0.961 (0.945, 0.977) | <0.001 | 0.960 (0.943, 0.977) | <0.001 |
| AA + ADP + COL PCA | 0.612 (0.512, 0.731) | <0.001 | 0.634 (0.528, 0.761) | <0.001 | 0.614 (0.501, 0.751) | <0.001 | 0.613 (0.494, 0.759) | <0.001 |
Logistic regression models for platelet aggregation variables and 30-day mortality.
Model 1 adjusted for Age. Model 2 adjusted for Age and GCS. Model 3 adjusted for Age, GCS, PT, SCr, and PLT_count. OR, odds ratio; CI, confidence interval; AA, arachidonic acid; ADP, adenosine diphosphate; COL, collagen; PCA, principal component analysis.
3.4 Supplementary model assessment
We examined discrimination, calibration, and decision curve analysis for the four fully adjusted models as supplementary assessments. As shown in Figure 3A, all four models showed good discrimination, with AUCs of 0.839 (95% CI, 0.794–0.884) for the AA model, 0.856 (95% CI, 0.814–0.899) for the ADP model, 0.855 (95% CI, 0.812–0.898) for the COL model, and 0.857 (95% CI, 0.814–0.899) for the PCA model. Calibration was acceptable for all models (Figure 4), with Hosmer-Lemeshow p-values >0.05 and MAE values ranging from 0.038 to 0.045. Decision curve analysis further showed favorable clinical net benefit for all four models (Figure 3B). All four fully adjusted models yielded positive net benefit over a threshold probability range of 1–78%; the ADP, COL, and PCA models maintained positive net benefit up to 80%, whereas the AA model showed positive net benefit up to 78%.
Figure 3
Figure 4
3.5 Sensitivity analyses
Sensitivity analyses included both extended multivariable adjustment and overlap weighting. After further adjusting the fully adjusted model for Hypertension, DM, Surgery, Hemorrhage_cause, and Gender (Supplementary Table S3), the associations remained robust: AA (OR, 0.975; 95% CI, 0.961–0.990; p < 0.001), ADP (OR, 0.956; 95% CI, 0.938–0.975; p < 0.001), COL (OR, 0.953; 95% CI, 0.935–0.972; p < 0.001), and the PCA score (OR, 0.573; 95% CI, 0.456–0.721; p < 0.001) remained significantly associated with 30-day mortality. AA, ADP, and COL were dichotomized according to the clinical reference cutoff of 52% by our hospital’s laboratory department, with values <52% defined as low aggregation status and values ≥52% defined as normal aggregation status. For the PCA score, the ROC/Youden-derived cutoff was −0.136. This divided the cohort into a lower-score group (n = 142) and a higher-score reference group (n = 193). In the overlap-weighted sensitivity analysis, all assessed balance terms achieved SMD < 0.1 after weighting (52/52 balance terms for AA, ADP, COL, and PCA; Supplementary Figures S4–S7). Low aggregation status remained associated with increased 30-day mortality for AA (OR, 2.605; 95% CI, 1.444–4.697; p = 0.002), ADP (OR, 2.660; 95% CI, 1.533–4.614; p < 0.001), COL (OR, 3.102; 95% CI, 1.761–5.463; p < 0.001), and PCA (OR, 3.068; 95% CI, 1.757–5.357; p < 0.001) Supplementary Table S2. Additional exploratory analyses using cohort-derived Youden cutoffs were reported in Supplementary Table S4 and Supplementary Figure S11.
To explore whether platelet aggregation measures provided information beyond APACHE II, AA, ADP, COL, and the AA + ADP + COL PCA score were separately added to the APACHE II-only model. We compared each expanded model with the APACHE II-only model using the DeLong test, continuous NRI, and IDI (Table 4). Compared with the APACHE II-only model (AUC, 0.771), the AUC increased to 0.805 after adding AA, 0.821 after adding ADP, 0.826 after adding COL, and 0.826 after adding the PCA score. DeLong, NRI, and IDI analyses supported incremental prognostic information. Because external validation was not available, we interpreted these findings as supportive. The corresponding incremental ROC curves were provided in Supplementary Figure S8.
Table 4
| Model | AUC (95% CI) | Delta AUC | DeLong P | NRI (95% CI) | NRI P | IDI (95% CI) | IDI P |
|---|---|---|---|---|---|---|---|
| APACHE II model | 0.771 (0.716, 0.825) | 0.000 | NA | NA | |||
| APACHE II + AA | 0.805 (0.754, 0.856) | 0.035 | 0.014 | 0.360 (0.151, 0.597) | 0.002 | 0.042 (0.010, 0.089) | 0.043 |
| APACHE II + ADP | 0.821 (0.770, 0.871) | 0.050 | 0.011 | 0.688 (0.430, 0.905) | <0.001 | 0.093 (0.036, 0.167) | 0.008 |
| APACHE II + COL | 0.826 (0.777, 0.875) | 0.055 | 0.004 | 0.789 (0.447, 0.933) | <0.001 | 0.088 (0.033, 0.160) | 0.007 |
| APACHE II + AA + ADP + COL PCA | 0.826 (0.777, 0.875) | 0.055 | 0.004 | 0.575 (0.379, 0.868) | <0.001 | 0.093 (0.037, 0.170) | 0.006 |
Incremental predictive value of platelet aggregation variables beyond APACHE II.
The APACHE II-only model was used as the reference model. NRI and IDI were calculated using bootstrap resampling. APACHE II, Acute Physiology and Chronic Health Evaluation II; AUC, area under the receiver operating characteristic curve; NRI, net reclassification improvement; IDI, integrated discrimination improvement; PCA, principal component analysis.
3.6 Subgroup analysis
Subgroup analyses were performed according to Hemorrhage_cause, Hypertension, DM, Gender, and Surgery. The associations of AA, ADP, COL, and AA + ADP + COL PCA with 30-day mortality were generally consistent across prespecified subgroups. A statistically significant interaction was observed only for ADP by hypertension status (P for interaction = 0.047), whereas the other interaction tests did not indicate clear effect modification. The subgroup forest plots were presented in Supplementary Figure S9.
3.7 Classifier sensitivity analysis
Random forest did not consistently improve discrimination over logistic regression in repeated cross-validation, although its mean Brier scores were lower (Supplementary Table S5 and Supplementary Figure S10).
4 Discussion
In this retrospective cohort of patients with cerebral hemorrhage, we found that lower platelet aggregation induced by AA, ADP, and COL was associated with higher 30-day mortality. The associations remained after adjustment for age, GCS score, PT, SCr, and platelet count, and were similar in the extended and overlap-weighted analyses. The PCA score based on the three aggregation measures showed the same direction of association. This pattern suggested that reduced platelet responsiveness was related to early mortality in this cohort, although the observational design precludes causal inference.
This association has biological plausibility. Early outcomes following cerebral hemorrhage are affected by neurological severity and hemorrhage burden (20, 21). Platelets participate in the hemostatic response after vascular disruption through adhesion, activation, secretion, and aggregation. Impaired aggregation may reduce formation of a stable platelet plug at the bleeding site. A previous Stroke study reported that lower platelet activity was associated with early clot growth and poorer 3-month outcome after intracerebral hemorrhage (22). Our findings were in line with this report. However, hematoma volume and hematoma expansion were not consistently available, so we could not determine whether reduced platelet aggregation was associated with mortality directly or through greater hemorrhage burden.
A main difference from many earlier studies was that we measured platelet aggregation directly rather than relying on platelet count or antiplatelet history as indirect markers. Platelet count does not measure platelet reactivity, and patients with similar platelet counts may respond differently to agonist stimulation. AA-induced aggregation mainly reflects cyclooxygenase-thromboxane signaling, whereas ADP-induced aggregation reflects P2Y receptor-mediated activation (7, 8). COL-induced aggregation reflects collagen-related platelet activation, and light transmission aggregometry is a standard method for assessing these responses (9, 10). The consistent findings across the three agonists pointed to a broader reduction in aggregation capacity.
Coagulopathy after trauma can involve factor consumption, fibrinolysis, endothelial activation, and inflammation (23). Platelet dysfunction is also part of trauma-related hemostatic disturbance (24), and studies in severe traumatic brain injury have linked platelet dysfunction with hemorrhagic progression, transfusion requirements, and adverse outcomes (25, 26). Although spontaneous and traumatic cerebral hemorrhage differ in cause and management, both involve intracranial bleeding and early hemostatic stress. We therefore examined the associations separately by hemorrhage cause. All four point estimates were in the same direction in both groups (Supplementary Figure S9). However, the confidence intervals for AA and ADP crossed 1 in the traumatic group. We found no clear statistical interaction, but the small subgroups limited our ability to detect effect modification. These findings did not establish equivalent associations or shared mechanisms across the two etiologies. We also explored gender, hypertension, diabetes mellitus and surgery. In general, we did not observe clear evidence of interaction across these subgroups. The only exception was a weak interaction between ADP-induced aggregation and hypertension status (P for interaction = 0.047). Given the number of subgroup comparisons and the borderline p-value, this finding should be interpreted cautiously. Overall, these subgroup analyses did not provide strong evidence that the association differed across clinically relevant strata, although the limited sample size reduced the power to detect true effect modification.
The findings should be separated from studies of antiplatelet reversal and anticoagulant-associated hemorrhage. The PATCH trial showed that platelet transfusion did not improve outcomes in antiplatelet-associated spontaneous intracerebral hemorrhage and may be harmful (11). Outcomes also differ across types of anticoagulant-associated intracerebral hemorrhage (27). Our study addressed a different question. Patients with long-term antiplatelet or anticoagulant exposure were excluded, and endogenous platelet aggregation was measured early after admission. The results should not be read as support for empirical platelet transfusion or as evidence for managing antithrombotic-associated hemorrhage.
We used PCA to summarize the shared variation among the correlated AA, ADP, and COL measures (12). In this analysis, the first principal component loaded in the same direction for all three aggregation measures, and a higher PCA score represented higher overall platelet aggregation and lower mortality risk. We regarded the PCA score as a statistical summary of the shared information, rather than as a validated biological scale. The AA, ADP, COL, and PCA models were not intended to compete as prediction tools. Their AUC values were compared descriptively to check whether the association was consistent across different ways of representing platelet aggregation.
Several additional analyses were performed to examine the stability of the results. First, the associations remained after adding comorbidities, hemorrhage cause, and surgery to the fully adjusted model, suggesting that the results were not explained solely by these clinical factors. Second, overlap weighting was used as an alternative adjustment strategy. After weighting, baseline covariates were balanced, and low aggregation status remained associated with higher 30-day mortality (19). This finding reduced the possibility that the main results were driven by measured baseline imbalance, but did not completely eliminate it. Third, adding each aggregation measure to APACHE II improved AUC comparison by DeLong testing and showed favorable NRI and IDI results (17, 18). APACHE II is a general critical illness score rather than a hemorrhage-specific prognostic tool (28). Adding aggregation measures to APACHE II did not establish additional value beyond the ICH Score or imaging-based hemorrhage severity. Because this study lacked external validation, we interpreted these analyses as supportive rather than definitive evidence that platelet aggregation may add prognostic information. Given the limited number of outcome events, we avoided entering all candidate variables into one model. Covariate selection instead combined LASSO regression, Boruta, random forest screening, clinical judgment, and stepwise adjustment. This approach limited the number of variables in the primary models, but it did not guarantee freedom from overfitting or residual confounding (29, 30). Random forest and Boruta were initially used for covariate screening. We subsequently compared a random forest classifier with logistic regression in exploratory internal validation. Random forest did not consistently improve AUC, although it had lower mean Brier scores. We retained the primary association models and interpreted the new analyses as supportive.
Our single-center retrospective design left room for residual confounding, and the number of deaths constrained adjustment in the primary models. Hematoma volume was unavailable for adjustment, although it is an established predictor of early mortality (3). Inconsistent documentation of serial imaging also limited assessment of hematoma expansion and other imaging markers (31, 32). We could not determine whether lower platelet aggregation reflected greater initial hemorrhage burden, contributed to subsequent bleeding, or both. GCS adjustment and the supportive APACHE II analyses could not resolve this uncertainty, and the missing imaging data prevented us from reconstructing the ICH Score or testing incremental value beyond it. Although the associations persisted after extended adjustment that included hemorrhage cause, the subgroup analyses could not rule out differences between spontaneous and traumatic hemorrhage. Patients who died before LTA was completed were outside the 388-patient screening cohort; their number was not recorded separately. Collecting blood alongside other admission samples did not remove potential selection or survivorship bias, and our findings may not apply to patients with rapidly fatal hemorrhage. Platelet aggregation was measured only once at admission, and exclusion of patients with antiplatelet or anticoagulant exposure or conditions affecting testing limited generalizability. Analyses using cohort-derived cutoffs and the internal classifier comparison may have yielded optimistic estimates.
In summary, lower AA-, ADP-, and COL-induced platelet aggregation was associated with higher 30-day mortality in patients with cerebral hemorrhage. The PCA score showed a similar association, and platelet aggregation measures appeared to add prognostic information beyond APACHE II. These associations require confirmation in larger cohorts with standardized imaging data and external validation. They do not establish an effect independent of hemorrhage burden or a role for routine clinical risk assessment.
Statements
Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.
Ethics statement
The studies involving humans were approved by the Ethics Committee for Clinical Medical Research, Suzhou Ninth People's Hospital (Approval number: KYLW2026-066-01). The studies were conducted in accordance with the local legislation and institutional requirements. The ethics committee/institutional review board waived the requirement of written informed consent for participation from the participants or the participants' legal guardians/next of kin due to the retrospective nature of the study, which used anonymized data.
Author contributions
FY: Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. WL: Data curation, Investigation, Resources, Writing – original draft. RH: Conceptualization, Methodology, Project administration, Writing – original draft. CX: Conceptualization, Methodology, Project administration, Supervision, Writing – original draft. QS: Supervision, Validation, Writing – review & editing.
Funding
The author(s) declared that financial support was received for this work and/or its publication. This study was supported by the Suzhou Ninth People’s Hospital Institutional Research Project (YK202440) and the 2024 Excellence Program for Young and Middle-aged Talents at Suzhou Ninth People’s Hospital (Document No. [2024] 23). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Acknowledgments
The authors expressed their sincere gratitude to all the participants involved in this study.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that Generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneur.2026.1962864/full#supplementary-material
SUPPLEMENTARY FIGURE S1LASSO regression for covariate selection. Panel (A) shows the LASSO coefficient path across different values of lambda. Panel (B) shows the cross-validation curve used to determine the penalty parameter. LASSO, least absolute shrinkage and selection operator.
SUPPLEMENTARY FIGURE S2Random forest variable importance for candidate covariates. Variables were ranked by MeanDecreaseGini in the random forest model. The plot was used as one component of covariate screening and was interpreted together with LASSO regression, Boruta, and clinical relevance.
SUPPLEMENTARY FIGURE S3Boruta variable selection plot. Green boxes indicate confirmed important variables, yellow boxes indicate tentative variables, red boxes indicate rejected variables, and shadow variables represent permuted benchmark features used by the Boruta algorithm.
SUPPLEMENTARY FIGURE S4Covariate balance before and after overlap weighting for AA-induced platelet aggregation. Points show absolute standardized mean differences (SMDs) before and after overlap weighting between the low-aggregation (<52%) and normal-aggregation (≥52%) groups. Absolute SMDs <0.1 indicated acceptable covariate balance.
SUPPLEMENTARY FIGURE S5Covariate balance before and after overlap weighting for ADP-induced platelet aggregation. Points show absolute standardized mean differences (SMDs) before and after overlap weighting between the low-aggregation (<52%) and normal-aggregation (≥52%) groups. Absolute SMDs <0.1 indicated acceptable covariate balance.
SUPPLEMENTARY FIGURE S6Covariate balance before and after overlap weighting for COL-induced platelet aggregation. Points show absolute standardized mean differences (SMDs) before and after overlap weighting between the low-aggregation (<52%) and normal-aggregation (≥52%) groups. Absolute SMDs <0.1 indicated acceptable covariate balance.
SUPPLEMENTARY FIGURE S7Covariate balance before and after overlap weighting for the AA + ADP + COL PCA score. Points show absolute standardized mean differences (SMDs) before and after overlap weighting between the lower-score (<−0.136) and higher-score (≥−0.136) groups. The cutoff of −0.136 was derived using the ROC/Youden method. Absolute SMDs <0.1 indicated acceptable covariate balance.
SUPPLEMENTARY FIGURE S8Incremental ROC curves for platelet aggregation indicators added to APACHE II. The incremental predictive value of AA, ADP, COL, and AA + ADP + COL PCA was assessed by comparing each expanded model with the APACHE II-only model.
SUPPLEMENTARY FIGURE S9Subgroup analyses of AA, ADP, COL, and AA + ADP + COL PCA. Forest plots show subgroup-specific odds ratios and 95% confidence intervals according to Hemorrhage_cause, Hypertension, DM, Gender, and Surgery. Interaction P-values were used to evaluate effect modification.
SUPPLEMENTARY FIGURE S10Exploratory internal classifier comparison. Circles show out-of-fold AUC and Brier scores from 10 repeats of fivefold cross-validation. Diamonds show the means across repeats. Clinical included age, GCS, PT, SCr, and platelet count. Each other specification added the indicated aggregation measure.
SUPPLEMENTARY FIGURE S11Covariate balance using Youden-derived cutoffs. Panels (A–D) show covariate balance for AA, ADP, COL, and the AA + ADP + COL PCA score using their respective Youden cutoffs. Orange circles and green triangles indicate balance before and after overlap weighting, respectively. The dashed line marks an absolute standardized mean difference (SMD) of 0.1.
References
1.
van AschCJLuitseMJRinkelGJvan der TweelIAlgraAKlijnCJ. Incidence, case fatality, and functional outcome of intracerebral haemorrhage over time, according to age, gender, and ethnic origin: a systematic review and meta-analysis. Lancet Neurol. (2010) 9:167–76. doi: 10.1016/S1474-4422(09)70340-0. 20056489,
2.
HemphillJC3rdBonovichDCBesmertisLManleyGTJohnstonSC. The ICH score: a simple, reliable grading scale for intracerebral hemorrhage. Stroke. (2001) 32:891–7. doi: 10.1161/01.str.32.4.891. 11283388,
3.
BroderickJPBrottTGDuldnerJETomsickTHusterG. Volume of intracerebral hemorrhage: a powerful and easy-to-use predictor of 30-day mortality. Stroke. (1993) 24:987–93. doi: 10.1161/01.str.24.7.987. 8322400,
4.
GreenbergSMZiaiWCCordonnierCDowlatshahiDFrancisBGoldsteinJNet al. Guideline for the management of patients with spontaneous intracerebral hemorrhage. Stroke. (2022) 53:e282–361. doi: 10.1161/STR.0000000000000407,
5.
KutcherMERedickBJMcCreeryRCCraneIMGreenbergMDCacholaLMet al. Characterization of platelet dysfunction after trauma. J Trauma Acute Care Surg. (2012) 73:13–9. doi: 10.1097/TA.0b013e318256deab,
6.
GreselePHarrisonPBuryLFalcinelliEGachetCHaywardCPet al. Diagnosis of suspected inherited platelet function disorders: results of a worldwide survey. J Thromb Haemost. (2014) 12:1562–9. doi: 10.1111/jth.12650. 24976115,
7.
BornGVR. Aggregation of blood platelets by adenosine diphosphate and its reversal. Nature. (1962) 194:927–9. doi: 10.1038/194927b0. 13871375,
8.
CattaneoMCerlettiCHarrisonPHaywardCPMKennyDNugentDet al. Recommendations for the standardization of light transmission aggregometry. J Thromb Haemost. (2013) 11:1183–9. doi: 10.1111/jth.12231,
9.
HaywardCPMMoffatKARabyAIsraelsSPlumhoffEFlynnGet al. Development of north American consensus guidelines for medical laboratories that perform and interpret platelet function testing using light transmission aggregometry. Am J Clin Pathol. (2010) 134:955–63. doi: 10.1309/AJCP9V3RRVNZMKDS. 21088160,
10.
PanicciaRPrioraRLiottaAAAbbateR. Platelet function tests: a comparative review. Vasc Health Risk Manag. (2015) 11:133–48. doi: 10.2147/VHRM.S44469. 25733843,
11.
BaharogluMICordonnierCAl-Shahi SalmanRde GansKKoopmanMMBrandAet al. Platelet transfusion versus standard care after acute stroke due to spontaneous cerebral haemorrhage associated with antiplatelet therapy (PATCH): a randomised, open-label, phase 3 trial. Lancet. (2016) 387:2605–13. doi: 10.1016/S0140-6736(16)30392-0,
12.
RingnerM. What is principal component analysis. Nat Biotechnol. (2008) 26:303–4. doi: 10.1038/nbt0308-303,
13.
TibshiraniR. Regression shrinkage and selection via the lasso. J R Stat Soc B Stat Methodol. (1996) 58:267–88. doi: 10.1111/j.2517-6161.1996.tb02080.x
14.
KursaMBRudnickiWR. Feature selection with the Boruta package. J Stat Softw. (2010) 36:1–13. doi: 10.18637/jss.v036.i11
15.
BreimanL. Random forests. Mach Learn. (2001) 45:5–32. doi: 10.1023/A:1010933404324
16.
Van CalsterBWynantsLVerbeekJFMVerbakelJYChristodoulouEVickersAJet al. Reporting and interpreting decision curve analysis: a guide for investigators. Eur Urol. (2018) 74:796–804. doi: 10.1016/j.eururo.2018.08.038. 30241973,
17.
DeLongERDeLongDMClarke-PearsonDL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. (1988) 44:837–45. doi: 10.2307/2531595,
18.
PencinaMJD'AgostinoRBSrSteyerbergEW. Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers. Stat Med. (2011) 30:11–21. doi: 10.1002/sim.4085. 21204120,
19.
LiFMorganKLZaslavskyAM. Balancing covariates via propensity score weighting. J Am Stat Assoc. (2018) 113:390–400. doi: 10.1080/01621459.2016.1260466
20.
CordonnierCDemchukAZiaiWAndersonCS. Intracerebral haemorrhage: current approaches to acute management. Lancet. (2018) 392:1257–68. 30319113. doi: 10.1016/S0140-6736(18)31878-6,
21.
PuyLParry-JonesARSandsetECDowlatshahiDZiaiWCordonnierC. Intracerebral haemorrhage. Nat Rev Dis Primers. (2023) 9:14. doi: 10.1038/s41572-023-00424-7,
22.
NaidechAMJovanovicBLieblingSGargRKBassinSLBendokBRet al. Reduced platelet activity is associated with early clot growth and worse 3-month outcome after intracerebral hemorrhage. Stroke. (2009) 40:2398–401. doi: 10.1161/STROKEAHA.109.550939. 19443791,
23.
MooreEEMooreHBKornblithLZNealMDHoffmanMMutchNJet al. Trauma-induced coagulopathy. Nat Rev Dis Primers. (2021) 7:30. doi: 10.1038/s41572-022-00360-y,
24.
SimmonsJWPowellMF. Acute traumatic coagulopathy: pathophysiology and resuscitation. Br J Anaesth. (2016) 117:iii31–43. doi: 10.1093/bja/aew328. 27940454,
25.
NekludovMBellanderBMBlombackMWallenHN. Platelet dysfunction in patients with severe traumatic brain injury. J Neurotrauma. (2007) 24:1699–706. doi: 10.1089/neu.2007.0322,
26.
RiojasCMEkaneyMLRossSWCunninghamKWFurayEJBrownCVRet al. Platelet dysfunction after traumatic brain injury: a review. J Neurotrauma. (2021) 38:819–29. doi: 10.1089/neu.2020.7301. 33143502,
27.
InoharaTXianYLiangLMatsouakaRASaverJLSmithEEet al. Association of intracerebral hemorrhage among patients taking non-vitamin K antagonist vs. vitamin K antagonist oral anticoagulants with in-hospital mortality. JAMA. (2018) 319:463–73. doi: 10.1001/jama.2017.21917. 29372247,
28.
KnausWADraperEAWagnerDPZimmermanJE. APACHE II: a severity of disease classification system. Crit Care Med. (1985) 13:818–29. doi: 10.1097/00003246-198510000-00009,
29.
CollinsGSReitsmaJBAltmanDGMoonsKGM. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med. (2015) 162:55–63. doi: 10.7326/L15-0078-4,
30.
van SmedenMMoonsKGMde GrootJAHCollinsGSAltmanDGEijkemansMJCet al. Sample size for binary logistic prediction models: beyond events per variable criteria. Stat Methods Med Res. (2019) 28:2455–74. doi: 10.1177/0962280218784726. 29966490,
31.
DowlatshahiDDemchukAMFlahertyMLAliMLydenPLSmithEE. Defining hematoma expansion in intracerebral hemorrhage: relationship with patient outcomes. Neurology. (2011) 76:1238–44. doi: 10.1212/WNL.0b013e3182143317. 21346218,
32.
LiZYouMLongCBiRXuHHeQet al. Hematoma expansion in intracerebral hemorrhage: an update on prediction and treatment. Front Neurol. (2020) 11:702. doi: 10.3389/fneur.2020.00702. 32765408,
Summary
Keywords
30-day mortality, adenosine diphosphate, arachidonic acid, cerebral hemorrhage, collagen, platelet aggregation
Citation
Yin F, Lu W, Hu R, Xu C and Shi Q (2026) Association between early platelet aggregation function and 30-day mortality in patients with cerebral hemorrhage: a retrospective neurocritical care cohort study. Front. Neurol. 17:1962864. doi: 10.3389/fneur.2026.1962864
Received
09 August 2026
Revised
21 October 2026
Accepted
22 October 2026
Published
04 November 2026
Volume
17 - 2026
Edited by
Zilong Zhao, Tianjin Medical University General Hospital, China
Reviewed by
Ting Zhou, New York University, United States
Achmad Rafli, University of Indonesia, Indonesia
Updates
Copyright
© 2026 Yin, Lu, Hu, Xu and Shi.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Qiang Shi, kiddsq@163.com
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.