Serum Peroxiredoxin 6 Levels and Clinical Outcomes After Acute Intracerebral Hemorrhage in Elderly Patients: A Multicenter Observational Analytical Study

Introduction

Spontaneous intracerebral hemorrhage (ICH) represents only 10–15% of all strokes, but has a higher mortality and disability rate than cerebral infarction and spontaneous subarachnoid hemorrhage.^1,2 Moreover, the elderly are susceptible to ICH, thereby causing poorer clinical outcomes.³ Except that hematoma causes direct damage to the brain tissue, it triggers a series of molecular reactions, including inflammation, oxidative stress, necrosis, and apoptosis.⁴ Admittedly, the National Institutes of Health Stroke Scale (NIHSS) and hematoma amount, two conventional indices of ICH severity, are constantly applied to predict poor prognosis after ICH.^5,6 The modified Rankin Scale (mRS) is a useful evaluation system for the neurological functional status after ICH.^7–9 Early neurological deterioration (END) is often encountered and predisposes ICH patients to poor prognosis. Consequently, early and accurate prognostication of both poor prognosis and END is equally meaningful. Of note, biomarkers have been increasingly exploited for their prognostic significance in ICH, while there is scarce literature regarding the clinical relevance of biomarkers in elderly ICH.^10–12

Peroxiredoxins (Prxs) belong to the redox protease family and function as antioxidant enzymes.¹³ Peroxiredoxin 6 (Prx6), the only mammalian 1-Cys member in this family, can bind to thioredoxin to scavenge reactive oxygen species.¹⁴ Prx6 is more abundant in brain tissues than in other tissues and, therefore, plays an essential role in the pathological processes of numerous neuropsychiatric disorders, such as neurodegenerative diseases, brain aging, stroke, neurotrauma and more.^15,16 Notably, Prx6 may have dual effects that are dependent on the intracellular and extracellular environments. Specifically, intracellular Prx6, as an antioxidant enzyme, may be neuroprotective.^17–20 In contrast, extracellular Prx6 is likely to be a detrimental factor because significantly elevated Prx6 levels in the extracellular fluid of the ischemic brain in both rats and stroke patients were highly linked to the release of inflammatory cytokines from glial cells.^21–23 In the setting of traumatic brain injury, blood Prx6 levels were substantially heightened, and recovery of Prx6 activity indicated more favorable patient outcome.²⁴ Moreover, elevated serum Prx6 levels were highly associated with albumin quotient in patients with multiple sclerosis.²⁵ Hence, Prx6 may be a biomarker of brain injury.

As far as we are concerned, it is unclear about serum Prx6 levels following ICH and its prognostic role. Elderly patients generally experience poorer prognoses and more complex recovery trajectories due to comorbidities, frailty, and age-related physiological decline, therefore leading to the notion that the elderly with ICH should be a clinically highly noted population.²⁶ In mixed-age cohorts, the strong confounding effect of age itself may obscure the specific predictive value of Prx6, whereas studying an elderly cohort allows its contribution to be more clearly isolated. This multicenter observational analytical study was designed to demonstrate that serum Prx6 may provide reliable prognostic information, so as to improve risk stratification and optimize personalized management in this vulnerable population.

Materials and Methods

Study Design and Participant Selections

This multicenter observational analytical study was performed between March 2021 and March 2024 at Lishui Hospital of Wenzhou Medical University (Lishui, People’s Republic of China), Third Affiliated Hospital of Jiaxing University (Jiaxing, People’s Republic of China), Linping Campus, Second Affiliated Hospital of Zhejiang University School of Medicine (Hangzhou, People’s Republic of China), and Affiliated Hangzhou First People’s Hospital and Westlake University School of Medicine (Hangzhou, People’s Republic of China). This study was divided into two groups: the cross-sectional and prospective cohort studies (Figure 1). In the former study, elderly controls and patients (aged > 60 years) were enrolled in accordance with criteria in Supplementary Table 1. Blood samples of controls were collected at study entry, those of all patients were obtained at their hospital admission, and those at days 1, 3, 5, 7, 10, and 14 following ICH were required from those patients consenting for serial sampling, and subsequently, longitudinal changes in serum Prx6 levels after ICH were ascertained. In the latter study, admission serum Prx6 levels were detected among all patients to determine their predictive role for poor 6-month prognosis and END, as well as to unravel the mediating effect of END on the association of serum Prx6 levels with poor prognosis post-ICH.

Figure 1 Flow-chart of study design. This research was comprised of the prospective cohort study and cross-sectional study. Patients with intracerebral hemorrhage and controls were eligible for clinical study, in order to explore the dynamic changes of serum peroxiredoxin 6 levels following acute intracerebral hemorrhage, as well as to elucidate its associations with the disease severity, early neurological deterioration and neurological prognosis. Abbreviations: ICH, intracerebral hemorrhage; Prx6, peroxiredoxin 6.

Ethical Considerations

This study adhered to the guidelines of the Declaration of Helsinki and its subsequent guidelines. The research proposal was approved by the Ethics Committees of the Lishui Hospital of Wenzhou Medical University (opinion number: 2020–001), Third Affiliated Hospital of Jiaxing University (opinion number: 2021–076), Linping Campus, Second Affiliated Hospital of Zhejiang University School of Medicine (opinion number: 2021–045) and Affiliated Hangzhou First People’s Hospital, Westlake University School of Medicine (opinion number: 2020-058-01). All patients were all admitted to hospital within 24 hours after ICH, and were accompanied with consciousness disturbance or other neurological impairments, such as headache, vomiting, paralysis, aphasia, sensory disturbance and others, necessitating a good rest. Moreover, it is difficult to communicate with the elderly. Also, ICH patients are vulnerable to rebleeding and we made a great effort to decrease stimuli, so as to avoid rebleeding. So, informed consent forms were independently signed by the patient’s legal representatives after being informed of study details; and controls did it themselves.

Data Collections

At entry into the emergency room, patients or relatives were asked about (1) general demographic information, such as age, gender, body weight, and height; (2) certain chronic diseases, such as tobacco smoking, alcohol consumption, hypertension, diabetes, dyslipidemia, chronic obstructive pulmonary disease, ischemic heart disease, and hyperuricemia; and (3) specific medications, such as statins, antiplatelet agents, and anticoagulants. In addition, intervals between the onset of ICH symptoms and hospital arrival as well as between the onset of ICH symptoms and blood drawing were noted. Routinely, a noninvasive technique was employed to measure both systolic blood pressure and diastolic blood pressure. Head computed tomography (CT) was strictly performed in accordance with the radiological regulations. The amount of bleeding lesion was ascertained using the formula 0.5×a×b×c,²⁷ and the bleeding loci were categorized into superficial and deep cavities. We determined whether bleeding had expanded into the ventricular or subarachnoid space. To evaluate the severity of bleeding, the NIHSS was used upon admission.²⁸ END was considered when the NIHSS score increased by ≥ 4 points or death occurred within 24 h of admission.²⁹ At six months post-ICH, the mRS was used for meticulous evaluation of neurological functional status. Patients with the scores ranging from 3 to 6 were considered to have an adverse prognosis.⁸

Immune Analysis

Peripheral blood of controls was drawn at study entry. In compliance with the voluntary principle, some patients consented for blood drawings at admission and on days 1, 3, 5, 7, 10, and 14 after ICH, and a fraction of patients only proffered blood samples at admission. Through venipuncture within the antecubital region, blood specimens were rapidly deposited into 5 mL gel-containing biochemistry tubes (Ningbo Siny Medical Technology Co., Ltd., China). As soon as coagulation occurred, blood samples were centrifuged at 2000 × g for 10 min. Subsequently, the supernatants were carefully transferred into Eppendorf Tubes (Eppendorf Tubes® BioBased, China) and immediately stored at −80 °C refrigerator for subsequent analysis. To prevent protein decomposition, serum Prx6 levels were determined in batches with strict quality control measures. Every quarter, a batch of serum samples was melted to measure serum Prx6 levels using a commercially available Enzyme-Linked Immunosorbent Assay (ELISA). The kit was purchased from Abmart Shanghai Co., Ltd. (item no. ABK775601). The detection range was 0.312–20 ng/mL. Intra-assay and inter-assay variabilities were below 10% and 12% respectively. Optical density was measured at 450 nm using an ELISA reader (Biobase Meihua Trading Co., Ltd., China). All quantification procedures were performed twice by the same technician, who was blinded to the clinical details. The values obtained from the two measurements were averaged for subsequent statistical evaluation. Batch-to-batch calibration standards were to use standardized Prx6 supplied by the kit so as to reestablish standard curve for every batch of samples in accordance with manufacturer’s instructions. Moreover, there was no significant correlation between serum Prx6 levels and batches (P>0.05; Supplementary Figure 1), indicating that batches may not statistically affect serum Prx6 levels.

Statistical Analysis

Data analysis was performed using SPSS (version 23.0; SPSS Inc., Chicago, IL, USA). Following normality assessment by the Kolmogorov–Smirnov test, quantitative variables were expressed as means (standard deviations) or medians (lower and upper quartiles) as applicable. Categorical variables were summarized as counts (percentages). Data comparisons, bivariate correlation analysis and multivariate analyses were done according to Supplementary Table 2. Besides significantly different variables on univariate analysis (P<0.05), blood-collection time was included as a covariate in all regression models. The variance inflation factor was reported to reflect multicollinearity in regression analysis. To control the type I error rate, Bonferroni correction was applied for multiple comparisons. By using the MedCalc 20 (MedCalc Software, Ltd, Ostend, Belgium), the receiver operating characteristic (ROC) curve analysis was performed and corresponding area under ROC curve (AUC) was yielded. The optimal cutoff value for serum Prx6 levels was ascertained using the Youden method. By using the R 3.5.1 (https://www.r-project.org), subgroup analyses were carried out to evaluate interactions. Restricted cubic splines were configured to facilitate more nuanced exploration of the underlying relationships within the data. Independent factors associated with END and unfavorable prognosis were systematically incorporated into respective comprehensive combinatorial model. Subsequently, the nomogram, calibration curve, and decision curve were presented visually. A mediation analysis was performed to determine the mediating effect of END on the relationship between serum Prx6 levels and poor prognosis. E-values were calculated to assess the robustness of the results to unmeasured confounders. A type 1 error (alpha) of 0.05, a test power (1-beta) of 0.95, and an effect size of 0.8 were considered for sample size estimation of statistical analysis about serum Prx6 levels, covering intergroup comparisons and correlation analysis. The sample size was validated for its accuracy via a priori power analysis in G*Power 3.1.9.4 (Heinrich-Heine Universität Düsseldorf, Germany). As a result, the estimated least sample numbers for outcome variables, that is poor prognosis, END, and correlation analyses, were 136, 194 and 98 respectively. Thus, the sample size of 306 in our study was adequate for clinical analysis. Prospective cohort study is marked by a lower data missing. Only two cases were lost to follow-up in this study, leading to outcome data missing. Deletion method was adopted to handle value missing and the two cases were directly eliminated from this study. Differences were regarded as statistically significant when two-sided P-values were less than 0.05.

Results

Study Populations and Baseline Characteristics

A total of 421 elderly patients with ICH was consecutively recruited, and then 115 cases were excluded, leading to final inclusion of 306 patients (Supplementary Figure 2). A total of 103 elderly patients voluntarily continued to supply blood samples on days 1, 3, 5, 7, 10, and 14. In addition, 100 elderly controls were recruited. In Table 1, no significant differences were observed in age, gender, body mass index, tobacco smoking, and alcohol consumption among three groups: all 306 patients, 103 patients for serial samplings, and controls (all P>0.05). Furthermore, 103 patients for serial sampling showed similar ICH-relevant parameters, as well as demographic and clinical variables, as compared to overall patient cohort (all P>0.05; Table 1).

Table 1 Baseline Features of Elderly Patients with Acute Intracerebral Hemorrhage and Elderly Controls

Alteration of Serum Prx6 Levels Post-ICH

In patients with serial blood sampling, serum Prx6 levels spiked promptly upon admission, escalated to a notable concentration by the first day post-injury, peaked on day three, and subsequently declined gradually from the fifth to the fourteenth day following ICH (P<0.001; Figure 2 and Supplementary Table 3). Notably, compared with the control group, serum Prx6 levels in patients with ICH remained significantly elevated throughout the 14-day observation period (P<0.001; Figure 2 and Supplementary Table 3). In addition, serum Prx6 levels in all patients were markedly higher than those in controls (P<0.001; Supplementary Figure 3).

Figure 2 Dynamic change of serum peroxiredoxin 6 levels following intracerebral hemorrhage. Serum peroxiredoxin 6 levels showed a prompt incremental trend at admission, with the highest peak at day 3, and then gradually declined until day 14. The levels were markedly higher during 14 days in patients than in controls (P<0.001). Abbreviations: ICH, intracerebral hemorrhage; Prx6, peroxiredoxin 6.

Serum Prx6 Levels and ICH Severity

Among the patients with serial sampling, serum Prx6 levels at all time points were notably correlated with NIHSS scores and hematoma volume (all P<0.01; Supplementary Table 4). Furthermore, serum Prx6 levels at admission in all patients were significantly correlated with admission NIHSS scores (P<0.001; Supplementary Figure 4) and hematoma volume (P<0.001; Supplementary Figure 5). As illustrated in Table 2, aside from NIHSS scores and hematoma volume, other variables were significantly correlated with admission serum Prx6 levels in all patients, including diabetes mellitus, expansion of blood clots into the intraventricular space, and blood glucose levels (all P<0.05). Moreover, NIHSS scores and hematoma volume were independently correlated with serum Prx6 levels in all patients after ICH (both P<0.001; Table 3). Systematic deviation of residual was not observed according to Supplementary Figures 6 and 7, no evident heteroscedasticity pattern was shown in the residual-fitted value plot (Supplementary Figure 8), Cook’s distance ranged from 0.001 to 0.092 (mean, 0.004; standard deviation, 0.008), model R² equaled to 0.540 and Durbin-Watson value was 1.789.

Table 2 Factors in Connection with Serum Peroxiredoxin 6 Levels After Acute Intracerebral Hemorrhage

Table 3 Multivariate Linear Regression Analysis of Serum Peroxiredoxin 6 Levels and Variables in Patients with Acute Cerebral Hemorrhage

Serum Prx6 Levels and Post-ICH Six-Month mRS

mRS scores were substantially correlated with serum Prx6 levels at all time points in patients with serial sampling (all P<0.01; Supplementary Table 4). Admission serum Prx6 levels of all patients were closely related to their mRS scores (P<0.001; Supplementary Figure 9). On univariate correlation analysis of all patients, factors that were closely related to mRS scores included age, diabetes, intraventricular and subarachnoidal expansion of hematoma, NIHSS scores, hematoma size, serum Prx6 levels, END, and blood glucose levels (all P<0.05; Supplementary Table 5). Supplementary Table 6 shows that serum Prx6 levels, NIHSS scores, END, and hematoma volume were independently linked to mRS scores six months after acute ICH (all P<0.05). Residual was normally distributed based on Supplementary Figures 10 and 11, the residual‑versus‑fitted plot (Supplementary Figure 12) displayed no evident heteroscedastic pattern, Cook’s distance spanned from 0.001 to 0.072 (mean, 0.004; standard deviation, 0.009), model R² reached 0.529 and Durbin-Watson value was 2.071.

When patients for serial samplings were grouped into seven subgroups based on ordinal mRS, serum Prx6 levels at all time points were gradually elevated in order of mRS from 0 to 6 (P≤0.01; Supplementary Table 7). Among all 306 patients, admission serum Prx6 levels differed substantially among seven subgroups with different mRS scores (P<0.001; Supplementary Figure 13 and Supplementary Table 8). Additionally, age, diabetes mellitus, NIHSS scores, intraventricular extension of hematoma, subarachnoid extension of hematoma, hematoma size, END, blood glucose levels, and serum Prx6 levels were markedly different among seven subgroups from all patients (all P<0.05; Supplementary Table 9). Supplementary Table 10 shows that NIHSS scores, hematoma size, serum Prx6 levels, and END appeared as four independent predictors of ordinal mRS six months after acute ICH (all P<0.05).

Serum Prx6 Levels and Post-ICH Six-Month Functional Outcome

In patients with dynamic sampling, serum Prx6 levels at all time points were profoundly higher in those with poor prognosis than in those with good prognosis (all P<0.01; Supplementary Table 11). Moreover, admission serum Prx6 levels had AUC analogous to those at other time points (all P>0.05; Supplementary Table 11). Among all patients, admission serum Prx6 levels were markedly higher in patients with poor prognosis, as compared to those with good prognosis (P< 0.001; Supplementary Figure 14). Admission serum Prx6 levels greater than 27.21 ng/mL predicted poor prognosis with the maximum Youden index under ROC curve (Figure 3). A linear relationship was observed between serum Prx6 levels at admission and likelihood of poor prognosis (P for nonlinear >0.05; Figure 4). Significantly different variables between subgroup with poor prognosis and that with good prognosis among all patients encompassed age, diabetes mellitus, extensions of hematoma into intraventricular and subarachnoidal cavities, END, NIHSS scores, hematoma size, blood glucose levels and serum Prx6 levels (all P <0.05; Table 4). In Table 5, NIHSS scores, hematoma volume, serum Prx6 levels, and END were independent predictors of poor prognosis (all P<0.05).

Table 4 Factors Related to Poor Prognosis at Six Months After Acute Intracerebral Hemorrhage

Table 5 Multivariate Logistic Regression Analysis of Six-Month Poor Prognosis and Variables in Patients with Acute Intracerebral Hemorrhage

Figure 3 The discriminatory role of serum peroxiredoxin 6 levels in predicting adverse prognosis six months post - acute intracerebral hemorrhage. Under the receiver operating characteristic curve, serum peroxiredoxin 6 levels demonstrated a remarkable capability to discriminate the risk of adverse prognosis after acute intracerebral hemorrhage, and the optimal cutoff value was determined in accordance with the Youden method. Abbreviations: AUC, area under curve; 95% CI, 95% confidence interval. Notes: Red arrow points to cutoff value.

Figure 4 Restricted cubic spline analysis elucidating the linear association between serum peroxiredoxin 6 levels and the probability of adverse outcomes in intracerebral hemorrhage. The concentration of serum peroxiredoxin 6 demonstrated a linear relationship with the risk of poor prognosis six months after intracerebral hemorrhage (P for nonlinear > 0.05). Abbreviation: Prx6, peroxiredoxin 6.

Serum Prx6 levels at admission of all patients were analogous to those of the other three independent predictors (all P>0.05; Figure 5), and did not exhibit any substantial interactions with age, gender, hypertension, or other surrogate variables (all P interaction >0.05; Supplementary Figure 15). In Supplementary Figure 16, E-value was 1.5653 (95% confidence interval, 1.4061–1.7237). The model integrating the four independent predictors of poor prognosis was pictorially depicted as a nomogram (Figure 6). Furthermore, compared with the predictive performance of serum Prx6 levels, NIHSS scores, and hematoma size, the model demonstrated a remarkable enhancement in predictive efficacy, as measured by the AUC (All P<0.05; Figure 5). The model exhibited remarkable stability based on calibration curve analysis (Figure 7), and manifested outstanding clinical applicability via decision curve analysis (Figure 8).

Figure 5 Receiver operating characteristic curve analysis of poor prognosis following intracerebral hemorrhage. A combined model was constructed to predict adverse outcomes six months after acute intracerebral hemorrhage. Among all the variables investigated, their combination showed a relatively high ability to predict a poor prognosis (all P<0.05). Abbreviations: Prx6, peroxiredoxin 6; NIHSS, National Institutes of Health Stroke Scale; AUC, area under the curve. Notes: *P<0.05, **P<0.01.

Figure 6 Nomogram evaluating predictive model for the six-month poor prognosis of individuals experiencing acute cerebral hemorrhage. Serum peroxiredoxin 6 levels, hematoma volume, and National Institutes of Health Stroke Scale scores were integrated to discriminate the probability of functional decline six months subsequent to acute cerebral hemorrhage. Abbreviations: NIHSS, National Institutes of Health Stroke Scale; Prx6, peroxiredoxin 6; END, early neurological deterioration.

Figure 7 The reliability of the nomogram in predicting the probability of adverse prognosis six months after intracerebral hemorrhage. With the aid of calibration curve analysis, it is possible for the combined model to forecast poor neurological consequences six months after the onset of acute intracerebral hemorrhage.

Figure 8 Clinical applicability of the combined model in estimating the probability of worse outcomes six months after intracerebral hemorrhage. Using the decision curve, the developed model exhibited remarkable clinical utility in forecasting adverse outcomes six months after the onset of acute intracerebral hemorrhage. Abbreviations: NIHSS, National Institutes of Health Stroke Scale; Prx6, peroxiredoxin 6; END, early neurological deterioration.

Serum Prx6 Levels and END

Serum Prx6 levels at all time points in patients with serial samplings were significantly higher in individuals with END than in those without (all P<0.05; Supplementary Table 12). Moreover, serum Prx6 levels at admission were in possession of AUC comparable to those at other time points (all P>0.05; Supplementary Table 12). Among all patients, admission serum Prx6 levels were significantly higher in patients with END than in those without END (P<0.001; Supplementary Figure 17). Serum Prx6 levels greater than 28.09 ng/mL predicted END after ICH with the maximum Youden index (Figure 9). A linear correlation was revealed between serum Prx6 levels and END probability (P for nonlinear >0.05; Figure 10). As listed in Table 6, patients with END, as opposed to the other remainders, exhibited notably increased NIHSS scores, hematoma volume, blood glucose levels, and admission serum Prx6 levels, as well as markedly heightened proportions of diabetes mellitus and extensions of hematoma into the intraventricular and subarachnoid cavities (all P<0.05). Moreover, NIHSS score, hematoma volume, and admission serum Prx6 levels were independently predictive of END (all P<0.05; Table 7).

Table 6 Factors in Relation to Early Neurological Deterioration After Acute Intracerebral Hemorrhage

Table 7 Multivariate Logistic Regression Analysis of Early Neurological Deterioration and Variables in Patients with Acute Cerebral Hemorrhage

Figure 9 Efficiency assessment of serum peroxiredoxin 6 as a predictive biomarker of early neurological deterioration following acute intracerebral hemorrhage. In the framework of the receiver operating characteristic curve, serum peroxiredoxin 6 levels showed a marked capability to discriminate the risk of early neurological deterioration after acute intracerebral hemorrhage, and the optimal cut-off value was determined in accordance with the Youden method. Abbreviations: AUC, area under the curve; 95% CI, 95% confidence interval. Notes: Read arrow points to cutoff value.

Figure 10 Restricted cubic spline evaluating linearity relationship between serum peroxiredoxin 6 levels and possibility of early neurological deterioration following acute cerebral hemorrhage. There was a linear correlation between serum peroxiredoxin 6 levels and the probability of early neurological deterioration in this cohort of patients with acute cerebral hemorrhage (P for nonlinear >0.05). Abbreviation: Prx6, peroxiredoxin 6.

The predictive ability of serum Prx6 levels was equivalent to those of the other two predictors (both P>0.05; Figure 11). No substantial interactions were existent between serum Prx6 levels and other surrogate variables, such as age, gender, hypertension and so on (all P interaction >0.05; Supplementary Figure 18). In Supplementary Figure 19, E-value reached 1.4973 (95% confidence interval, 1.3154–1.6655), The model encompassing the three independent predictors of END showed significantly higher AUC than any of the three independent predictors (all P<0.05; Figure 11), was visualized as a nomogram (Figure 12), possessed an acceptable goodness of fit through calibration curve analysis (Figure 13), and exhibited an impactful clinical benefit when employing the decision curve evaluation (Figure 14).

Figure 11 Utility of combined model in predicting early neurological deterioration post-ICH. By employing receiver operating characteristic curve analysis, the combined model outperformed the individual variables in predicting early neurological deterioration following acute intracerebral hemorrhage (all P<0.05), showing its potential clinical utility. Abbreviations: Prx6 stands for peroxiredoxin 6; NIHSS, National Institutes of Health Stroke Scale; AUC, area under the curve. Notes: *P<0.05, **P<0.01.

Figure 12 Nomogram visualizing the combined model of early neurological deterioration in acute intracerebral hemorrhage. The three determinants of early neurological deterioration after acute intracerebral hemorrhage in this cohort, i.e., serum peroxiredoxin 6 levels, National Institutes of Health Stroke Scale scores, and hematoma volume, corresponded to designated scoring points. The summed total scores were used to assess the risk of early neurological deterioration following acute cerebral hemorrhage. Abbreviations: NIHSS, National Institutes of Health Stroke Scale; Prx6, peroxiredoxin 6; END, early neurological deterioration.

Figure 13 Clinical stability of the combined model of early neurological deterioration in acute intracerebral hemorrhage. The serum peroxiredoxin 6 levels, National Institutes of Health Stroke Scale scores, and hematoma volume, which were the three independent predictive factors of early neurological deterioration following acute intracerebral hemorrhage, were merged to form a combination model. This model showed good clinical stability under the calibration curve. Abbreviation: END, early neurological deterioration.

Figure 14 Clinical validity of the combined model of early neurological deterioration in acute intracerebral hemorrhage. The model, in which serum peroxiredoxin 6 levels, National Institutes of Health Stroke Scale scores, and hematoma volume were incorporated, in contrast to any of the above three components, had a relatively high clinical benefit in predicting early neurological deterioration after acute intracerebral hemorrhage under the decision curve. Notes: NIHSS, National Institutes of Health Stroke Scale; Prx6, peroxiredoxin 6.

Mediation Effect of END on Prognostic Association of Serum Prx6 Levels

Mediation analysis was performed to assess whether the association between serum Prx6 levels and likelihood of poor prognosis was partially mediated by END. As delineated in Figure 15, the association of admission Prx6 levels with probability of poor prognosis was partially mediated by END, and END accounted for a mediation ratio of 38.9% of this association, and sensitivity parameters were relatively high (both ρ=0.4; Supplementary Figure 20), indicating that END may in part interpret link between serum Prx6 levels and poor 6-month prognosis.

Figure 15 Mediating analysis determining mediating effect of early neurological deterioration following acute cerebral hemorrhage. The association between admission serum peroxiredoxin 6 levels and the probability of poor prognosis was partially mediated by early neurological deterioration, which accounted for a mediation ratio of 38.9%. Prx6 indicates peroxiredoxin 6; β, beta; β indirect1, effect of serum peroxiredoxin 6 levels on early neurological deterioration; β indirect2, effect of early neurological deterioration on poor prognosis; β direct, effect of serum peroxiredoxin 6 levels on poor prognosis; β indirect, β indirect1 multiplied by β indirect2; β total, β indirect plus β direct; mediation effect, β indirect divided by β total.

Discussion

To the best of our knowledge, changes in serum Prx6 levels after ICH remain unclear. In this study, serum Prx6 levels in ICH elderly patients exhibited a substantial and rapid escalation subsequent to ICH; serum Prx6 levels in patients were strongly associated with both NIHSS scores and hematoma volume; an independent association was discerned between serum Prx6 levels and poor clinical outcome observed six months after the ICH event; serum Prx6 levels and combined models showed effective prognostic prediction ability; and END partially mediated the association between serum Prx6 levels and poor prognosis. Overall, serum Prx6 level may be a potential prognostic biomarker for ICH in the elderly population.

The inflammatory response following ICH serves as a crucial pathological foundation and has been implicated in the ICH process.³⁰ Inflammation regression is considered an essential endogenous route capable of rescuing host cells, tissues, and organs from ongoing or overly intense inflammatory responses.³¹ Prx6 is an antioxidant enzyme and naturally confers neuroprotection inside brain cells.¹⁸ However, a large amount of intracellular Prx6 may be released into the extracellular environment of the brain tissue during the acute phase of cerebral ischemia, which is likely to activate infiltrating macrophages, induce the generation of inflammatory cytokines, and in turn drive the inflammatory response after cerebral ischemia.³² In a mouse model of middle cerebral artery occlusion, infarct area was significantly smaller in animals treated with anti-Prx than in their untreated counterparts.²² Collectively, Prx6 may be a detrimental factor in extracellular milieu. Blood-brain barrier is usually disrupted following ICH,³³ therefore leading to possible release of Prx6 from damaged cells to peripheral blood. In the present study, serum Prx6 levels were markedly elevated following ICH in the elderly, further supporting the conception that serum Prx6 may be a biomarker of brain injury.

Severity assessment is an essential element in the management of ICH.³⁴ Broadly speaking, NIHSS and hematoma size are extensively regarded as two key indicators of ICH severity.³⁵ In univariate analysis, the two indicators were closely related to serum Prx6 levels in this cohort of elderly patients after acute ICH. Moreover, this relationship remained unchanged in multiple linear regression analysis. From an alternative perspective, this finding implies that hemorrhagic brain injury may contribute to alterations of serum Prx6 levels in ICH patients Also, serum Prx6 could serve as a potential biomarker for evaluating severity of ICH, therefore aiding in risk stratification during clinical management of ICH.

Prognosis prediction is generally recognized as an extremely vital step in the treatment of ICH.³⁶ Among many scales, such as the Glasgow outcome scale, extended Glasgow outcome scale and activities of daily living scale, mRS had been widely adopted to evaluate neurological outcome.^37,38 Generally, patients are followed up for at least three months and commonly six months following stroke.³⁹ NIHSS scores and hematoma size are the two most common prognostic determinants of acute ICH.^40,41 In our cohort, serum Prx6 levels and three other adverse prognostic indicators, that is NIHSS scores, END and hematoma volume, were integrated to develop predictive model. The model exhibited robust clinical efficacy, validity, and stability via multiple statistical approaches including calibration curve, decision curve, and ROC curve analyses. Furthermore, serum Prx6 emerged as an independent predictor not only for poor prognosis but also for END, and mediation analysis revealed that END could partially mediate the association between serum Prx6 and poor prognosis. Specifically, significantly elevated serum Prx6 levels may be linked to pathophysiological mechanisms underlying END. However, owing to partial mediating role of END, serum Prx6 may serve as a key contributing factor to poor prognosis in patients with ICH. Collectively, additive value of serum Prx6 in combined prognosis prediction model, coupled with its association with poor prognosis mediated by END, further supports the notion that serum Prx6 is likely to become a promising prognostic indicator of poor outcomes in the elderly with ICH.

Once Prx6 is released from intracellular compartments, it would exert detrimental effects.⁴² Nevertheless, intracellular Prx6 functions as an antioxidant enzyme and is neuroprotective;^17–20 so, without Prx6 to clean up debris after neuronal cell death and cortical necrosis, mass effect from cerebral edema can lead to much bigger issues and worse long-term neurological outcomes. Likewise, current finding that serum Prx6 levels were elevated and correlated with worse outcomes of patients does not necessarily mean they are likely to be a driving factor for poor outcomes. Overall, Prx6 should be studied further in the clinical setting around the world, as drug therapies could be designed to target this protein and improve outcomes among elderly patients with ICH who often have poor outcomes due to multiple pre-existing comorbidities anyway.

This study was obligated to address strengths. Compared with our published work investigating the prognostic predictive value of resolvin D2 and hypoxia-inducible factor 1 alpha in ICH,^43,44 the current study has the following advantages. First, the present study is likely to be an inaugural series of in-depth investigations about elderly ICH patients, which is specifically designed to accurately measure and analyze the levels of serum Prx6 subsequent to acute ICH in human subjects. Second, compared to a single-center design, a multicenter study incorporates patients from diverse geographical regions and healthcare settings, therefore reflecting real-world heterogeneity more accurately, and subsequently ensuring conclusions’ generalizability to the broader population of elderly ICH patients. By utilizing data from this multicenter sample-size study, we successfully unearthed several highly fascinating results, which may serve to firmly validate that serum Prx6 levels have the potential to act as a crucial prognostic biomarker of ICH. Third, to minimize recall bias and measurement heterogeneity, this study adopted a prospective design and incorporated both cross-sectional and longitudinal components into a single study. By combining static associations with dynamic processes, it can not only reflect the potential value of real-time monitoring of pathophysiological processes during the acute phase but also assess whether their changing patterns (such as peak levels and decline slopes) have stronger prognostic significance than a single baseline value. The current study integrated the temporal trends of serum Prx6 levels post-ICH, thereby generating more information. Fourth, the current study incorporated END as a secondary outcome variable and a multivariate model was meticulously constructed with the aim of comprehensively examining the prognosis associations of serum Prx6 levels, and this association may be partially mediated by END, therefore leading to the notion that this novel mediation analysis may provide unexpected and valuable insights.

There are several limitations in this study. Serum Prx6 levels have been currently identified as a promising indicator for predicting adverse outcomes after ICH. However, this conclusion warrants verification through a larger-scale cohort survey. Also, a total of 103 patients were willing to accept blood drawings at multiple time points in order to determine the prognostic ability of serum Prx6 levels. It is possible that the designated time intervals are not completely adequate for this clinical analysis; and therefore, increasing the time points will be an optimal choice for future research. Alternatively, although known confounding factors were controlled through statistical methods and mediating effect of END was identified in this multicenter, observational study, the influence of unmeasured factors (such as disease severity perception, health literacy, and family support) on the results cannot be completely excluded. Therefore, the discovered association requires further experimentation to verify the causality. Time parameters have been added to multivariate model for adjusting, ensuring reliability and scientificity of results of regression analysis. Nevertheless, prompt variation of serum Prx6 levels may exist within 24 hours following ICH, and therefore this variability may lead to discrepancy in prognostic predictive ability of serum Prx6 levels. It may be a valuable research direction to meticulously investigate such nuance differences. Limitations in statistical power must be considered to interpret noninteraction of age with serum Prx6 levels in prognostic associations. Our sample size, particularly within ≥ 80-year-old subgroup (only 45 such patients in this study), may have been insufficient to detect a true, albeit modest, effect modification. Thus, the absence of a significant interaction may reflect limited statistical power rather than definitive biological homogeneity. So, it may be optimal to enhance size sample for discovering such a phenomenon.

Conclusions

Serum Prx6 levels in elderly patients with ICH are significantly elevated upon admission and are closely associated with severity, END, and adverse outcomes six months after acute ICH. Moreover, serum Prx6 levels display efficacious prognostic predictive ability, and END may partially elucidate the link between serum Prx6 levels and poor prognosis in elderly patients with ICH. These preliminary findings suggest that serum Prx6 may emerge as a potential prognostic biomarker for ICH in the elderly population. However, validation in independent cohorts, standardization of measurement timing, and comparison with existing ICH prognostic tools need to be done before clinical implementation. Moreover, it is meaningful to compare elderly versus non-elderly populations to establish age-specific utility of serum Prx6 in ICH.