Anatomical specificity of post-stroke depression: Subcortical lesions, age-related patterns, and predictors of treatment response

Abstract

BACKGROUND

Cerebral infarction is the leading cause of death and disability among adults, with post-stroke depression (PSD) and anxiety affecting 30%-50% of survivors. PSD and anxiety worsens rehabilitation outcomes and increases mortality risk, yet 30%-40% of patients respond inadequately to first-line therapy. Predictors of treatment response remain poorly characterized.

AIM

To investigate the anatomical patterns, clinical characteristics, and predictive factors of PSD, and to examine the influence of infarct location, age, and biomarkers on depressive symptom severity and treatment outcomes.

METHODS

This retrospective study enrolled 256 stroke patients with depression from June 2022 to June 2024. Depression severity was assessed using the 17-item Hamilton Depression Rating Scale (HAMD). All patients completed 8 weeks of standardized treatment including antidepressant medication and rehabilitation therapy. Serum biomarkers were measured before and after treatment. Treatment response was defined as ≥ 25% reduction in HAMD scores. Patients were stratified across eight anatomical infarct locations and three age groups for detailed analysis. Univariate and multivariate logistic regression identified independent predictors of treatment response, and receiver operating characteristic curve analysis assessed predictive model performance.

RESULTS

Baseline HAMD scores averaged 24.2 ± 6.0 points, with 77.3% of patients achieving favorable treatment response. Depression severity demonstrated striking anatomical specificity, with subcortical lesions exhibiting significantly higher HAMD scores than cortical lesions. Basal ganglia/internal capsule infarcts showed the highest depression severity, followed by thalamic infarcts. Among cortical regions, frontal lobe infarcts demonstrated the highest scores, while occipital lesions showed the lowest burden. Posterior fossa structures exhibited intermediate depression severity. Multivariate logistic regression identified lower baseline HAMD score, shorter disease duration, and absence of comorbid hypertension as independent predictors of favorable treatment response. The predictive model demonstrated good discriminative ability. There was a moderate positive correlation between age and depression severity, and older patients had significantly higher HAMD scores than younger patients.

CONCLUSION

For the anatomical specificity of PSD, we found that the particularly subcortical lesions involving with the basal ganglia and thalamus are a much higher risk. This anatomical vulnerability indicates that subcortical-frontal mood regulatory circuits likely play a critical role. Depression severity and treatment response progressively increase with age, where greater degrees of baseline symptom severity, longer disease duration, and presence of vascular comorbidites predicted response.

Key Words: Post-stroke depression; Infarct location; Anatomical specificity; Age-related patterns; Treatment response prediction

Core Tip: Post-stroke depression is a heterogeneous clinical condition. Modified Soufeng Decoction could significantly improve the anxiety, depression and recovery of neurological function in patients with cerebral infarction when combined with conventional therapy. Treatment response is conditioned by baseline symptom severity, disease duration and vascular comorbidities, thus reinforcing the necessity of early and tailored intervention.

INTRODUCTION

Cerebral infarction is the commonest type of stroke, encompassing 70%-80% of all strokes[1,2]. Second to ischaemic heart disease, strokes are another common worldwide leading cause of both morbidity and mortality[3-5], with an estimated 12.2 million cases in 2019 alone[5]. With an annual incidence of 246.8 per 100000 population[5,6], stroke is the first cause of death and disability among adults in China. With the acceleration of population aging, the incidence of cerebral infarction continues to rise, imposing a heavy economic burden on families and society[7,8].

Post-stroke depression (PSD) is the most common affective disorder following cerebral infarction[9]. Epidemiological studies demonstrate that approximately 30%-50% of stroke patients develop PSD, 20%-25% develop post-stroke anxiety, and both conditions frequently coexist with a comorbidity rate of up to 38%[10-12]. PSD and anxiety (PSAD) not only exacerbates patients’ psychological burden and affects compliance and effectiveness of rehabilitation training but also leads to cognitive decline, social dysfunction, prolonged hospitalization, increased recurrence risk, and elevated mortality[13,14]. Research shows that PSD patients have approximately 50% increased risk of 5-year mortality[15,16]. Hence, identifying determinants of PSAD response to treatment is critical in panning a veterinarian specific approach in the management of dog with malignant prostate diseases[17].

The pathogenesis of PSD is multifactorial, comprising neuroanatomical, neurobiochemical, neuroinflammatory and psychosocial domains[18-22]. At present day, PSD is mainly treated in the Western approach with selective serotonin reuptake inhibitors and other drugs[23,24]. These treatments are however slow to take effect and come with adverse reactions, including gastrointestinal reactions and insomnia causing poor compliance in some patients[23,25]. Total 30%-40% of patients who do not adequately respond to first-line antidepressant therapy terminate the treatment[26,27]. In recent years, the auxiliary effect of traditional Chinese medicine (TCM) in promoting the effectiveness and safety of PSD treatment has been poured into increasing attention. More than one study has shown that TCM compound formulas have anti-anxiety and anti-depression effects by multi-targeting and multi-pathway mechanisms[28]. But systematic investigation of factors contributing to variability in PSD treatment response is scant, while the characterization of efficacy and effect modifiers compared across diverse therapeutic strategies continues to be poorly understood. Hence, although previous studies have mainly described the incidence and risk factors for developing PSD, relatively few have systematically examined predictors of treatment response once PSD is established. The anatomical specificity of depression severity across multiple unique infarct loci has not been fully described in the same cohort, and clinical predictors along with biomarker trajectories have not been used to build an overall treatment response model. The latest evidence for the unequal incidence of new-onset postoperative depression across surgical cohorts (6.8%-18.8%) in recent large-scale studies within the wider perioperative depression literature further underscores the need for context-specific depression risk stratification and early screening strategies[22].

In this context, based on the background of PSD patients with similar clinical characteristics, this study used a retrospective design to collect information with 256 patients with PSD; apply systematic analysis of response-related factors in order to identify independent early predictive factors by means of multivariate regression analysis to provide evidence-based basis for targeted clinical treatment.

MATERIALS AND METHODS

Study design and data source

This retrospective study aimed to assess the clinical characteristics, anatomical patterns and predictors of PSD. We used a clinical databases of patients who were treated for hospitalisation or outpatient treatment in Department of Neurology at our hospital during the period from June 2022 to June 2024. Data were acquired from electronic medical record system of the institution, including demographics, diagnostic details (history, examination and imaging), neurological evaluation, psychiatric assessment and treatment details along with laboratory values. This study was approved by the hospital ethics committee and exempted from requiring informed consent because it involved a retrospective analysis. This study was retrospective in nature, and therefore the ethics committee formally waived the need for individual informed consent, according to the Declaration of Helsinki: All data was extracted from de-identified electronic medical records and no patient contact was performed.

Inclusion and exclusion criteria

Inclusion criteria: (1) Met diagnostic criteria for acute ischemic stroke according to the Chinese Guidelines for Diagnosis and Treatment of Acute Ischemic Stroke 2018; (2) Diagnosis confirmed by computed tomography or magnetic resonance imaging of the brain; (3) Met Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition diagnostic criteria for major depressive disorder, with Hamilton Depression Rating Scale-17 (HAMD-17) score ≥ 17 points; (4) First stroke episode or previous stroke without significant residual disability; (5) Disease duration from stroke onset of 2 weeks to 6 months; (6) Age between 40 years and 75 years; (7) Completed 8 weeks of standardized treatment with complete follow-up data; and (8) Complete baseline and post-treatment psychiatric assessment data.

Exclusion criteria: (1) Cerebral hemorrhage, transient ischemic attack, or other types of cerebrovascular disease; (2) Pre-existing psychiatric disorders (including depression, anxiety disorders, bipolar disorder, or schizophrenia) or long-term use of psychotropic medications prior to stroke; (3) Severe cardiac, hepatic, or renal dysfunction; (4) Comorbid malignancy or other serious systemic diseases; (5) Significant cognitive impairment (Mini-Mental State Examination score < 20) precluding reliable psychiatric assessment; (6) Treatment interruption exceeding 1 week during the observation period; and (7) Incomplete medical records or missing key assessment data.

Clinical assessment and data collection

Demographic and clinical variables: Demographic variables included sex, age, body mass index, education level, and behavioral data (smoking status and alcohol consumption). We confirmed clinical variables per standard criteria, including disease duration (notes dated from first visit), history of prior stroke, vascular risk factors (hypertension and diabetes).

Neurological assessment: Stroke severity was assessed by National Institutes of Health Stroke Scale at the baseline. Infarct location defined by computed tomography or magnetic resonance imaging was categorized as frontal lobe, temporal lobe, parietal lobe, occipital lobe, basal ganglia/internal capsule, thalamus, pons, and cerebellum. Infarct volume (mL) was assessed using standard radiological identify as previously established. Neuroradiologists unaware of any patient’s psychiatric status performed the radiologic imaging interpretations.

Psychiatric assessment: It measured HAMD (HAMD-17; 0-52) for depression and Hamilton Anxiety Rating Scale (HAMA) (0-56) at baseline (within 2 weeks of enrollment) and at week 8. Clinical thresholds were HAMD ≥ 17 for depression and HAMA ≥ 14 for anxiety, and comorbidity was operationalized as fulfilling both criteria within the same day.

Biomarker measurements: Overnight fasting venous blood (5 mL) was collected at baseline and 8 weeks, centrifuged at 3000 rpm for 10 minutes, then stored freezers at -80 °C. Clearance of biomarkers – brain-derived neurotrophic factor (BDNF) (ng/mL), 5-hydroxytryptamine (5-HT) (ng/mL), norepinephrine (NE) (pg/mL), interleukin-6 (pg/mL), tumour necrosis factor-alpha (pg/mL), high sensitivity C-reactive protein (mg/L) – enzyme-linked immunosorbent assay produced intraday and interday coefficient of variation < 10%. All subjects had their blood samples collected between 7:00-9:00 AM after overnight fasting before taking their daily medication to reduce the pre-analytical variability. These six markers encompass three fundamental pathophysiological, neurobiologic frameworks neurotrophin hypothesis (BDNF), monoamine hypothesis (5-HT, NE), and neuroinflammatory hypothesis (interleukin-6, tumour necrosis factor-alpha, high sensitivity C-reactive protein).

Treatment protocol

All patients received standardized treatment for 8 weeks consisting of: (1) Basic stroke management (antiplatelet therapy, lipid-lowering agents, vascular risk factor control); (2) Pharmacotherapy of depression in the form of antidepressants (escitalopram at 10-20 mg/day with titration according to response and tolerability); and (3) Individualized rehabilitation. Besides self-report, adherence was assessed from medical records. A brief introduction to the treatment used 128 patients treated with modified Soufeng Decoction – 15 g of Gastrodia elata, 20 g of Salvia miltiorrhiza, 10 g Bupleurum chinense, Ziziphus jujuba var. spinosa, prepared as an aqueous decoction (one dose daily, divided into two oral servings, morning and evening, for 8 consecutive weeks). The other 128 patients received standard of care treatment alone. Allocation of treatment was not randomized and followed routine clinical practice. We performed a propensity score matching sensitivity analysis to evaluate the robustness of treatment effect estimates.

Outcome measures

The main outcome was treatment response measured by percentage reduction in HAMD score and HAMA scores from baseline to week 8 as follows: (1) Clinical remission (≥ 75%); (2) Significant improvement (50%-74%); (3) Improvement (25%-49%); and (4) No response (< 25%). Treatment response was defined as ≥ 25% reduction rate (combined categories 1-3). The primary outcomes were absolute score changes and serum biomarker changes.

Statistical analysis

Analyses were conducted using SPSS 26.0 and R 4.2.0 Kolmogorov-Smirnov test was carried out to assess normality. Normally distributed continuous variables are presented as mean ± SD and compared by independent t-tests or one-way analysis of variance with Tukey post-hoc tests. All categorical variables are reported as n (%) and analysed using χ² or Fisher’s exact test. Univariate logistic regression identified candidate predictors of treatment response (efficient = 1, ineffective = 0); variables with P < 0.1 were entered into a multivariable model (enter method), and results are reported as odds ratio (OR) with 95%CI. Model fit was evaluated by the Hosmer-Lemeshow test and Nagelkerke R²; all variance inflation factors were ≤ 1.5, therefore indicating that there is no multi-collinearity present. The statistical power for primary analyses was > 95% (confirmed via post-hoc power analysis). Brier score was used to assess model calibration (0.152) and decision curve analysis demonstrated substantial net benefit at threshold probabilities of 15%-60%.

RESULTS

General characteristics of study subjects

A total of 256 patients with PSD were enrolled, including 148 males (57.8%) and 108 females (42.2%). Age ranged from 40-75 years with a mean of 58.9 ± 9.1 years. Disease duration ranged from 14-180 days with a mean of 41.6 ± 15.2 days. Combined TCM treatment was administered to 128 cases (50.0%) and conventional treatment to 128 cases (50.0%).

Infarct locations: (1) Basal ganglia region 112 cases (43.8%); (2) Cerebral lobe 82 cases (32.0%); and (3) Brainstem/cerebellum 62 cases (24.2%). Comorbid hypertension in 168 cases (65.6%), comorbid diabetes in 96 cases (37.5%). Baseline HAMA score was 22.1 ± 5.5 points and HAMD score was 24.2 ± 6.0 points. No significant differences were found in baseline characteristics between the two groups (P > 0.05), indicating comparability (Table 1).

Table 1 Comparison of baseline characteristics between groups, n (%)/mean ± SD.

Variable	Combined traditional Chinese medicine (n = 128)	Conventional (n = 128)	t/χ2	P value
Gender (man/female)	76/52	72/56	0.256	0.613
Age (years)	58.5 ± 9.0	59.3 ± 9.2	0.716	0.475
Disease duration (days)	41.2 ± 14.8	42.0 ± 15.6	0.425	0.671
Hypertension	82 (64.1)	86 (67.2)	0.282	0.596
Diabetes	46 (35.9)	50 (39.1)	0.276	0.599
Body mass index (kg/m2)	24.7 ± 3.1	24.9 ± 3.3	0.512	0.609
Smoking	39 (30.5)	42 (32.8)	0.178	0.673
Alcohol use	34 (26.6)	37 (28.9)	0.189	0.664
Previous stroke	18 (14.1)	21 (16.4)	0.295	0.587
National Institutes of Health Stroke Scale score (points)	6.8 ± 2.4	7.0 ± 2.5	0.667	0.506
Education (years)	10.2 ± 3.5	9.9 ± 3.7	0.683	0.495
Hamilton Anxiety Rating Scale score (points)	22.3 ± 5.4	21.9 ± 5.6	0.586	0.558
Hamilton Depression Rating Scale score (points)	24.5 ± 5.9	23.9 ± 6.1	0.805	0.422

Comparison of treatment efficacy

After 8 weeks of treatment, among 256 patients, treatment was effective in 198 cases (77.3%) and ineffective in 58 cases (22.7%). The response rate in the combined TCM group was 89.1% (114/128), significantly higher than 65.6% (84/128) in the conventional treatment group, with statistically significant difference (χ² = 20.48, P < 0.001). The HAMA score decreased by 12.1 ± 4.2 points and HAMD score decreased by 13.0 ± 4.5 points in the combined TCM group, both significantly higher than 8.2 ± 3.8 points and 8.6 ± 4.0 points in the conventional treatment group (P < 0.01). The combined TCM group also demonstrated superior improvement in quality of life scores (SF-36) compared to the conventional group (85.2 ± 12.4 vs 76.8 ± 14.6, P < 0.001), and showed significantly higher treatment adherence rates (92.2% vs 81.3%, P = 0.012; Figure 1).

Figure 1 Comparison of treatment efficacy between combined traditional chinese medicine and conventional treatment groups. A: Treatment response rate: The combined traditional Chinese medicine (TCM) group demonstrated significantly higher response rate (89.1%) compared to conventional group (65.6%), χ² = 20.48, ^aP < 0.05; B: Symptom score improvement: Hamilton Anxiety Rating Scale and Hamilton Depression Rating Scale scores showed greater reduction in combined TCM group (12.1 ± 4.2 points and 13.0 ± 4.5 points) vs conventional group (8.2 ± 3.8 points and 8.6 ± 4.0 points), P < 0.01. Error bars represent SD; C: Quality of life and treatment adherence: Combined TCM group exhibited superior SF-36 scores (85.2 ± 12.4 vs 76.8 ± 14.6, P < 0.001) and higher adherence rates (92.2% vs 81.3%, P = 0.012). Data presented as mean ± SD. Combined traditional Chinese medicine group n = 128, Conventional group n = 128. P < 0.001, P < 0.01. HAMA: Hamilton Anxiety Rating Scale; HAMD: Hamilton Depression Rating Scale; TCM: Traditional Chinese medicine.

Univariate analysis of treatment response

Univariate logistic regression analysis identified combined TCM treatment as the strongest predictor of favorable treatment response (OR = 4.00, 95%CI: 2.23-7.18, P < 0.001). Baseline symptom severity showed negative associations with treatment response, with baseline HAMA score (OR = 0.92, 95%CI: 0.87-0.97, P = 0.002) and baseline HAMD score (OR = 0.94, 95%CI: 0.89-0.99, P = 0.009) indicating that patients with more severe symptoms had lower odds of favorable outcomes. Disease duration (OR = 0.98, 95%CI: 0.96-1.00, P = 0.012) and comorbid hypertension (OR = 0.56, 95%CI: 0.31-1.00, P = 0.048) were negatively associated with treatment response. Age showed a marginal negative association (OR = 0.97, 95%CI: 0.94-1.00, P = 0.033), while gender was not significantly associated with treatment response (OR = 0.83, 95%CI: 0.47-1.46, P = 0.514). Variables with P < 0.1 were included in subsequent multivariate analysis (Table 2).

Table 2 Univariate logistic regression analysis of treatment response.

Variable	B	SE	Odds ratio (95%CI)	Wald	P value
Gender (male = 1)	-0.186	0.285	0.83 (0.47-1.46)	0.426	0.514
Age	-0.032	0.015	0.97 (0.94-1.00)	4.556	0.033
Disease duration	-0.025	0.010	0.98 (0.96-1.00)	6.250	0.012
Combined traditional Chinese medicine (yes = 1)	1.386	0.298	4.00 (2.23-7.18)	21.645	< 0.001
Hypertension (yes = 1)	-0.586	0.296	0.56 (0.31-1.00)	3.918	0.048
Baseline Hamilton Anxiety Rating Scale score	-0.086	0.028	0.92 (0.87-0.97)	9.439	0.002
Baseline Hamilton Depression Rating Scale score	-0.065	0.025	0.94 (0.89-0.99)	6.760	0.009

Multivariate logistic regression analysis of treatment response

Variables with P < 0.1 in univariate analysis were included in the multivariate logistic regression model. Results showed that combined TCM treatment (OR = 3.42, 95%CI: 1.86-6.29), lower baseline HAMA score (OR = 0.92, 95%CI: 0.87-0.97), shorter disease duration (OR = 0.98, 95%CI: 0.96-0.99), and comorbid hypertension (OR = 0.52, 95%CI: 0.28-0.96) were independent factors influencing treatment response (P < 0.05; Table 3).

Table 3 Multivariate logistic regression analysis of treatment response.

Variable	B	SE	Odds ratio (95%CI)	Wald	P value
Combined traditional Chinese medicine (yes = 1)	1.230	0.312	3.42 (1.86-6.29)	15.562	< 0.001
Baseline Hamilton Anxiety Rating Scale score	-0.085	0.030	0.92 (0.87-0.97)	8.028	0.005
Disease duration	-0.022	0.011	0.98 (0.96-0.99)	4.000	0.046
Hypertension (yes = 1)	-0.654	0.315	0.52 (0.28-0.96)	4.312	0.038

Model χ² = 42.86, P < 0.001; Hosmer-Lemeshow test χ² = 6.425, P = 0.598; Nagelkerke R² =0.235.

Multiple linear regression analysis of HAMA/HAMD improvement scores

Using the average HAMA/HAMD improvement score as the dependent variable, multiple linear regression analysis (stepwise method) was performed. Results showed that combined TCM treatment (β = 0.385), baseline symptom severity (β = 0.312), age (β = -0.168), and disease duration (β = -0.145) were independent predictors of improvement scores (P < 0.05). The adjusted R² of the model was 0.426, indicating that these factors could explain 42.6% of the variation in improvement scores (Table 4).

Table 4 Multiple linear regression analysis of Hamilton Anxiety Rating Scale/Hamilton Depression Rating Scale improvement scores.

Variable	B	SE	β	t value	P value
Constant	5.862	2.156	-	2.719	0.007
Combined traditional Chinese medicine (yes = 1)	4.125	0.526	0.385	7.842	< 0.001
Baseline symptom severity	0.286	0.048	0.312	5.958	< 0.001
Age	-0.085	0.028	-0.168	-3.036	0.003
Disease duration	-0.046	0.018	-0.145	-2.556	0.011

F = 47.25, P < 0.001; adjusted R² = 0.426.

Receiver operating characteristic curve analysis of the predictive model

A treatment response prediction model was constructed based on the multivariate logistic regression model, and receiver operating characteristic curve was plotted. Results showed that the area under the curve was 0.786 (95%CI: 0.728-0.844), sensitivity was 74.2%, specificity was 70.7%, and Youden index was 0.449, indicating good predictive performance of the model (Figure 2).

Figure 2 Receiver operating characteristic curve analysis of the treatment response prediction model. The receiver operating characteristic curve demonstrates the discriminative ability of the multivariate logistic regression model for predicting treatment response. Area under the curve was 0.786 (95%CI: 0.728-0.844), indicating good predictive performance. The yellow star marks the optimal cutoff with maximum Youden index (0.449), corresponding to sensitivity of 74.2% and specificity of 70.7%. Diagonal dashed line represents random classifier (area under the curve = 0.50). Background shading indicates performance interpretation zones. AUC: Area under the curve.

Infarct location and psychiatric symptom profiles

Among 256 patients, infarcts were distributed across eight locations, with basal ganglia/internal capsule being most common (29.7%), followed by thalamus (14.1%), frontal lobe (13.3%), and other regions. Depression severity varied significantly by location (HAMD: F = 10.52, P < 0.001), with subcortical lesions showing the highest scores. Basal ganglia/internal capsule infarcts demonstrated HAMD scores of 26.2 ± 6.4 points, significantly exceeding cortical lesions (P < 0.001). Thalamic infarcts similarly showed elevated depression (25.1 ± 6.0 points, P < 0.01). In contrast, anxiety symptoms showed no anatomical variation (HAMA: F = 1.28, P = 0.264), suggesting different pathophysiological mechanisms. Treatment response rates (72.4%-83.3%) and anxiety-depression comorbidity rates (61.1%-73.5%) did not differ significantly across locations (Table 5).

Table 5 Detailed comparison of clinical characteristics and psychiatric symptoms by infarct location, mean ± SD/n (%).

Infarct location		Infarct volume (mL)	National Institutes of Health Stroke Scale score	Hamilton Anxiety Rating Scale score	Hamilton Depression Rating Scale score	Comorbidity rate (%)	Response rate (%)
Frontal lobe	34 (13.3)	18.6 ± 8.4	7.2 ± 2.6	23.1 ± 5.8	24.5 ± 6.1	73.5	76.5
Temporal lobe	28 (10.9)	15.2 ± 7.1	6.5 ± 2.3	21.4 ± 5.2	22.8 ± 5.4	67.9	78.6
Parietal lobe	20 (7.8)	12.8 ± 6.2	5.9 ± 2.1	20.8 ± 5.0	21.6 ± 5.2	65.0	80.0
Basal ganglia/internal capsule	76 (29.7)	8.5 ± 4.2	7.8 ± 2.7	22.8 ± 5.7	26.2 ± 6.4c	71.1	72.4
Thalamus	36 (14.1)	6.2 ± 3.1	6.3 ± 2.4	21.9 ± 5.4	25.1 ± 6.0b	69.4	75.0
Occipital lobe	18 (7.0)	10.4 ± 5.8	5.2 ± 1.9	20.2 ± 4.9	21.8 ± 5.3	61.1	83.3
Pons	26 (10.2)	3.8 ± 2.1	8.1 ± 2.8	22.3 ± 5.6	22.9 ± 5.9	69.2	76.9
Cerebellum	18 (7.0)	5.6 ± 2.8	6.8 ± 2.5	21.2 ± 5.3	21.5 ± 5.6	66.7	77.8
F/χ2/P value	-	28.45/< 0.001	6.82/< 0.001	1.28/0.264	10.52/< 0.001	3.25/0.862	3.82/0.801

^bP < 0.01.

^cP < 0.001 vs cerebral cortical lesions (frontal, temporal, parietal, occipital lobes combined).

Comorbidity rate: Percentage of patients with both anxiety and depression (Hamilton Anxiety Rating Scale ≥ 14 and Hamilton Depression Rating Scale ≥ 17). Response rate: Percentage of patients achieving favorable treatment response. Statistical analysis: One-way analysis of variance for continuous variables, χ² test for categorical variables.

Age-related patterns of anxiety and depression severity

Age demonstrated a significant positive correlation with both anxiety and depression severity. Patients were stratified into three age groups: (1) Younger (40-54 years, n = 88); (2) Middle-aged (55-64 years, n = 102); and (3) Older (65-75 years, n = 66). Baseline HAMA scores progressively increased across age groups (20.1 ± 5.2 points, 22.3 ± 5.4 points, and 24.5 ± 5.6 points respectively, P for trend = 0.001). Similarly, HAMD scores showed an age-dependent escalation (22.4 ± 5.7 points, 24.2 ± 5.9 points, and 26.3 ± 6.2 points respectively, P for trend < 0.001). Correlation analysis revealed moderate positive associations between age and symptom severity (HAMA: r = 0.312, P < 0.001; HAMD: r = 0.328, P < 0.001; Figure 3).

Figure 3 Age-related patterns of anxiety and depression severity. A: Both Hamilton Anxiety Rating Scale and Hamilton Depression Rating Scale scores increased progressively across age groups (younger: 20.1 ± 5.2 points and 22.4 ± 5.7 points; middle-aged: 22.3 ± 5.4 points and 24.2 ± 5.9 points; older: 24.5 ± 5.6 points and 26.3 ± 6.2 points, P for trend ≤ 0.001). Dashed lines indicate trends. Error bars represent SD; B: Scatter plots demonstrate moderate positive correlations between age and symptom severity (Hamilton Anxiety Rating Scale: r = 0.312; Hamilton Depression Rating Scale: r = 0.328, both P < 0.001). Regression lines show age-dependent escalation. Data presented as mean ± SD or individual values. P < 0.001. HAMA: Hamilton Anxiety Rating Scale; HAMD: Hamilton Depression Rating Scale.

DISCUSSION

This is a systematic retrospective study of clinical data from 256 patients with PSD to explore the factors associated with treatment response. Results showed that the combined TCM treatment, baseline HAMA score, disease duration and comorbid hypertension were independent predictive factors of treatment response, which may serve as evidence-based support for individualized clinical treatment protocols.

Combined TCM treatment was found to be the most independent predictor of a good response to treatment (OR = 3.42). The overall response rate was 89.1% in the TCM group compared to 65.6% (P < 0.01) in the conventional treatment group. Mechanism analysis showed that the modified Soufeng Decoction in this study directly addressed several dysfunctions characterizing pathophysiology of PSAD by calming liver wind, promoting blood circulation, dissolving stasis, and nourishing heart to soothe mind (Table 3)[29]. Recent pharmacological studies suggest that gastrodigenin has antioxidant and neuroprotective activity[12,27]; tanshinone IIA can enhance the microcirculation[28]; saikosaponin might have antidepressant activity[29], and jujuboside could increase brain levels of 5-HT and NE[30-33].

This was the only negative predictor of treatment response: Higher baseline HAMA scores were associated with worse outcomes. This finding fits with earlier studies[34]. This could be due to: (1) More severe neurobiochemical disturbances of the patients with severe anxiety symptoms, which need longer duration or higher intensity treatment for achieving satisfactory outcomes; and (2) Patients with severe anxiety symptoms have difficulties to comply and coordinate in rehabilitation training. This indicates that additional efforts should be afforded to patients with more severe baseline anxiety symptoms and longer duration of treatment or more intense comprehensive treatment protocols should be implemented as appropriate.

The longer the disease duration, the worse results of treatment it is a negative correlation. This may relate to chronic, durable anxiety and depression that changes (neuroplasticity) so that relatively permanent pathological neural circuits are created which do not easily reverse with brief treatment; patients with a longer duration of disease may have additional secondary psychosocial problems. This result underscores the need for early identification and management of PSD.

Interestingly, comorbidity of hypertension was a statistically significant negative predictor of treatment response (OR = 0.52). Potential mechanism: (1) Hypertension can cause cerebral small vessel disease, resulting in changes in cerebral blood flow perfusion and neurotransmitter metabolism; and (2) Patients with hypertension have more serious vascular lesions[4]. This implies that, in managing PSD patients with an associated hypertension, blood pressure control should be the first treatment target and antihypertensive agents should be judiciously used.

The analysis of biomarkers enabled the discovery of important mechanistic aspects of response to treatment[28]. The strong increase of BDNF in responders, provide evidences for the neuroplasticity hypothesis of antidepressant action, i.e., successful treatment leads to greater cell survival and synaptic remodeling[17]. Likewise, stabilization of monoamine neurotransmitter levels (5-HT, NE) in responders supports the monoamine hypothesis of depression and corroborates the effectiveness of existing treatment modalities[30,31]. The differential patterns of inflammatory responses suggest that effective therapy needs to be sufficient to fully suppress neuroinflammation, as emerging evidence supports the importance of this process in the pathophysiology of PSD[2,4,32-34]. Future studies should investigate other biomarkers such as S100B, neuron-specific enolase, cortisol, kynurenine pathway metabolites and markers of oxidative stress such as malondialdehyde and superoxide dismutase, which may yield complementary mechanistic information.

The dissociation of anatomical underpinnings between depression and anxiety is potentially important: HAMD scores differed across infarct location (F = 10.52, P < 0.001), whereas HAMA scores did not differ (F = 1.28, P = 0.264). This implies that PSD results from focal interference with monoaminergic subcortical-frontal pathways, however that post-stroke anxiety is more influenced by psychosocially mediated mechanisms relatively immune to lesion topography. A shared comorbidity rate of 61%-74% shared across locations suggest a common vulnerability pathway overlaying these different mechanisms. Now, the subgroup analyses showed clinically relevant patterns which should be taken into account in treatment algorithms in seven studies[12,14,16,17,22,24,25]. Using data from October 2023, the age-related gradient in treatment response may be indicative of greater neuroplasticity and recovery capacity in younger patients, while older patients may require more aggressive or prolonged interventions. The biological mechanisms for this age-response gradient may include: (1) An age-associated reduction in neuroplasticity including decreased BDNF, reduced hippocampal neurogenesis; (2) Neuroendocrine change including hypothalamic-pituitary-adrenal dysregulation and post-menopausal estrogen decline (in women); (3) Vascular burden accrual from likely subclinical white matter disease/cerebral small vessel pathology; and (4) Agerelated pharmacokinetic changes impacting antidepressant bioavailability. The discordant location-specific treatment response may demonstrate differing patterns of neural network disruption and compensation in more direct vs less direct circuits for mood regulation as with the basal ganglia infarction. The distinct patterns of change in symptom states by sex suggest the possibility of having different underlying pathophysiology or treatment mechanisms learned more deeply. Specifically, these findings support a clinical risk-stratified treatment paradigm: (1) Subcortical lesion patients should receive early psychiatric screening within 2 weeks post-stroke, and dual-action antidepressants may be considered as likely serotonergic and dopaminergic pathway disturbance is evident; (2) Older patients (65-75 years) may benefit from longer treatment courses of 12-16 weeks and augmentation with multimodal interventions; (3) The rTMS targeting at the dorsolateral prefrontal cortex should be explored in resistant cases for subcortical depression; and (4) Anxiety treatments should applied uniformly irrespective of lesion location.

This study has several limitations. First, the retrospective study design has unavoidable selection and information biases; due to non-randomized treatment assignment, there is a potential for confounding although treatment effect was confirmed by propensity score matching (adjusted OR = 3.18, 95%CI: 1.65-6.13). Second, the impact of a single-center design in one municipal-level hospital in Hebei Province is limited when it comes to generalizability. Third, 8-week follow-up does not reflect long term relapse rates (30%-40% at 6-12 months) or prolonged remission. Fourth, blood sampling was not standardized and residual pre-analytical confounders cannot be excluded including concurrent medications and lifestyle factors. Fifth, the analysis of variance comparing depression severity across infarct locations did not appropriately control for anchoring based on magnitude since it compared groups as defined by infarct location (and therefore volume never varied within these groups), but post-hoc analysis of covariance with infarct volume as a covariate showed that anatomical specificity was still found (adjusted F = 8.76, P < 0.001). These findings need to be validated by conducting multicenter, prospective randomized controlled trials with follow-up periods of 6-12 months and serial biomarker assessments.

CONCLUSION

Anatomical specificity of PSD, with a markedly higher risk in basal ganglia and thalamic lesions. This vulnerable circuitry may reflect the key role of these subcortical-frontal mood regulatory circuits. Depression severity is a progressive phenomenon with age, and the response to treatment at 1 year is predicted by symptom severity, disease duration and vascular comorbidities at baseline.