Sleep disturbance and inflammatory markers associated with post-stroke depression, anxiety, and cognitive impairment

Abstract

BACKGROUND

Neuropsychiatric complications, such as post-stroke depression (PSD), post-stroke anxiety (PSA) and post-stroke cognitive impairment (PSCI), are common after stroke and have important impacts on functional recovery and quality of life. Although sleep disturbances and inflammation have been implicated, their combined role remains unclear.

AIM

To assess how post-stroke sleep disturbance and acute inflammation relate to and predict depression, anxiety, and cognitive impairment.

METHODS

This retrospective study included 160 stroke patients treated at The Second People’s Hospital of Nantong (2022-2025). Acute-phase inflammatory markers were collected during hospitalization, and sleep quality was evaluated at 1-3 months applying the Pittsburgh Sleep Quality Index (PSQI). Outcomes (PSD, PSA, PSCI) were evaluated utilizing the 17-item Hamilton Depression Rating Scale, Hamilton Anxiety Rating Scale, and Montreal Cognitive Assessment. A principal component analysis was used to create a composite inflammatory indicator. Logistic regression and receiver operating characteristic curve analyses were conducted.

RESULTS

The incidences of PSD, PSA, and PSCI were 28.13%, 25.63%, and 40.63%, respectively. Sleep disturbance (PSQI > 5) significantly raised the risk for all outcomes [odds ratio (OR) range 2.22-2.34, all P < 0.05]. Both sleep disturbance and inflammatory burden were independent risk indicators (OR for inflammatory burden: 1.45-1.56, all P < 0.05). Area under the curves of the combined model (sleep + inflammation) for prediction of PSD, PSA, and PSCI were 0.664, 0.682, and 0.653, respectively, showing modestly better discriminative power compared with the single indicators.

CONCLUSION

Post-stroke sleep disturbance is independently associated with PSD, PSA, and PSCI as is acute inflammation. The sum of their assessment enhances risk stratification, helping in early intervention and personalized management.

Key Words: Post-stroke depression; Post-stroke anxiety; Post-stroke cognitive impairment; Sleep disturbance; Inflammatory markers; Retrospective study

Core Tip: This study systematically examined effects of sleep disturbance and inflammatory burden in the acute phase on post-stroke depression, anxiety and cognitive impairment. Sleep disturbance and inflammatory burden were both found to be independent risk factors for neuropsychiatric-cognitive adverse post-stroke outcomes. A combined assessment model further enhanced the discriminative performance and provided a scientific basis for early identification of high-risk patients and personalized intervention strategy.

INTRODUCTION

Stroke is a main cause of death and disability, and is the second most common reason of death around the world[1]. Significant improvements in the survival rates have been achieved through improvements in acute phase treatment. Neuropsychiatric-cognitive complications, however, have emerged as core issues affecting long-term prognosis, with post-stroke depression (PSD), post-stroke anxiety (PSA), and post-stroke cognitive impairment (PSCI) all figuring prominently[2-4]. Such complications exhibit a profound impact on quality of life and are strongly correlated with higher long-term risk of dementia, lower functional independence and mortality[5].

Post-stroke neuropsychiatric cognitive disorders are thought to be related not only to psychological stress, but to a combination of neurobiological abnormalities and disruption of behavioral rhythms. Two important pathological processes that are believed to be important, and correlated, are neuroinflammation and sleep-wake cycle regulation. When the blood-brain barrier (BBB) is compromised due to stroke, peripheral inflammatory mediators and immune cells can traverse the BBB and enter the central nervous system to initiate or exacerbate neuroinflammatory processes[6]. Sleep disturbances are present in 40%-60% of patients with stroke, with the most common symptoms being difficulty falling asleep, maintaining sleep continuity, and premature awakening[7]. Previous research suggests that disrupted sleep can have a negative impact on emotional control and cognitive performance by disrupting neuroreparative mechanisms, by reducing metabolic waste clearance from the brain and by altering the plasticity of synapses[8].

Significantly, inflammatory and sleep disturbance are not mutually exclusive. There is increasing evidence that sleep disruption can trigger inflammatory reactions both peripherally and centrally, and inflammatory mediators can interfere with sleep architecture and/or circadian stability, which can then further trigger sleep disruption, forming a vicious cycle[9,10]. Thus, inflammation and sleep disturbance could interact on emotional regulation and cognitive processing, which could, in turn, impact neuropsychiatric-cognitive outcomes following stroke.

These mechanisms have been described independently, and the current literature on post-stroke neuropsychiatric-cognitive disorders still tends to be limited to the study of single biological or behavioral aspects without a systematic approach, which combines inflammatory burden and sleep disturbance[11,12]. It is thus uncertain whether a combined evaluation of inflammation and sleep status can better identify patients at high risk for such complications and thus improve risk stratification and early intervention.

In order to fill this gap, the present study aims to assess PSD, PSA, PSCI, sleep disturbance and to measure acute phase inflammatory markers in acute stroke. It also investigates the interaction of the two dimensions, inflammation and sleep, in the prediction of neuropsychiatric-cognitive disorders following stroke and to offer a move towards clinical applicability for early identification of high-risk patients and design of subsequent intervention strategies.

MATERIALS AND METHODS

Study design and subjects

A total of 160 stroke patients were hospitalized in The Second People's Hospital of Nantong during October 2022 to October 2025, and included in this single-center retrospective study. All patients had an ischemic/hemorrhagic stroke, which was confirmed by a cranial computed tomography or magnetic resonance imaging. Patients were analyzed using a common follow-up window of 1-3 months after stroke, during which they had been evaluated for depression, anxiety, and cognitive function. This study obtained approval from the Institutional Review Board of The Second People’s Hospital of Nantong. The patient selection process, inclusion and exclusion criteria, and assessment time points are summarized in Figure 1.

Figure 1 Flowchart of patient selection and study procedures. PSD: Post-stroke depression; PSA: Post-stroke anxiety; PSCI: Post-stroke cognitive impairment; PSQI: Pittsburgh Sleep Quality Index; HAMD-17: 17-item Hamilton Depression Rating Scale; MoCA: Montreal Cognitive Assessment; HAMA: Hamilton Anxiety Rating Scale.

Inclusion and exclusion criteria

Inclusion criteria: (1) Age ≥ 18 years; (2) Completion of the 17-item Hamilton Depression Rating Scale (HAMD-17)[13], Hamilton Anxiety Rating Scale (HAMA)[14], and Montreal Cognitive Assessment (MoCA)[15] at 1-3 months post-stroke; (3) Measurement of inflammatory markers within 72 hours of stroke admission; and (4) Complete clinical and follow-up data for analysis.

Exclusion criteria: (1) Pre-stroke history of diagnosed depression, anxiety, or other major psychiatric issues; (2) Severe disturbance of consciousness, severe aphasia, or other conditions precluding scale assessment; (3) Acute infection, autoimmune disease, or malignancy; (4) Long-term use of corticosteroids or immunosuppressants; and (5) Missing key clinical or follow-up data.

Data collection

Sleep disturbance assessment: Post-stroke sleep disturbance was assessed at 1-3 months follow-up applying the Pittsburgh Sleep Quality Index (PSQI)[16], and indicated by a score > 5.

Acute-phase inflammatory markers: C-reactive protein (CRP), interleukin-6 (IL-6), tumor necrosis factor-α (TNF-α), and neutrophil-to-lymphocyte ratio (NLR) measured within 72 hours of admission, were obtained from medical records and laboratory reports. CRP was measured by immunoturbidimetry (Beckman Coulter, United States), while IL-6 and TNF-α were determined via ELISA (R&D Systems, United States). NLR was calculated from the complete blood count. All assays were carried out by the clinical laboratory of our hospital as per the manufacturers’ instructions. Reference ranges were as follows: 0-5 mg/L for CRP, 0-7 pg/mL for IL-6, and 0-8 pg/mL for TNF-α.

Covariates: Demographic and clinical covariates (such as age, sex, stroke type, stroke severity, assessed by the National Institutes of Health Stroke Scale)[17], and comorbidities (e.g., hypertension, diabetes and cardiovascular disease), were directly obtained from medical records.

To achieve uniformity in recording all the variables, data collection was done following standard procedures. Completeness of original medical records and laboratory data was checked. Post-stroke neuropsychiatric and cognitive evaluations were made by trained neurological researchers blinded to patients’ clinical information during analysis to minimize assessment bias.

Outcome assessment

Post-stroke neuropsychiatric and cognitive outcomes were assessed within a unified follow-up window of 1-3 months post-stroke, defined as follows: HAMD-17 contains 17 items (score range 0-52), with higher scores indicating severer depression. PSD was defined as a HAMD-17 score ≥ 7. HAMA includes 14 items (score range 0-56), with a score ≥ 7 indicating PSA. MoCA assesses eight cognitive domains (total score 30), and PSCI was defined as a MoCA score < 26.

Statistical analysis

Statistical analysis was carried out applying SPSS 26.0. After being tested for normality, continuous data with a normal distribution were expressed as mean ± SD and compared utilizing independent samples t-tests, while data without a normal distribution were presented as median (interquartile range), with comparisons through the Mann-Whitney U test. Categorical data were expressed as n (%) and compared via the χ² or Fisher’s exact test. Univariate logistic regression was employed for the preliminary screening of potential factors associated with PSD, PSA, and PSCI and those with P < 0.05 were entered into the multivariate logistic regression models. For multivariate analyses to avoid skewering results due to the influence of various scales, acute-phase inflammatory markers were Z score transformed to remove the scale effect. Next, principal component analysis (PCA) was conducted on these four variables after they were standardized in order to reduce the number of dimensions, and 4 factors were retained (accounting for 73.9% of the cumulative variance) with each having an eigenvalue > 1. A composite score was created that would reflect overall inflammatory burden, which was then added to the multivariable models. To minimize the risk of overfitting, the events per variable principle was followed for variable selection in multivariate logistic regression. The discrimination of the combined predictive model was determined by the receiver operating characteristic (ROC) curves and the area under the curve (AUC). All tests were two-sided, with P < 0.05 considered statistically significant.

RESULTS

Demographic and clinical characteristics of the subjects

Totally, 160 patients with stroke were included. Complete data on demographics, stroke type, stroke severity, comorbidities, and acute-phase inflammatory markers were obtained from medical records and laboratory reports (Table 1).

Table 1 Baseline demographic and clinical characteristics of patients with stroke, n (%).

Characteristics	Overall cohort (n = 160)
Age, years, mean ± SD	62.24 ± 10.33
Sex
Male	62 (38.75)
Female	98 (61.25)
Stroke type
Ischemic stroke	123 (76.88)
Hemorrhagic stroke	37 (23.13)
NIHSS score, median (IQR)	5 (4, 7)
Comorbidities
Hypertension	97 (60.63)
Diabetes mellitus	47 (29.38)
Cardiovascular disease	42 (26.25)
Acute-phase inflammatory markers, median (IQR)
CRP, mg/L	8.46 (6.66, 14.21)
IL-6, pg/mL	8.32 (6.81, 10.35)
TNF-α, pg/mL	9.51 (8.04, 11.22)
NLR	2.66 (2.25, 3.74)
PSQI score, median (IQR)	7 (4, 10)

NIHSS: National Institutes of Health Stroke Scale; IQR: Interquartile range; CRP: C-reactive protein; IL-6: Interleukin-6; TNF-α: Tumor necrosis factor-α; NLR: Neutrophil-to-lymphocyte ratio; PSQI: Pittsburgh Sleep Quality Index.

Incidents of PSD, PSA, and PSCI

Assessments using the HAMD-17, HAMA, and MoCA were conducted at 1-3 months post-stroke. The incidences of PSD, PSA, and PSCI were 28.13%, 25.63%, and 40.63%, respectively (Table 2).

Table 2 Distribution of psychiatric and cognitive outcomes after stroke.

Characteristics	Overall cohort (n = 160)
HAMD-17 score, median (IQR)	5 (3.14)
PSD (HAMD-17 ≥ 7), n (%)	45 (28.13)
HAMA score, median (IQR)	5 (3.11)
PSA (HAMA ≥ 7), n (%)	41 (25.63)
MoCA score, median (IQR)	26 (20.28)
PSCI (MoCA < 26), n (%)	65 (40.63)

HAMD-17: 17-item Hamilton Depression Rating Scale; PSD: Post-stroke depression; HAMA: Hamilton Anxiety Rating Scale; PSA: Post-stroke anxiety; MoCA: Montreal Cognitive Assessment; PSCI: Post-stroke cognitive impairment; IQR: Interquartile range.

Comparison of demographic and clinical characteristics by sleep status

Based on PSQI scores, patients were classified into sleep disturbance (PSQI score > 5, n = 96) and non-sleep disturbance (PSQI score ≤ 5, n = 64) groups. As shown in Table 3, the sleep disturbance group exhibited markedly higher incidences of PSD [33 (34.38%) vs 12 (18.75%), P = 0.031], PSA [30 (31.25%) vs 11 (17.19%), P = 0.046], and PSCI [46 (47.92%) vs 19 (29.69%), P = 0.021] relative to the non-sleep disturbance group.

Table 3 Comparison of demographic and clinical characteristics in patients with stroke who have sleep disturbances vs those who do not, n (%).

Characteristics	Sleep disturbance (PSQI > 5, n = 96)	Non-sleep disturbance (PSQI ≤ 5, n = 64)	t/χ2/Z	P value
Age, years, mean ± SD	61.00 ± 9.77	64.09 ± 10.93	-1.871	0.063
Sex			0.317	0.573
Male	35 (36.46)	27 (42.19)
Female	61 (63.54)	37 (57.81)
Stroke type			0.709	0.400
Ischemic stroke	76 (79.17)	47 (73.44)
Hemorrhagic stroke	20 (20.83)	17 (26.56)
NIHSS score, median (IQR)	5 (4, 7)	4 (3, 6)	-1.905	0.057
Comorbidities
Hypertension	60 (62.50)	37 (57.81)	0.353	0.552
Diabetes mellitus	27 (28.13)	20 (31.25)	0.181	0.671
Cardiovascular disease	25 (26.04)	17 (26.56)	0.005	0.942
Acute-phase inflammatory markers, median (IQR)
CRP, mg/L	8.44 (6.77, 13.51)	8.49 (6.57, 14.92)	-0.233	0.815
IL-6, pg/mL	8.67 (7.09, 10.23)	7.87 (6.26, 10.59)	-1.118	0.264
TNF-α, pg/mL	9.85 (8.51, 11.27)	9.19 (7.16, 11.19)	-1.027	0.304
NLR	2.67 (2.30, 3.94)	2.65 (2.15, 3.52)	-0.998	0.318
PSD (HAMD-17 ≥ 7)	33 (34.38)	12 (18.75)	4.638	0.031
PSA (HAMA ≥ 7)	30 (31.25)	11 (17.19)	3.984	0.046
PSCI (MoCA < 26)	46 (47.92)	19 (29.69)	5.290	0.021

NIHSS: National Institutes of Health Stroke Scale; IQR: Interquartile range; CRP: C-reactive protein; IL-6: Interleukin-6; TNF-α: Tumor necrosis factor-α; NLR: Neutrophil-to-lymphocyte ratio; PSD: Post-stroke depression; PSA: Post-stroke anxiety; PSCI: Post-stroke cognitive impairment; PSQI: Pittsburgh Sleep Quality Index; HAMD-17: 17-item Hamilton Depression Rating Scale; HAMA: Hamilton Anxiety Rating Scale; MoCA: Montreal Cognitive Assessment.

Univariate analysis

Univariate analysis indicated that higher levels of inflammatory markers and the presence of sleep disturbance were significantly associated with PSD, PSA, and PSCI (P < 0.05), whereas demographic characteristics and pre-existing comorbidities had no significant effect (P > 0.05, Table 4).

Table 4 Univariate analysis of factors associated with post-stroke depression, post-stroke anxiety, and post-stroke cognitive impairment.

Variables	PSD		PSA		PSCI
	t/χ2/Z	P value	t/χ2/Z	P value	t/χ2/Z	P value
Age	-0.965	0.336	0.284	0.777	0.320	0.750
Sex	0.855	0.355	0.492	0.483	1.109	0.292
Stroke type	0.442	0.506	0.050	0.824	3.690	0.055
NIHSS score	-0.703	0.482	-1.861	0.063	-1.600	0.110
CRP	-2.150	0.032	-2.253	0.024	-0.215	0.829
IL-6	-2.118	0.034	-2.017	0.044	-2.486	0.013
TNF-α	-1.133	0.257	-1.477	0.140	-3.676	0.000
NLR	-3.355	0.001	-2.322	0.020	-2.785	0.005
Hypertension	0.010	0.919	0.003	0.958	0.038	0.845
Diabetes mellitus	0.734	0.392	0.605	0.437	2.093	0.148
Cardiovascular disease	2.321	0.128	1.775	0.183	0.502	0.478
Sleep disturbance	4.638	0.031	3.984	0.046	5.290	0.021

PSD: Post-stroke depression; PSA: Post-stroke anxiety; PSCI: Post-stroke cognitive impairment; NIHSS: National Institutes of Health Stroke Scale; CRP: C-reactive protein; IL-6: Interleukin-6; TNF-α: Tumor necrosis factor-α; NLR: Neutrophil-to-lymphocyte ratio.

Multivariable logistic regression analysis

Multivariable logistic regression demonstrated that both sleep disturbance and overall inflammatory burden were independent predictors of PSD, PSA, and PSCI (P < 0.05; Table 5).

Table 5 Multivariable logistic regression analysis of factors associated with post-stroke outcomes.

Outcome	Variables	β	SE	χ2	P value	OR (95%CI)
PSD	Sleep disturbance	0.849	0.394	4.65	0.031	2.34 (1.08-5.06)
	Inflammatory burden	0.387	0.177	4.79	0.029	1.47 (1.04-2.08)
PSA	Sleep disturbance	0.825	0.409	4.08	0.043	2.28 (1.03-5.08)
	Inflammatory burden	0.443	0.181	6.01	0.014	1.56 (1.09-2.22)
PSCI	Sleep disturbance	0.799	0.349	5.20	0.023	2.22 (1.12-4.40)
	Inflammatory burden	0.374	0.167	5.01	0.025	1.45 (1.05-2.02)

PSD: Post-stroke depression; PSA: Post-stroke anxiety; PSCI: Post-stroke cognitive impairment; OR: Odds ratio; CI: Confidence interval.

Discriminative performance of the combined assessment model

ROC curve analysis showed that sleep disturbance alone had statistically significant predictive value for PSD and PSA (P < 0.05), whereas inflammatory burden alone was not a significant predictor (P > 0.05, Figure 2). The combined model showed significant increase in discriminative power for all three outcomes with AUC of 0.664 for PSD, 0.682 for PSA and 0.653 for PSCI.

Figure 2 Receiver operating characteristic curves for the combined assessment model predicting post-stroke outcomes. A: Receiver operating characteristic (ROC) curves for post-stroke depression; B: ROC curves for post-stroke anxiety; C: ROC curves for post-stroke cognitive impairment. AUC: Area under the curve.

DISCUSSION

This retrospective study systematically examined the effects of post-stroke sleep disturbance and inflammatory markers in the acute phase on PSD, PSA, and PSCI in a single-center study. Using PCA, the four markers of CRP, IL-6, TNF-α, and NLR were reduced to a composite feature that reflects a systemic inflammatory burden. The results revealed that sleep disturbance was independently associated with PSD, PSA and PSCI, and inflammatory burden was independently associated with these measures, suggesting potential use as risk stratifiers. In addition, the combined assessment model showed incremental improvement with respect to discriminative ability compared with single-indicator assessment models, suggesting that simultaneously assessing sleep and inflammatory status immediately after stroke could offer a more comprehensive approach to identifying patients at high risk of these disorders and may have clinical utility. The moderate AUC values seen are plausible, reflecting the polyfaceted nature of post-stroke neuropsychiatric-cognitive outcomes and are in keeping with the findings from past observational research.

We observed a positive and statistically significant correlation of sleep disturbances with poor neuropsychiatric-cognitive outcomes following stroke, in accordance with the results of prior studies[18]. In a multicenter prospective clinical study, it was found that patients with stroke with continuous poor sleep at 3 months had heightened depression and anxiety risks, suggesting that poor sleep quality following stroke is a risk factor for PSD and PSA[19]. In another study, roughly 58.7% of patients suffered from sleep disturbances to different extents within the first 1 month following acute ischemic stroke, with sleep quality being significantly associated with cognitive function, depression and anxiety scores[20]. The results indicate that sleep disruption affects emotional state, and may also impact on poor psychiatric outcomes via cognitive pathways. Mechanisms could involve disruption of the glymphatic clearance system[21], dysfunction in synaptic plasticity[22], and dysfunction of the hypothalamic-pituitary-adrenal (HPA) axis[23]. These disruptions may disrupt neural homeostasis, leading to a worsening of functional abnormalities in neural networks relevant to emotional regulation and cognitive functioning.

Inflammatory response is a key component of the pathogenesis of PSD, PSA and PSCI. Our discovery of a relationship between increased acute-phase inflammatory burden and several negative outcomes is consistent with previous research[24]. In acute ischemic stroke patients, the concentration of IL-6, TNF-α and CRP were all found to be significantly correlated with depression and anxiety and cognitive decline by Li et al[25]. Pinzi et al[26] also pointed out that inflammatory cytokines are important in emotional and cognitive disorders through the effect on neurotransmitter metabolism, the activation of the HPA axis and the disruption of neuroplasticity. In addition, a higher level of inflammatory markers is generally observed in PSCI and is associated with an inverse correlation with cognitive scores, suggesting inflammation plays a role in PSCI[27]. Recent reviews suggest that sleep disturbance can be not only a result of inflammation, but can also trigger inflammatory responses at the systemic, cellular and genetic level, creating a “sleep-inflammation dual hit” in addition to the inflammatory burden[28]. Such a mechanism may heighten the functional deficits of neural circuits underlying mood regulation, reward processing and emotional recognition, thus considerably increasing the risk of depression. Thus, sleep disturbance and inflammatory response might link to each other to initiate and maintain post-stroke neuropsychiatric-cognitive disorders.

In the clinical context, this study highlights the potential of systematic evaluation of sleep quality and inflammatory status in the early post-stroke period for a more accurate identification of subgroups at high risk of PSD, PSA, and PSCI. The incremental increase in the discriminative ability of the combined model may help clinical risk stratification and individualized follow-up, as it offers discriminative capability at multiple outcomes. Thus, treating sleep disturbance (e.g., through cognitive-behavioral therapy and sleep hygiene education) and taking a holistic approach toward sleep inflammatory modulation could lessen the burden of such complications and create better rehabilitative outcomes. These findings are consistent with those from past psychiatric research[29]. The added value of the combined assessment is in the possibility to inform multidimensional, early and accurate intervention strategies to optimize long-term mental health management of patients with stroke.

A number of limitations of this study exist. Firstly, the study is a single-center retrospective design, which could present selection biases, and so the findings may not be widely generalizable. Second, measurement of sleep quality was based mostly on the PSQI questionnaire and did not include any objective measure like polysomnography which might create information bias. Third, the inflammatory markers were only measured during the acute admission period, and hence dynamics of change and long-term inflammatory burden could not be assessed. Lastly, important potential confounding factors, such as medication use and psychosocial factors, were not fully controlled, and this may influence the validity of the multiple models. Future studies should be conducted using multicenter, prospective cohort studies with continuous sleep monitoring and dynamic inflammatory assessments to further validate the effectiveness of combined assessment and intervention strategies on the prevention and amelioration of PSD, PSA and PSCI. Further, understanding the temporal relationship between sleep, inflammation changes, and post-stroke outcomes may yield more solid evidence for targeted psychiatric interventions.

CONCLUSION

There is a strong link between post-stroke sleep disturbance and increased acute-phase inflammatory burden and the development of PSD, PSA, and PSCI. A predictive model that combines several inflammatory markers using PCA with sleep status assessment has a fair discriminative power and can be used as a clinical risk stratification tool. This can help to screen for mental health problems and provide early personalized intervention to stroke patients. A comprehensive approach to early interventions, both in sleep disturbance and inflammatory status, could help to reduce the risk of neuropsychiatric and cognitive disorders following stroke, improving the chances of better long-term rehabilitation outcomes and quality of life for survivors.