High-density lipoprotein (HDL) in older patients with schizophrenia: implications for clinical and cognitive outcomes

Abstract

Background

This study investigated the role of high-density lipoprotein (HDL) in older patients with schizophrenia (SCZ), examining its associations with clinical symptoms, cognitive function, and other conventional lipid parameters (triglycerides, total cholesterol), as well as its longitudinal changes during a 6-month follow-up period under routine clinical care.

Methods

A total of 77 participants (30 healthy controls [HC] and 47 SCZ patients) were stratified into high- and low-HDL subgroups based on a cutoff of 0.91 mmol/L. Baseline comparisons revealed significantly lower HDL levels in the SCZ group compared to HC. The SCZ group, as expected, exhibited significant psychopathology (assessed by PANSS) and impaired cognitive performance across multiple RBANS domains.

Results

HDL subgroup analysis demonstrated that among SCZ patients, the high-HDL subgroup exhibited higher PANSS scores and worse performance in specific delayed memory domains compared to the low-HDL subgroup. In SCZ patients, correlation analysis showed a positive association between HDL and PANSS scores (r = 0.28, p < 0.05) and a negative correlation with performance in one delayed memory task (Figure Recall: r = -0.30, p = 0.04). Longitudinal follow-up after 6 months in the SCZ group revealed a reduction in HDL levels (though not statistically significant), along with decreased PANSS scores and improved RBANS total scores.

Conclusion

These findings reveal a complex relationship between HDL and SCZ. While SCZ patients have lower HDL than healthy controls, within the patient group, higher HDL levels were paradoxically associated with greater symptom severity and poorer memory performance. This suggests that in older patients with chronic SCZ, the functional quality of HDL may be more critical than its absolute concentration.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-026-07842-3.

Introduction

Schizophrenia (SCZ) is a severe and chronic psychiatric disorder characterized by persistent psychotic symptoms, progressive cognitive impairments, and significant functional decline [1]. The clinical management of older patients with SCZ presents unique challenges, as the interplay between the long-term course of the illness, cumulative exposure to antipsychotic medications, and age-related physiological decline often results in exacerbated metabolic disturbances, accelerated cognitive deterioration, and increased treatment resistance compared to younger populations [2]. Among these comorbidities, dysregulation of lipid metabolism has emerged as a critical factor potentially bridging systemic health with brain pathophysiology and clinical outcomes in SCZ [3, 4].

High-density lipoprotein (HDL), traditionally termed “good cholesterol,” is renowned for its cardioprotective role, primarily through reverse cholesterol transport. Beyond this, HDL exerts potent anti-inflammatory, antioxidant, and endothelial-protective effects, which are increasingly recognized as vital for neuroprotection and maintaining cerebral vascular health [5]. Conversely, metabolic syndrome—a cluster of conditions including dyslipidemia (often featuring low HDL), insulin resistance, and central obesity—is highly prevalent in patients with SCZ, particularly those on long-term antipsychotic regimens [6, 7]. This metabolic dysregulation is not merely a peripheral side effect; it can drive a state of chronic, low-grade systemic inflammation [8, 9], whereas elevated low-density lipoprotein (LDL) and abnormal triglyceride levels are associated with increased inflammation and cardiovascular risk, which are prevalent in SCZ [10, 11]. This inflammatory milieu is implicated in damaging the central nervous system through multiple pathways, including promoting neuroinflammation, increasing oxidative stress, accelerating neuronal loss, disrupting synaptic pruning mechanisms, and impairing white matter integrity (dysconnection), thereby contributing directly to both cognitive deficits and symptom severity in SCZ [12–14].

Investigating HDL in SCZ is therefore complex. While group-level comparisons frequently report lower HDL in patients versus healthy controls [15], the relationship within the patient population appears nonlinear and may be confounded by several key factors. Antipsychotic medications, especially second-generation agents, have well-documented adverse effects on weight and lipid profiles, variably impacting HDL levels [16]. Treatment-resistant SCZ may constitute a distinct biological subtype with different metabolic correlates and inflammatory states [17]. Significant sex differences exist in lipid metabolism, immune response, and the clinical presentation of SCZ, necessitating careful consideration in analysis [18–21] Furthermore, the concept of cognitive reserve—the brain’s resilience to pathology—adds another layer of complexity. Interestingly, some evidence suggests that in aging populations, higher cognitive reserve (a proxy for more robust neural circuitry) might paradoxically be associated with greater susceptibility to the detrimental cognitive effects of systemic inflammation, indicating a nuanced interaction between brain resilience, metabolic health, and inflammatory insults [22].

Despite this background, critical gaps remain. Most studies on lipids and SCZ are cross-sectional, focusing on younger or first-episode cohorts. The longitudinal trajectory of HDL levels, its specific associations with psychopathology and discrete cognitive domains, and how these relationships evolve with routine treatment in older patients with chronic SCZ are poorly understood. Furthermore, the interplay between HDL, symptom dimensions (positive, negative, general), and cognitive impairment against the backdrop of the confounders requires clarification.

Therefore, this study aims to: (1) Compare baseline lipid profiles, clinical symptoms, and cognitive function between older patients with SCZ and matched healthy controls (HC); (2) Stratify SCZ patients based on HDL levels to investigate the cross-sectional associations between HDL status and psychopathology (measured by PANSS) and cognition (measured by RBANS); and (3) Examine the longitudinal changes in HDL, clinical symptoms, and cognitive performance over a 6-month period of routine clinical care to explore how HDL dynamics relate to treatment response and cognitive outcomes. By addressing these aims, we seek to elucidate the nuanced role of HDL in older SCZ patients, potentially informing more personalized management strategies for this vulnerable population.

Methods

Participants

A total of 77 participants were enrolled in this study, comprising 30 HC and 47 patients with SCZ. Participants were consecutively recruited from the outpatient and inpatient services of our hospital between January 2025 to August 2025. Given the exploratory nature of this study and the challenge of recruiting older patients with specific criteria, a formal power analysis was not conducted a priori. The sample size was determined based on feasibility and available resources over the recruitment period. Participants were further stratified into high- and low-HDL subgroups based on a cutoff value of 0.91 mmol/L [23, 24], ensuring subgroup comparisons were grounded in metabolic profiles. The HC group consisted of 11 males and 19 females, while the SCZ group included 18 males and 19 females. All participants were aged between 60 and 70 years and met the DSM-5 criteria for SCZ group or were free of any psychiatric disorders (HC group), ensuring demographic and diagnostic consistency across the cohort. This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethical Review Committee of the Third People’s Hospital of Foshan (Ethics approval number: FSSY-LS202411). Every participant provided their consent, and written informed consent was obtained from all participants or their legal guardians. Clinical trial number: not applicable.

Inclusion and exclusion criteria

Participants were selected based on strict inclusion and exclusion criteria to minimize confounding factors. Inclusion criteria required participants to be aged ≥ 60 years, meet DSM-5 diagnostic criteria for SCZ (SCZ group) or have no history of psychiatric disorders (HC group), and maintain a stable medication regimen for at least 3 months prior to enrollment, ensuring baseline stability. Exclusion criteria eliminated individuals with a history of neurological disorders, substance abuse, or major medical comorbidities (diabetes, cardiovascular disease), as well as those currently using medications affecting lipid metabolism or cognitive function. Additionally, participants with cognitive impairment due to non-psychiatric causes (dementia, traumatic brain injury) were excluded to isolate the effects of SCZ and metabolic status on outcomes.

Clinical assessments

Clinical evaluations were conducted to comprehensively assess psychopathology, cognitive function, and metabolic parameters. Psychopathology was measured using the Positive and Negative Syndrome Scale (PANSS) [25] in the SCZ group, providing a standardized metric for symptom severity. Cognitive function was evaluated with the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) [26], which quantifies domains such as immediate memory, delayed memory, attention, and visuospatial construction. Metabolic profiles were determined via fasting blood samples, which measured lipid levels (triglycerides [TG], total cholesterol [CHOL], HDL, LDL) and fasting glucose levels, ensuring a holistic assessment of metabolic health alongside clinical and cognitive metrics.

The PANSS comprises 30 items divided into three subscales: Positive (7 items), Negative (7 items), and General Psychopathology (16 items), each rated from 1 (absent) to 7 (extreme) [27]. The RBANS assesses five cognitive domains: Immediate Memory (List Learning and Story Memory), Visuospatial/Constructional (Figure Copy and Line Orientation), Language (Picture Naming and Semantic Fluency), Attention (Digit Span and Coding), and Delayed Memory (List Recall, List Recognition, Story Recall, and Figure Recall) [28]. Fasting blood samples were collected between 7:00 and 8:00 AM after a 12-hour fast. Serum levels of TG, CHOL, HDL, and LDL were measured using standard enzymatic methods on an automated biochemical analyzer.

Study design and procedures

This was a prospective, observational cohort study with a 6-month follow-up period. At baseline (T0), all participants (SCZ and HC) underwent comprehensive clinical, cognitive, and metabolic assessments. The SCZ patients then continued to receive their routine, clinically prescribed antipsychotic treatment throughout the study. No experimental interventions targeting metabolic or cognitive parameters were introduced. The treatment regimen, primarily consisting of olanzapine, risperidone, or aripiprazole, was stabilized for at least 3 months prior to enrollment and adjusted only as clinically necessary by the treating psychiatrist. All baseline assessments were repeated at the 6-month follow-up visit (T1) for the SCZ group to evaluate naturalistic changes. Treatment adherence was monitored through self-report and pill counts.

Clinical outcomes

The primary outcomes focused on changes in PANSS scores, RBANS subdomain scores, and metabolic parameters (HDL, TG, CHOL, LDL, fasting glucose) at baseline and after 6 months of follow-up, capturing both short-term and longitudinal trends. During follow-up, all SCZ patients continued to receive routine antipsychotic treatment (primarily including olanzapine, risperidone, aripiprazole, etc.), with no specific experimental intervention targeting metabolic indices introduced. Secondary outcomes emphasized the association between HDL levels and clinical/cognitive variables, as well as the longitudinal effects of interventions on disease progression. These outcomes were designed to evaluate the interplay between metabolic status, psychopathology, and cognitive function, while also assessing the impact of therapeutic interventions over time.

Follow-up

Participants underwent a 6-month follow-up to monitor changes in clinical and metabolic parameters. At baseline and at the 6-month visit, all assessments were repeated to ensure consistency and reliability of measurements. Adherence to treatment protocols was closely monitored, and any modifications to medication or health status were documented to account for potential confounding variables. This structured follow-up approach enabled the evaluation of long-term trends and the efficacy of interventions in maintaining or improving clinical and metabolic outcomes.

Statistical analysis

Baseline comparisons were conducted using independent sample t-tests or Mann-Whitney U tests for continuous variables (e.g., age, BMI, PANSS scores, RBANS subdomains) and chi-square tests for categorical variables (e.g., gender, education level), ensuring appropriate statistical methods for data distribution. HDL subgroup analysis employed ANOVA or Kruskal-Wallis tests to compare differences in PANSS scores and RBANS subdomains between high- and low-HDL groups. Pearson’s correlation coefficient (r) was calculated to quantify relationships between HDL levels and clinical/cognitive variables. To evaluate longitudinal changes following the 6-month treatment period, paired t-tests were used to compare outcomes between baseline and follow-up within the SCZ patient group. Given the exploratory nature of the subgroup and correlation analyses, and to avoid over-correction in a modestly sized sample, we reported uncorrected p-values but interpreted findings with caution, focusing on effect sizes and consistency across related measures. For the primary longitudinal comparison (T0 vs. T1 in SCZ patients), paired t-tests were used. A two-tailed p < 0.05 was considered statistically significant, with adjustments for multiple comparisons to maintain rigor.

Results

Baseline comparisons between HC and SCZ groups

Table 1 revealed significant differences in baseline characteristics between the HC and SCZ groups. Specifically, HDL levels were significantly lower in the SCZ group compared to the HC group (p < 0.01), while other demographic and metabolic parameters (age, BMI, fasting glucose) showed no significant differences. Additionally, the SCZ group exhibited higher PANSS scores (p < 0.01) and impaired cognitive performance across multiple RBANS domains (e.g., immediate memory, delayed memory, attention) compared to the HC group as shown in Table 1.

Table 1.

Demographic and baseline characteristics of the participants

	HC	SCZ
Participants	30	47
Age (years)	63.67 ± 2.52	64.26 ± 2.59
Gender (M/F)	11/19	18/19
BMI	23.34 ± 3.37	24.25 ± 4.37
Education
Below 9 years, n (%)	12 (40%)	24 (51.1%)
9–12 years, n (%)	6 (20%)	7 (14.9%)
13–17 years, n (%)	12 (40%)	16 (34.0%)
Baseline PANSS score	30.07 ± 0.25	70.51 ± 16.95**
P	7 ± 0	15.06 ± 7.34**
N	7.03 ± 0.18	21.81 ± 5.76**
G	16.03 ± 0.18	33.62 ± 5.76**
RBANS
Immediate memory (Learning)	24.5 ± 6.27	17.13 ± 6.86**
Immediate memory (Story Memory)	11.93 ± 5.02	5.72 ± 4.34**
Visuospatial Construction	17.07 ± 2.46	15.21 ± 4.55*
Language	17.43 ± 4.48	11.85 ± 4.30**
Attention (Digit span)	13.4 ± 2.33	10.57 ± 2.43**
Attention (Coding)	45.07 ± 14.34	27.72 ± 12.57**
Delayed memory (List Recall)	5.47 ± 3.08	2.89 ± 2.60**
Delayed memory (List Recognition)	19.33 ± 12.4	17.40 ± 2.76**
Delayed memory (Story Recall)	6.67 ± 3.30	2.70 ± 2.65**
Delayed memory (Figure Recall)	12.9 ± 4.30	8.57 ± 5.53**
TG	1.43 ± 1.15	1.54 ± 0.97
CHOL	4.81 ± 0.82	4.41 ± 0.80
HDL	1.21 ± 0.28	1.02 ± 0.28**
LDL	2.70 ± 0.65	2.61 ± 0.71
Fasting blood glucose	5.76 ± 1.06	5.33 ± 1.03

Data are presented as mean ± standard deviation. Abbreviations: HC, healthy control; SCZ, schizophrenia; BMI, body mass index; PANSS, Positive and Negative Syndrome Scale (P, Positive; N, Negative; G, General Psychopathology); RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; TG, triglycerides; CHOL, total cholesterol; HDL, high-density lipoprotein; LDL, low-density lipoprotein. *p < 0.05, **p < 0.01 (vs. HC group)

HDL subgroup analysis: impact on PANSS and cognitive function

Participants were stratified into high- and low-HDL subgroups (Table 2). Contrary to expectations, among SCZ patients, the high-HDL subgroup showed significantly higher PANSS total scores compared to the low-HDL subgroup (p < 0.05), indicating more severe psychopathology. Furthermore, the high-HDL subgroup performed worse on specific delayed memory tasks (List Recall and Figure Recall, both p < 0.01). No significant differences were observed in demographic variables across subgroups.

Table 2.

Baseline characteristics stratified by diagnosis and HDL level

	HC (≤ 0.91 mmol/L)	HC (>0.91 mmol/L)	SCZ (≤ 0.91 mmol/L)	SCZ (>0.91 mmol/L)
Participants	5	25	19	28
Age (years)	63.2 ± 1.92	63.76 ± 2.65	64.47 ± 2.55	64.11 ± 2.66
Gender (M/F)	3/2	8/17	9/10	9/9
BMI	22.17 ± 1.76	23.57 ± 3.60	25.61 ± 3.29	23.32 ± 4.81
PANSS score	30.2 ± 0.45	30.04 ± 0.2	64.79 ± 14.00	74.39 ± 17.9*
P	7 ± 0	7 ± 0	13.16 ± 6.41	16.36 ± 7.76
N	7.2 ± 0.45	7 ± 0	20.89 ± 6.05	22.43 ± 5.57
G	16 ± 0	16.04 ± 0.04	30.74 ± 6.56	35.57 ± 9.08*
RBANS
Immediate memory (Learning)	25.2 ± 5.85	24.36 ± 6.45	18.63 ± 7.78	16.11 ± 6.09
Immediate memory (Story Memory)	15 ± 2.55	11.32 ± 5.19	6.74 ± 4.85	5.04 ± 3.90
Visuospatial Construction	17.6 ± 1.67	16.96 ± 2.61	15.68 ± 5.13	14.89 ± 4.18
Language	16 ± 3.39	17.72 ± 4.67	12.74 ± 4.2	11.25 ± 4.34
Attention (Digit span)	13.2 ± 3.83	13.44 ± 20.2	10.79 ± 2.39	10.43 ± 2.49
Attention (Coding)	45.4 ± 10.9	45 ± 15.12	29.63 ± 2.40	26.43 ± 12.22
Delayed memory (List Recall)	6.8 ± 2.17	5.2 ± 3.2	3.11 ± 2.92	2.75 ± 2.40**
Delayed memory (List Recognition)	20 ± 0	19.2 ± 1.32	18.16 ± 2.09	16.89 ± 3.06
Delayed memory (Story Recall)	9 ± 1.41	6.2 ± 3.39	3.47 ± 2.95	2.18 ± 2.33
Delayed memory (Figure Recall)	14.6 ± 3.36	12.56 ± 4.45	10.58 ± 5.72	7.21 ± 5.05*
TG	2.33 ± 2.58	1.25 ± 0.52	1.96 ± 1.36	1.26 ± 0.38*
CHOL	4.38 ± 1.08	4.90 ± 0.76	4.39 ± 0.86	4.43 ± 0.78
HDL	0.78 ± 0.16	1.30 ± 0.22	0.77 ± 0.11	1.18 ± 0.24**
LDL	2.37 ± 0.64	2.76 ± 0.64	2.66 ± 0.71	2.57 ± 0.72
Fasting blood glucose	6.84 ± 2.51	5.59 ± 0.55	5.13 ± 1.13	5.47 ± 0.95

Data are presented as mean ± standard deviation. Abbreviations: HC, healthy control; SCZ, schizophrenia; BMI, body mass index; PANSS, Positive and Negative Syndrome Scale (P, Positive; N, Negative; G, General Psychopathology); RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; TG, triglycerides; CHOL, total cholesterol; HDL, high-density lipoprotein; LDL, low-density lipoprotein. *p < 0.05, **p < 0.01 for comparisons between SCZ (> 0.91 mmol/L) and SCZ (≤ 0.91 mmol/L) subgroups

Correlation analysis of HDL with clinical and cognitive variables

Within the SCZ patient group (n = 47), we examined the foundational relationship between symptom severity and cognitive performance (Supplementary Table 1). As expected, more severe negative symptoms (N) were significantly associated with worse global cognition (RBANS total score: r = -0.41, p = 0.005), particularly in the domains of language (r = -0.59, p < 0.01) and attention (coding: r = -0.40, p < 0.01). General Psychopathology (G) scores also correlated with poorer delayed memory (list recognition r = -0.34, p = 0.02). In contrast, positive symptoms (P) showed no significant correlation with any cognitive index. This pattern confirms the established link between negative symptom burden and cognitive impairment in our cohort.

Table 3 presented the correlation analysis between HDL and clinical/cognitive variables. A positive correlation was found between HDL and PANSS total scores (r = 0.28, p < 0.05), indicating that higher HDL levels were associated with more severe psychopathology. However, correlations between HDL and cognitive performance were generally weak and not statistically significant, except for a negative correlation with Delayed memory (Figure Recall) (r = -0.30, p = 0.04).

Table 3.

Correlation analysis between HDL levels and clinical/cognitive variables in the schizophrenia patient group (n = 47)

		P	N	G	PANSS	RBANS	Immediate memory (Learning)	Immediate memory (Story Memory)	Visuospatial Construction	Language	Attention (Digit span)	Attention (Coding)	Delayed memory (List Recall)	Delayed memory (List Recognition)	Delayed memory (Story Recall)	Delayed memory (Figure Recall)
HDL	r	0.22	0.13	0.28	0.28	-0.18	-0.19	-0.09	-0.09	-0.17	-0.07	-0.13	-0.07	-0.23	-0.24	-0.3
	p	0.14	0.38	0.05	0.05	0.22	0.39	0.56	0.53	0.20	0.65	0.39	0.65	0.32	0.1	0.04

Abbreviations: PANSS, Positive and Negative Syndrome Scale (P, Positive; N, Negative; G, General Psychopathology); RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; HDL, high-density lipoprotein

Longitudinal outcomes after 6-month intervention

Table 4 demonstrated the longitudinal changes over 6 months of routine clinical care in the SCZ group. HDL levels showed a decreasing trend, but this change did not reach statistical significance. Concurrently, significant reductions in PANSS scores (p < 0.01) and improvements in RBANS total and several subdomain scores (p < 0.01) were observed.

Table 4.

Comparison of baseline and 6-month follow-up outcomes in the schizophrenia patient group under routine antipsychotic treatment

	Baseline	6-Month follow-up	Between-group difference (95% CI)
Participants	47	36
PANSS score	70.51 ± 16.95	51.08 ± 11.88**	-19.43 ± 15.87 (-26.02 to -12.83)
P	15.06 ± 7.34	9.11 ± 3.83**	-5.95 ± 6.63 (-8.63 to -3.27)
N	21.81 ± 5.76	17.69 ± 5.76**	-4.11 ± 6.52 (-6.79 to -1.44)
G	33.62 ± 8.43	24.28 ± 5.05**	-9.34 ± 7.70 (-12.50 to -6.18)
RBANS
Immediate memory (Learning)	17.13 ± 6.86	23.53 ± 6.66**	6.40 ± 7.17 (3.42 to 9.38)
Immediate memory (Story Memory)	5.72 ± 4.34	10.03 ± 5.83**	4.30 ± 5.53 (2.08 to 6.53)
Visuospatial Construction	15.21 ± 4.55	16.31 ± 3.40	1.09 ± 4.32 (-0.71 to 2.90)
Language	11.85 ± 4.30	12.47 ± 3.71	0.62 ± 4.29 (-1.17 to 2.41)
Attention (Digit span)	10.57 ± 2.43	10.50 ± 2.71	-0.07 ± 2.72 (-1.20 to 1.05)
Attention (Coding)	27.72 ± 12.57	31.69 ± 12.41	3.97 ± 13.24 (-1.54 to 9.48)
Delayed memory (List Recall)	2.89 ± 2.60	5.22 ± 2.62**	2.33 ± 2.61 (1.18 to 3.48)
Delayed memory (List Recognition)	17.40 ± 2.76	18.39 ± 3.49	0.98 ± 3.36 (-0.38 to 2.35)
Delayed memory (Story Recall)	2.70 ± 2.65	5.06 ± 3.04**	2.35 ± 3.06 (1.11 to 3.60)
Delayed memory (Figure Recall)	8.57 ± 5.53	11.11 ± 5.75*	2.54 ± 6.00 (0.06 to 5.02)
TG	1.54 ± 0.97	1.48 ± 0.74	-0.06 ± 0.91 (-0.45 to 0.33)
CHOL	4.41 ± 0.80	4.26 ± 0.60	-0.15 ± 0.78 (-0.47 to 0.17)
HDL	1.02 ± 0.28	0.93 ± 0.20	-0.09 ± 0.22 (-0.20 to 0.03)
LDL	2.61 ± 0.71	2.49 ± 0.61	-0.12 ± 0.67 (-0.42 to 0.18)
Fasting blood glucose	5.33 ± 1.03	4.96 ± 0.43	-0.36 ± 0.93 (-0.73 to 0.00)

Data are presented as mean ± standard deviation. Abbreviations: PANSS, Positive and Negative Syndrome Scale (P, Positive; N, Negative; G, General Psychopathology); RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; TG, triglycerides; CHOL, total cholesterol; HDL, high-density lipoprotein; LDL, low-density lipoprotein. *p < 0.05, **p < 0.01 (vs. Baseline). Note: During the follow-up period, patients received routine antipsychotic treatment (primarily atypical antipsychotics). No specific experimental intervention targeting lipid metabolism was implemented

Discussion

The HDL paradox in schizophrenia

Consistent with prior observations of metabolic dysregulation in SCZ [29], this study identified several critical associations between HDL levels and clinical and cognitive outcomes in older patients with SCZ. For our study, HDL levels were significantly lower in the SCZ group compared to HC, consistent with prior observations of metabolic dysregulation in SCZ. Notably, HDL showed a positive correlation with PANSS scores, indicating that higher HDL levels were associated with more severe psychopathology. However, its associations with cognitive performance were largely non-significant, with only a specific negative correlation observed in the delayed memory domain. These findings suggest that in older patients with SCZ, higher HDL levels are associated with greater disease severity, while their relationship with cognitive impairment appears more limited and domain-specific.

Potential mechanisms: dysfunctional HDL and systemic inflammation

After 6 months of follow-up, a declining trend in HDL levels was observed in the SCZ group, coinciding with reduced PANSS scores and enhanced cognitive function (RBANS total scores) [30]. Whether the association between HDL and cognition persists after treatment requires further investigation. This pattern—where a non-significant decrease in HDL coincides with significant symptomatic and cognitive improvement—raises questions about the role of HDL in treatment response.

The confirmed link between negative symptoms and cognitive deficits in our sample provides context for interpreting HDL’s complex associations. HDL correlated positively with general psychopathology (which itself linked to worse memory) but negatively with delayed memory. This pattern suggests that elevated HDL in this context may not be protective. Instead, it could mark a state of dysfunctional lipid metabolism tied to a broader pathological process. This process might simultaneously drive increased general symptom burden and accelerated cognitive decline, possibly through shared inflammatory pathways [11]. The specificity of findings—linking HDL to general (not positive) symptoms and memory—hints that HDL dysregulation may be more relevant to the deficit or inflammatory dimension of SCZ. The positive correlation between HDL and PANSS scores challenges the conventional view of HDL as a “protective” lipid [31–33]. While HDL is traditionally linked to anti-inflammatory and neuroprotective properties [34], its elevation in SCZ may reflect compensatory mechanisms in response to chronic inflammation or oxidative stress [35, 36]. Alternatively, elevated HDL could be a byproduct of metabolic dysregulation, such as hypertriglyceridemia or insulin resistance, which are common in antipsychotic-treated populations [37, 38]. The negative association with cognition, however, aligns with evidence that HDL dysfunction impairs synaptic plasticity and neurovascular integrity, potentially exacerbating cognitive deficits [39]. Furthermore, biological and preclinical data suggest that systemic metabolic-inflammatory crosstalk can directly impact CNS integrity, promoting neuronal loss and white matter alterations (dysconnection) [14, 40], which could underlie the observed link between dysregulated HDL profiles and specific cognitive impairments.

Association with symptom dimensions and cognitive deficits

The longitudinal data reveal a complex interplay between HDL, treatment response, and cognitive recovery. The reduction in HDL following intervention may reflect metabolic improvements (e.g., reduced inflammation or lipid-lowering effects of antipsychotics), which could explain the decline in PANSS scores. However, the persistence of the HDL-cognition link suggests that HDL dynamics are not solely driven by symptomatic changes but may independently influence neurocognitive resilience. This highlights the need to consider HDL as a potential therapeutic target, even in the context of effective psychiatric treatment.

Longitudinal perspectives and treatment implications

Our findings align with studies demonstrating the role of lipid metabolism in SCZ, particularly in older adults [41]. For instance, lower HDL has been associated with poorer executive function and increased risk of dementia in aging populations, while HDL dysregulation may exacerbate symptoms in SCZ [42]. However, the longitudinal perspective here adds novel insights, as previous research has primarily focused on cross-sectional associations [43, 44]. The persistence of the HDL-cognition link despite treatment underscores the importance of addressing metabolic comorbidities in older patients with SCZ, as residual cognitive impairments may limit functional recovery. Furthermore, it is important to note that our study excluded patients with major diabetes or cardiovascular disease, and the sample did not include individuals on lipid-lowering medications. Therefore, our findings primarily reflect the association patterns of HDL in relatively ‘metabolically stable’ older SCZ patients and may not be directly generalizable to populations with more severe metabolic comorbidities.

Future studies should explore the causal mechanisms underlying HDL’s dual role in psychopathology and cognition, including its interactions with inflammatory markers, oxidative stress, and neuroplasticity. Additionally, interventions targeting HDL (e.g., lifestyle modifications, lipid-modifying therapies) could be tested to determine their impact on both symptom severity and cognitive outcomes. Given the high prevalence of metabolic and cognitive comorbidities in older patients with SCZ, a holistic approach integrating psychiatric, metabolic, and neurocognitive care is essential to optimize long-term patient outcomes.

Limitations

This study has limitations, including its observational design, which precludes causal inferences, and the relatively small sample size, which may limit generalizability. Additionally, the reliance on self-reported adherence and the lack of detailed pharmacological data (e.g., specific antipsychotic agents) could introduce confounding variables. We also lacked direct measures of HDL functionality and specific inflammatory biomarkers to validate the proposed mechanistic links. The role of cognitive reserve, which may modulate susceptibility to metabolic-inflammatory insults [22], was not directly assessed. Finally, while patients received routine clinical care (antipsychotic treatment), no experimental metabolic or cognitive interventions were applied, which clarifies the naturalistic context of the observed longitudinal changes. Despite these constraints, the results highlight the clinical relevance of HDL in older patients with SCZ and underscore the need for further research to clarify its role as a biomarker and therapeutic target. Furthermore, while we have elucidated the correlations between PANSS dimensions and cognitive scores, the underlying mechanisms linking dysfunctional HDL to specific symptom domains (e.g., general vs. negative) and discrete cognitive circuits remain speculative. Future studies incorporating path analysis or mediation models in larger samples are needed to dissect these direct and indirect relationships.

Conclusion and future directions

In conclusion, our study in older patients with chronic SCZ reveals a paradoxical association where higher HDL levels are linked to greater symptom severity and specific memory impairment, challenging a simplistic view of HDL as a protective marker. This highlights the critical importance of investigating the functional quality of HDL rather than just its concentration in this population.

Future research should employ longitudinal designs with larger samples, integrating functional lipid assays, neuroimaging, and immune profiling to delineate causal pathways. From a clinical perspective, managing metabolic health in SCZ requires moving beyond basic lipid panels. Promising non-pharmacological approaches that may concurrently benefit metabolic health and cognition in SCZ include interventions targeting functional mobility [45], aerobic exercise [46], and combined aerobic exercise with cognitive remediation [47]. Implementing such holistic, lifestyle-oriented interventions into the routine clinical care of older patients with SCZ could serve as a valuable adjunct to pharmacotherapy and is a crucial step toward mitigating cognitive decline and improving long-term functional outcomes.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

BMI

Body Mass Index

CHOL

Total Cholesterol

HC

Healthy Controls

HDL

High-Density Lipoprotein

LDL

Low-Density Lipoprotein

PANSS

Positive and Negative Syndrome Scale

RBANS

Repeatable Battery for the Assessment of Neuropsychological Status

SCZ

Schizophrenia

TG

Triglycerides

Author contributions

SY and JQ contributed equally to this work. FF and SJZ designed the study. SY, JQ, CY, CJ, WM and ZC constructed the database, analyzed the data, and contributed to the statistical design of the study. SY and JQ prepared the first draft of the manuscript. All authors contributed to data interpretation and manuscript preparation, and all have approved the final manuscript.

Funding

This study was supported by the project of Foshan Science and Technology Bureau (2420001003774).

Data availability

The datasets are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethical Review Committee of the Third People’s Hospital of Foshan (Ethics approval number: FSSY-LS202411). Written informed consent was obtained from all participants or their legal guardians.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.