Alteration pattern of serum albumin levels in schizophrenia from first episode through remission to relapse: a longitudinal study

Abstract

Serum albumin, a reverse acute-phase protein, tends to decrease in response to acute clinical conditions. We hypothesized that albumin levels would exhibit state-dependent dynamics, with distinct patterns between acute episodes and remission states in schizophrenia. To test this, we conducted a retrospective longitudinal study to investigate the dynamic serum albumin levels in 148 schizophrenia patients, starting from their first episode through remission and subsequent relapse. A matched general population sample served as the control group. Classification models were developed using albumin levels (Albumin _current) and changes (ΔAlbumin₁ = Albumin _current−Albumin _{previous remission}, and ΔAlbumin₂ = Albumin _current−Albumin _{previous acute episode}) to distinguish clinical states. Model performance was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC). Serum albumin levels were significantly lower during acute episodes (first episode: 45·6, standard deviation (SD) 4·0 g/L; relapse: 44·9, SD 4·0 g/L) compared to remission (48·6, SD 2·9 g/L) and matched controls (48·6, SD 3·4 g/L). Patients in remission showed albumin levels comparable to controls. These findings remained consistent after adjustment for potential confounders using mixed-effects model and in sex-stratified analyses. The classification model incorporating ΔAlbumin₁ and ΔAlbumin₂ achieved an AUC of 0·88 (95% CI: 0·84, 0·92) in distinguishing acute episodes from remission. These findings highlight serum albumin’s potential as a clinically useful biomarker of illness activity/state and mental stress in schizophrenia, with utility in differentiating between acute and remitted states.

Introduction

Numerous studies have consistently demonstrated lower serum albumin levels in patients with schizophrenia compared to healthy controls [1–18]. Previous investigations of serum albumin levels in schizophrenia have predominantly focused on hospitalized patients during acute episodes, including first-episode cases (see Table 1 [1–21]). However, data on patients in clinical remission remain limited and inconsistent. Only four studies have focused on stable, remitted, or recovered states: two cross-sectional studies [2, 10], one longitudinal observational follow-up from acute to remission [20], and one clinical trial with an on–off–on haloperidol design [17]. Wong et al. [2] and Lu et al. [10] both found reduced albumin in remitted patients compared with controls, whereas Solberg et al. [20] reported normalization to control levels during remission. Yao et al. [17], in a clinical trial with an on–off–on haloperidol design, observed persistently lower albumin levels in clinically stable male patients irrespective of haloperidol use. These conflicting results may arise from variations in diagnostic criteria for clinical stability/remission/recovery, differences in patient age, illness duration, or medication regimens. To date, no longitudinal studies have examined the dynamic changes in serum albumin levels from the first episode, through clinical remission to subsequent relapse in patients with schizophrenia. Investigation beginning from the initial stage of disease would provide critical advantages: (1) better controlling for age, illness duration and chronicity confounders, and (2) determining whether reduced albumin levels are specifically associated with acute episodes, independent of chronic illness-related confounding factors.

Table 1.

Overview of studies on plasma/serum albumin levels in patients affected by schizophrenia.

Study ID	Design	Participants characteristics	Control for confounders	Main results
Wong 1993 Wong 1996	Cross-sectional	Series 1: SZ (n = 98): (Recovery 44, Chronic 39, Acute 15); HC (n = 41) Series 2: Acute phase SZ (n = 50); HC (n = 50)	NA	Series1: SZ at different stages ↓ vs. HC Series2: acute stage SZ ↓ vs. HC
Ebuehi 2001	Cross-sectional	Newly admitted drug-free SZ (n = 20); HC (n = 20)	NA	SZ vs. HC (NS)
Huang TL 2022	Cross-sectional	Acute phase SZ (n = 106); HC (n = 32)	Age	SZ ↓ vs. HC
Reddy 2003	Cross-sectional	Neuroleptic-naïve FES (n = 31); HC (n = 40)	Age, sex	FES ↓ vs. HC
Garcia-Unzueta 2003	Cross-sectional	SZ outpatients on stable treatment (n = 54); HC (n = 54)	Sex, age	SZ ↓ vs. HC
Pae 2004	Cross-sectional	1. Drug-naïve FES (n = 47); 2. Risperidone-treated chronic schizophrenia inpatients (RCS) (n = 55); 3. HC (n = 68)	Age, gender, BMI, smoking	SZ ↓ vs. HC; FES vs. RCS (NS)
Dadheech 2012	Cross-sectional	SZ (n = 50): (Acute 40, Chronic 10); HC (n = 50)	Age	SZ ↓ vs. HC. Chronic ↓ vs. Acute state
M Uzbekov 2013	Cross-sectional	Drug-naive FES (n = 26); HC (n = 15)	NA	FSZ ↓ vs. HC
Chen 2018; Yin 2022	Cross-sectional	SZ inpatient without drug in the recent 2 weeks (n = 34); HC (n = 136)	Age, gender, education	SZ ↓ vs. HC
Lu 2020	1. Cross-sectional 2. Longitudinal	1. Acute stage AS (n = 107: antipsychotics-use AU:56, antipsychotics-naïve/free ANF: 51); Remission stage RS (n = 101); HC (n = 273) 2. SZ in the acute stage were followed up for 12 weeks n = 77	Stratified analysis by gender	1. SZ acute state ↓ vs. HC, SZ AU ↓ ANF↓ vs. HC; SZ - RS↓ vs. HC 2. Post treatment ↓ vs. Pre
Huang K 2022	Cross-sectional	FES (n = 53); anti-NMDAR encephalitis (n = 53); HC (n = 59)	NA	FSZ ↓ vs. HC
Yan 2022	Cross-sectional	Drug-naïve FES (n = 79); HC (n = 36)	NA	SZ ↓ vs. HC
Xu 2023	Cross-sectional	Acute stage SZ (n = 5577); HC (n = 5000)	Age, sex	SZ ↓ vs. HC
Li 2024	Cross-sectional	SZ inpatient on a stable dose of antipsychotics for at least 8 weeks. (n = 133); HC (n = 120)	Gender, age, education, BMI	SZ ↓ vs. HC
Yeşilkaya 2024	Cross-sectional	Drug-naïve FES with current suicidal behavior (FES-S) (n = 31); drug-naive FES without suicidal behavior (FES-NS) (n = 69); HC (n = 127)	Age, gender, BMI	FSZ ↓ vs. HC
Solberg 2019	Longitudinal follow up 5 years	Baseline: SZ in the acute phase; Follow-up: SZ in a stable phase SZ (n = 55); HC (n = 51)	Sex, age, smoking habit	SZ acute phase↓ vs. SZ stable phase
Yao 2000	Clinical trial, on/off/on haloperidol	SZ, clinically stable outpatients treated with antipsychotics for 3 months on (haloperidol) (n = 46), off (drug free) (n = 35); HC (n = 31)	Age	SZ ↓ vs. HC. SZ haloperidol vs. SZ drug - free (NS)
Yuan 2020	Cross-sectional	SZ (n = 163): aripiprazole (n = 51), olanzapine (n = 48), paliperidone (n = 33), amisulpride (n = 25), and non-medication (n = 6); HC (n = 75)	Age, sex, ethnicity	SZ ↓ vs. HC. Among individual antipsychotics and Nonmedication (NS)
Zhai 2018	Longitudinal	Drug-naive adolescents and young adults FES (n = 342); HC (n = 342)	Age, sex, ethnicity	Drug-naive FSZ ↓ vs. HC; Post treatment ↓ vs. Pre

SZ schizophrenia, FES first-episode schizophrenia, HC health control, NA not available, ↓/↑, Significant Lower/Higher, NS Not Significant.

Acute episodes of schizophrenia, encompassing both first episodes and relapses, are frequently characterized as acute bio-toxic events [21–25]. As a reverse acute phase protein, albumin, decreases under clinical conditions marked by acute inflammation and severe stress, with lower levels consistently correlating with poor clinical outcomes [26–31]. We hypothesized that serum albumin levels in schizophrenia would exhibit a state-dependent alteration pattern. Specifically, we proposed that albumin levels would be lower during acute episodes (including both first episode and relapse) than during clinical remission. To test this hypothesis, we conducted a retrospective longitudinal study in a real-world setting to analyze the dynamic changes in serum albumin levels from first episode (baseline), through subsequent clinical remission, to first relapse in the same patient. Furthermore, we aimed to evaluate the potential utility of albumin levels and/or changes for classifying clinical states (acute episode and remission), using clinical status classification models that are applicable to various visit-level scenarios.

Methods

Study design, data source and study population

This retrospective longitudinal study was approved by the Ethics Committees of “the Second Affiliated Hospital of Xinxiang Medical University”, also known as the “Henan Mental Hospital”. Data were obtained from the records of the Henan Severe Mental Disorder Management and Treatment Program at this institution. This ongoing program, initiated in 2013, is embedded real-world clinical practice and aims to standardize the treatment of mental disorders, improve disease management, and enhance patient well-being. All personally identifiable information was anonymized using unique scrambled identification numbers to protect patient confidentiality. The requirement for informed consent was waived for this retrospective analysis.

Based on a priori sample size calculations, a minimum of 54 patients was required for the longitudinal analyses (paired design, effect size d = 0·5, α = 0·05, power = 0·95), and 105 participants per group were needed for case-control comparisons (independent-samples design, same parameters). Our final cohort of 148 patients exceeded these minimum requirements, thereby ensuring statistical power remained at or above 95%.

Patient selection was based on individuals enrolled in the Henan Severe Mental Disorder Management and Treatment Program who were diagnosed with first-episode schizophrenia between January 2016 and December 2019. First-episode status was defined as an illness duration not exceeding two years. Eligible patients comprised those who: (1) received inpatient treatment for their first acute episode until achieving clinical remission, (2) were subsequently followed as outpatients, and (3) were later rehospitalized for a relapsed episode. All schizophrenia diagnoses and clinical state classifications (first episode, remission, relapse) were independently assessed by two staff psychiatrists and confirmed by a senior psychiatrist, in strict accordance with the “Chinese National Guidelines for Schizophrenia Diagnosis and Treatment” and coded using the “International Classification of Diseases, 10^th revision (ICD-10)”. Exclusion criteria included: (1) history of alcohol or other substance abuse; (2) current pregnancy or lactation; (3) impaired hepatic or renal function; and (4) presence of other serious medical conditions.

A general population control group was selected from participants in the Chinese National Physique Health Database (CNPHD; http://cnphd.bmicc.cn/chs/cn/analysis.php). Exclusion criteria for controls included: (1) any lifetime diagnosis of psychotic disorders (e.g., schizophrenia or affective disorders); (2) history of alcohol or other substance abuse; (3) current pregnancy or lactation; and (4) impaired hepatic or renal function. Patients were matched (1:1) with control subjects randomly selected from the CNPHD based on sex, age ( ± 2 years), ethnicity, and residence (same district).

Data collection

Demographic and clinical data were extracted from the Henan Severe Mental Disorder Management and Treatment Program database. The collected variables included sex, age, ethnicity, diagnosis, illness duration (months), preadmission medications, medications during hospitalization, and maintenance treatment during remission, and laboratory values (alanine aminotransferase [ALT], creatinine, and albumin levels, and corresponding measurement dates.

During hospitalizations for both the first episode and relapse, antipsychotic doses were gradually titrated according to individual treatment response and tolerability, typically stabilizing approximately two weeks after initiation. Medication regimens and dosages were therefore recorded near this two-week time point during both hospitalization periods. For the clinical remission phase, maintenance treatment regimens were documented at the time of albumin measurement.

For both the first episode and relapse hospitalizations, albumin levels measured during the following time windows were extracted: (1) at admission, (2) on treatment days 15–30, and (3) on treatment days 31–60. For the clinical remission phase, the albumin value obtained 3–6 months post-discharge was selected, prioritizing the measurement closest to the 4-month timepoint.

The study timeline is schematically illustrated in Fig. 1.

Fig. 1. The schematic diagram of the study timeline.

↑: the time of measurement of albumin. FES: first-episode schizophrenia.

Statistical analysis

Descriptive analysis

Continuous variables conforming to a normal distribution were described using mean and standard deviation (SD), while those deviating from a normal distribution were described using median and interquartile range (IQR). Categorical variables were presented as frequencies and percentages. Homogeneity of variances between groups was assessed using Levene’s test.

Medication use and serum albumin levels

Serum albumin levels at admission were compared between patients with different medication statuses (medicated vs. drug-free) using independent samples t-tests. “Drug-free” status at baseline was defined as: (1) no previous exposure to psychiatric medications (including antipsychotics, mood stabilizers, and antidepressants); (2) psychiatric medication use for less than two weeks; or (3) no use of psychiatric medications for at least two weeks prior to admission. Changes in albumin levels from admission (pre-treatment baseline) to post-treatment during hospitalization were analyzed using repeated-measures analysis of variance (ANOVA). Additional subgroup analyses were stratified by: (1) sex, (2) admission medication status (medicated vs. drug-free), and (3) treatment strategy (monotherapy vs. combination therapy).

Comparison of serum albumin levels in healthy controls and across clinical states

Serum albumin levels were compared between patients and healthy controls using independent samples t-tests. Levels across different clinical states (first episode, relapse, and remission) were analyzed using repeated-measures ANOVA. For statistically significant findings, post hoc pairwise comparisons were conducted between specific clinical states. All analyses were repeated with sex stratification.

Multiple linear regression was employed to assess the association between baseline albumin levels and potential confounding factors, including age (years), sex (male vs. female), illness duration (months), body mass index (BMI), and medication status (medicated vs. drug-free).

A linear mixed-effects model was implemented to evaluate the association between clinical state and serum albumin levels while adjusting for covariates including age, sex, and illness duration.

Development of classification models

To evaluate the utility of albumin levels and/or their dynamic changes for distinguishing clinical states, we developed clinical status classification models applicable to various clinical scenarios using visit-level data. Based on available serum albumin measurements at each visit, we defined four scenarios: (1) current albumin level only (variable: Albumin _current); (2) current albumin level and previous remission value (variables: Albumin _current and ΔAlbumin₁, where ΔAlbumin₁ = Albumin _current−Albumin _{previous remission}); (3) current albumin level and previous acute episode value (variables: Albumin _current and ΔAlbumin₂, where ΔAlbumin₂ = Albumin _current−Albumin _{previous acute episode}); and (4) current albumin levels, previous remission, and previous acute episode albumin values (variables: ΔAlbumin₁ and ΔAlbumin₂). Four logistic regression models were developed to distinguish between acute and remissive states using these albumin parameters at the visit level. Each model corresponded to one of the clinical scenarios described above. Model performance to distinguish clinical states (i.e., acute episode vs. remission) was assessed by calculating the area under the receiver operating characteristic (ROC) curve (AUC).

All analyses were conducted using IBM SPSS 27·0, and statistical significance was set at a P-value ≤ 0·05 for a two-tailed test.

Results

Study population characteristics

A total of 148 cases of schizophrenia and 148 matched healthy controls were included in the study. At baseline, the median age of patients experiencing their first episode of schizophrenia was 19·0 years (IQR: 9·0) and the median illness duration was 5·0 months (IQR:11·0). There was no statistically significant difference in BMI between patients at baseline (median 21·3 kg/m², IQR: 4·7) and matched controls (median 21·8 kg/m², IQR: 5·1). The study subjects demonstrated normal liver and kidney function test. The median duration of the first hospitalization was 68 days (IQR: 33). The median interval from discharge to albumin measurement during the remission state was 4·7 months (IQR: 3·1). The median interval from discharge to albumin measurement at admission for their relapse was 18·5 months (IQR:13·8).

Medication use and serum albumin levels

At admission of the first hospitalization (at baseline), 100 (67·6%) patients were drug-free. At readmission of the second hospitalization, 81 (54·7%) patients were drug-free. There was no significant difference in serum albumin levels between drug-free and medicated patients at admission, in either the first or second hospitalization.

The medication treatment during hospitalization for their first episode and the relapse, and the maintenance treatment in remission are detailed in Table 2. Patients exhibited a modest but statistically significant decrease in albumin levels after 15–30 days and 31–60 days of hospitalization treatment compared to admission levels, with this pattern being consistent across both hospitalizations (Supplementary Table 1). This albumin reduction pattern was independent of admission medication status (medicated or drug-free at admission, Supplementary Table 2) and treatment strategy during hospitalization (monotherapy or combination therapy) (Supplementary Table 3). Similarly, no significant differences in albumin levels were observed between monotherapy and combination therapy patients in remission state (Supplementary Table 3). No patients had clinically identified hypoalbuminemia (albumin < 35 g/L). No other indicators of liver and kidney function were significantly changed.

Table 2.

The medication treatment during hospitalization for their first episode and relapse, and the maintenance treatment in remission state.

Drug	1st hospitalization for the first episode		Remission		2nd hospitalization for the relapse
	n	%	n	%	n	%
Antipsychotics
Olanzapine	66	44·6	58	39·2	52	35·1
Risperidone	28	18·9	32	21·6	23	15·5
Aripiprazole	18	12·2	16	10·2	19	12·8
Amisulpride	9	6·1	9	6·1	7	4·7
Ziprasidone	5	3·4	3	2·0	1	0·7
Clozapine			10	6·8	21	14·2
Quetiapine	14	9·5	16	10·8	12	8·1
Perphenazine	2	1·4			1	0·7
Paliperidone	4	2·7	3	2·0	2	1·4
Sulpiride	2	1·4	1	0·7	1	0·7
Fluphenazine					1	0·7
Perospirone					7	4·7
Mood stabilizers
Lithium	3	2·0	4	2·7	10	6·8
Valproate	54	36·5	38	25·7	50	33·8
Oxcarbazepine	6	4·1	6	4·1	4	2·7
Antidepressants	19	12·8	29	19·6	20	13·5
Anxiolytics	86	58·1	28	18·9	70	47·3
Treatment strategy
Monotherapy	84	56·8	75	50·7	91	61·9
Combination therapy	64	43·2	73	49·3	56	38·1

We did not further analyze the differences among individual mono-antipsychotic therapies and their dose-response effects on albumin levels, because there was no enough statistical power due to the limited sample size of subgroups.

Comparison of serum albumin levels across healthy controls and schizophrenia patients in acute episodes or remitted states

Patients with schizophrenia during either first-episode or relapse states exhibited significantly lower serum albumin levels compared to matched healthy controls (Fig. 2). Among female patients, mean serum albumin levels were 45·0 g/L (SD 4·0) for the first episode and 44·4 g/L (SD 4·1) for relapse states, compared to 48·4 g/L (SD 3·5) in healthy controls (P < 0·01, t-test). Similarly, male patients showed levels of 46·5 g/L (SD 3·9) for the first episode and 45·6 g/L (SD 3·8) for relapse states versus 48·9 g/L (SD 3·2) in healthy controls (P < 0·01, t-test). Notably, schizophrenia patients in clinical remission demonstrated serum albumin levels comparable to those of age-matched healthy controls.

Fig. 2. Serum albumin levels in healthy controls and schizophrenia patients across the first episode, remission, and relapse states.

A Serum albumin levels were described as mean and standard deviation (SD). There was no statistically significant difference in variance between the groups subjected to statistical comparison. **: Compared to the control group using independent t-tests, P < 0·01. ^##: Compared to the remission state using repeated measures ANOVA with post hoc tests, P < 0·01. B Serum albumin levels were presented as mean with 95% confidence intervals (95% CI). A linear mixed-effects model was used to compare albumin levels across the first episode, remission, and relapse states, adjusting for age, sex, and duration of illness. ^##: Compared to the remission state, P < 0·01.

Serum albumin levels were significantly lower during both first-episode and relapse states compared to the remission states (Fig. 2A). Among female patients, mean levels were 45·0 g/L (SD 4·0) during the first episode and 44·4 g/L (SD 4·1) during relapse, versus 48·1 g/L (SD 3·0) during remission states (n = 84) (P < 0·01, repeated-measures ANOVA with post hoc testing). Similarly, male patients showed mean levels of 46·5 g/L (SD 3·9) during first episode and 45·6 g/L (SD 3·8) during relapse versus 49·4 g/L (SD 2·7) during remission (n = 64) (P < 0·01, repeated-measures ANOVA with post hoc testing). The effect sizes (Cohen’s d) are presented in Supplementary Table 4. Although a trend toward lower levels was observed during relapse compared to first episode, this difference did not reach statistical significance (P > 0·05, repeated-measures ANOVA with post hoc testing; Supplementary Table 4).

In multiple linear regression models of patients at baseline, serum albumin levels showed significant associations with age and sex (Supplementary Table 5). Results from the linear mixed-effects model, adjusted for potential confounding factors, were consistent with those from repeated-measures ANOVA with post hoc testing (Fig. 2B).

Performance of albumin levels and changes in classifying clinical states

We developed four classification models to differentiate between clinical states (acute episode vs. remission) using albumin levels and/or their dynamic changes. The model incorporating both ΔAlbumin₁ (ΔAlbumin₁ = Albumin _current−Albumin _{previous remission}) and ΔAlbumin₂ (ΔAlbumin₂= Albumin _current−Albumin _{previous acute episode}) demonstrated the highest classification performance, achieving an area under the ROC curve of 0·88 (95% confidence interval: 0·84, 0·92). The classification model equation, along with corresponding AUC values, sensitivity, and specificity, are presented in Table 3. Corresponding ROC curves are shown in Fig. 3.

Table 3.

Clinical state classification models using serum albumin levels and/or changes as inputs.

Models: input and equation	AUC (95%CI)	Sensitivity	Specificity
Ln(P/1-P) = −16·049 - 0·469 sex + 0·18 age (year) + 0·12 illness duration (month)+ 0·336 Albumin current	0·81 (0·76, 0·86)	0·86	0·70
Ln(P/1-P) = −6·243 + 0·144 Albumin current + 0·330 ΔAlbumin1	0·84 (0·78, 0·89)	0·96	0·78
Ln(P/1-P) = −9·965 + 0·202 Albumin current + 0·342 ΔAlbumin2	0·86 (0·81, 0·91)	0·77	0·93
Ln(P/1-P) = 0·329 ΔAlbumin1 + 0·329 ΔAlbumin2	0·88 (0·84, 0·92)	0·85	0·83

P Probability of remission state, 1-P Probability of relapse state.

Sex: male=1, female=0 (reference).

ΔAlbumin₁ =Albumin _current−Albumin _{previous remission}.

ΔAlbumin₂ = Albumin _current−Albumin _{previous acute episode.}

Fig. 3. ROC curves for clinical status classification models based on serum albumin measurements at different clinical visits.

$\begin{array}{c}\mathrm{\Delta }{\mathrm{Albumin}}_1={\mathrm{Albumin}}_{\mathrm{current}}-{\mathrm{Albumin}}_{\mathrm{previous}\mathrm{remission}}\hfill \\ \mathrm{\Delta }{\mathrm{Albumin}}_2={\mathrm{Albumin}}_{\mathrm{current}}-{\mathrm{Albumin}}_{\mathrm{previous}\mathrm{acute}\mathrm{episode}}\hfill \end{array}$.

Discussion

Main findings

This longitudinal study reveals characteristic alteration patterns in serum albumin levels across different clinical states of schizophrenia, with levels significantly decreasing during acute phases (both first episode and relapse) and normalizing during clinical remission. The findings remained robust after adjustment for potential confounders including age, sex, and illness duration using mixed-effects modeling. Furthermore, classification models utilizing albumin levels and/or their changes demonstrated high performance in distinguishing between clinical states.

Interpretations

Multiple factors are known to affect serum albumin levels. The primary pathological conditions known to lower serum albumin levels include nephrotic syndrome, hepatic dysfunction, and transcapillary escape in the immune-inflammation process [32]. However, these conditions were not present in our study population. Other established influencing factors include aging, smoking, and malnutrition [28, 33–35]. Although low serum albumin levels are frequently suspected to be attributed to malnutrition, current evidence indicates only a weak association between albumin levels and nutritional status [27, 28, 36]. As noted in previous research, serum albumin levels “do not consistently or predictably change with weight loss, calorie restriction, or nitrogen balance” [36], and typically decrease during disease states independent of nutrient intake [28]. Previous investigations in schizophrenia populations similarly suggest that suboptimal nutrition alone cannot fully explain the lower albumin levels observed in these patients [6, 33, 34, 37, 38]. Our results are consistent with this view, as multiple linear regression models showed no significant association between serum albumin levels and BMI at baseline in either patients or healthy controls. In our study, patients at baseline had BMI values comparable to those of matched controls. Owing to the unavailability of smoking information, we were unable to control for potential confounding effects of smoking on serum albumin levels.

Several studies have investigated the effects of antipsychotic drug on the serum albumin levels. Lu et al. reported that schizophrenia patients in the acute stage have lower albumin levels than healthy controls, while no difference in albumin levels between medicated and antipsychotics-naïve/free subgroup [10]. Our present study also found there was no significant difference in serum albumin levels between drug-free and medicated patients at admission, in either the first or second hospitalization. In a clinical trial using an on-off-on haloperidol treatment design, Yao et al. reported no significant difference in albumin levels between haloperidol treatment (“on” phase) and drug-free (“off” phase), indicating albumin levels were not significantly affected by haloperidol treatment and withdrawal [17]. On the other hand, our previous research indicated a modest but statistically significant decrease in serum albumin levels following initial antipsychotic exposure compared to baseline (drug-naïve) levels [21]. Our present study similarly demonstrated a modest yet statistically significant decline in albumin levels at 15–30 and 31–60 days of treatment compared to admission levels, a pattern consistent across both first-episode and relapse hospitalizations. The modest reduction in albumin during acute-phase treatment has also been reported by other research group [10]. Due to the limited subgroups sample size, we did not analyze the effects of individual antipsychotics or dosages on serum albumin level changes during acute-phase treatment. Yuan et al. reported there were no significant differences in albumin levels among various antipsychotics and non-medication subgroup in a cross-sectional study [18]. Notably, our present study suggests that medication effects on albumin level is reversible, as patients in remission receiving maintenance antipsychotic treatment exhibited serum albumin levels comparable to healthy controls. These findings indicate that while medication may exert modest effects on albumin levels, illness activity, clinical state, and associated stress responses likely represent the primary drivers underlying the dynamic alterations in albumin levels observed in our study.

Albumin serves several crucial physiological functions. As the predominant antioxidant in circulation, it accounts for over 70% of free radical-trapping activity in serum [28, 35]. Reduced serum albumin levels have been consistently observed in pathological conditions characterized by immune-inflammatory activation and oxidative stress [28, 32, 35, 39–41]. The consistent decrease in albumin levels during acute critical illness has led to its recognition as a “reverse acute-phase reactant” [41]. Functioning as a sacrificial antioxidant with a high turnover rate (half-life of approximately 17·3 days) and no recycling pathway [28], serum albumin levels typically normalize rapidly following recovery from acute phase, provided hepatic biosynthesis remains unimpaired.

Dysfunction of the immune-inflammatory system is considered to play a critical role in the pathophysiology of schizophrenia, [22, 23]. Notably, prior studies have demonstrated elevated cerebrospinal fluid (CSF) albumin levels or increased CSF/ serum (plasma) albumin ratios in patients with schizophrenia compared to healthy controls [42–48]. Most such investigations have been conducted in acute inpatient settings [45–48]. The elevated CSF albumin parameters may reflect combined alterations in blood-brain barrier integrity, increased permeability, and neuroinflammatory processes [42, 49, 50]. However, longitudinal data on CSF dynamics across different illness states remain limited [42]. Although serum albumin has been proposed as a potential trait marker in schizophrenia, with reported reductions across various stages compared to healthy controls [1, 2, 10], these previous studies employed cross-sectional designs, and it was difficult to determine whether reduced albumin levels were independent of chronic illness-related confounding factors. Serum albumin did not exhibit trait-marker properties in our present longitudinal study, as albumin levels demonstrated significant state-dependent fluctuations. These dynamic alterations more likely reflect illness activity and stress responses than trait mechanisms. The observed pattern of serum albumin levels is consistent with that of a “reverse acute-phase reactant” potentially serving as an indicator of the severity of illness activity, mental stress, immune-inflammatory activation, and oxidative stress across different clinical states (i.e., acute episodes versus remission). The high performance of clinical state classification models suggests albumin’s potential utility in differentiating between acute and remitted states in schizophrenia—particularly when assessed through longitudinal measurements.

Clinical implications

The finding of reduced serum albumin levels during acute episode (both first episode and relapse) but not during remission holds significant clinical relevance. This suggests serum albumin’s potential utility as a state biomarker for objectively differentiating clinical states in schizophrenia.

Current diagnostic practices for schizophrenia and its clinical states (remission, deterioration, and relapse) rely primarily on behavioral observation and symptomatic assessment, lacking reliable objective biomarkers. The condition’s symptomatic heterogeneity and frequent psychiatric comorbidities further complicate state classification. Objective state markers are particularly valuable [22, 23, 51, 52], as they can signal treatment response and relapse, aiding in patient stratification for early warning and interventions to prevent full relapse. The dynamic alterations in serum albumin levels across illness states suggest their potential as an objective, state-specific biomarker for relapse identification and early warning. Serum albumin testing offers practical advantages: it is simple, minimally invasive, inexpensive, and routinely monitored during antipsychotic treatment. Additionally, albumin levels demonstrate low inter- and intra-individual variability [21, 27, 28]. Our clinical state classification models utilizing albumin parameters showed high discriminatory performance, suggesting potential utility for patient stratification in clinical practice. The discovery of albumin’s dynamic patterns may create opportunities for early intervention and relapse prevention, particularly during critical early illness stages where timely intervention significantly improves prognostic outcomes. Future clinical studies should investigate correlations between albumin dynamics and illness severity, treatment response, and long-term prognosis in schizophrenia.

Strengths and limitations

First, to our knowledge, this represents the first longitudinal study to evaluate dynamic changes in serum albumin levels across different clinical states (first episode, remission, and relapse) in schizophrenia. The longitudinal design, featuring repeated albumin measurements during these clinical states, combined with a matched general population control group, enabled clear characterization of albumin alteration patterns in schizophrenia and direct comparison of levels between acute and remitted states. Second, by including only patients during their first episode at baseline, we minimized potential confounding effects related to chronic illness duration and comorbidities. Third, we developed four clinical state classification models based on albumin parameters, which demonstrated high discriminatory performance. These models show promise as practical tools for patient stratification in real-world settings.

This study has several limitations. First, selection bias may exist in this real-world study, as hospitalization decisions were based on doctor assessments and consent from patients or their guardians. Second, the lack of smoking information precluded control for its confounding effects on serum albumin levels. Third, due to the limited number of patients receiving monotherapy with any specific antipsychotic drug, we could not conduct meaningful comparisons of individual antipsychotics or dosages effects on serum albumin levels. This limitation precludes definitive conclusions about medication-specific influences on albumin dynamics. Fourth, there was a decreasing trend in serum albumin levels during relapse states compared to their first episodes. It remains unclear whether the dynamic changes would persist in schizophrenia patients with a long illness duration. Fifth, while the observed reductions in serum albumin during acute episodes may represent a peripheral dynamic process of a “reverse acute-phase reactant”, the absence of paired CSF immune-inflammatory markers to prevents us from connecting these peripheral changes to central neuroinflammatory processes.

Conclusion

This longitudinal study identified a state-dependent pattern of serum albumin dynamics in schizophrenia, characterized by marked decreases during acute episodes and normalization during remission, highlighting albumin’s potential as a clinical biomarker of illness activity/state and mental stress, and offering an entry point for further research into its utility in differentiating acute and remitted states.

Funding

The funding sources had no contribution in study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the article for publication.

Data availability

The hospital record data used in this study contain sensitive personal health information and cannot be made publicly available due to privacy and confidentiality restrictions. De-identified data may be made available to qualified researchers upon reasonable request, subject to approval by the institutional ethics board and the hospital’s data access committee. Proposals for the use of data and requests for access should be directed to zds@xxmu.edu.cn and zhangruilingxx@163.com. To gain access, researchers will need to sign a data access agreement with the study sponsor (The Second Affiliated Hospital of Xinxiang Medical University, Xinxiang, China).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41398-026-03885-y.