Skip to main content
NCBI home page
Brain, Behavior, & Immunity - Health logoLink to Brain, Behavior, & Immunity - Health
. 2026 Jul 23;56:101312. doi: 10.1016/j.bbih.2026.101312

Transdiagnostic determinants of Growth Differentiation Factor-15 (GDF-15) in psychiatric in-patients: age, renal indices, interleukin-6 (IL-6) and soluble urokinase plasminogen activator receptor (suPAR)

Evangelia Maria Tsapakis a,⁎, Eirini Keskilidou b, Apostolos Karapatis a, Theocharis Kyziridis a, Prokopios Prokopiou a, Paraskevi Karalazou b, Aikaterini Thisiadou b, Kali Makedou b, Konstantinos N Fountoulakis a,c,d
PMCID: PMC13453585  PMID: 42577004

Abstract

Growth differentiation factor-15 (GDF-15) is a stress-responsive cytokine implicated in systemic inflammatory and renal stress biology. It remains unclear whether GDF-15 reflects diagnosis-specific biology or transdiagnostic physiological burden at acute psychiatric admission. In this cross-sectional study, 190 acutely admitted in-patients with schizophrenia spectrum disorders (SSD, n = 126) or mood disorders (MD, n = 64) underwent blood sampling within 12 h of emergency presentation. Serum GDF-15, interleukin-6 (IL-6), high-sensitivity C-reactive protein (hsCRP), and soluble urokinase plasminogen activator receptor (suPAR) were assayed, alongside creatinine, urea, and estimated glomerular filtration rate (eGFR). Group comparisons were conducted using independent-samples t-tests or Mann-Whitney U tests for continuous variables and χ2 or Fisher's exact tests for categorical variables. Following Bonferroni correction across 31 comparisons (p < 1.6 × 10−3), antidepressant burden and previous hospitalizations were the only significant between-group differences. GDF-15 showed a non-significant trend towards higher values in MD than SSD patients. In the primary complete-case multivariable model, diagnostic category was not independently associated with ln (GDF-15) after adjustment for age, urea, ln (IL-6), and ln (suPAR) (MD vs SSD: B = −0.012, β = −0.010, p = 0.900). Instead, ln (GDF-15) was associated with older age (B = 0.013, p = 3.0 × 10−4), higher urea (B = 0.010, p = 0.027), higher ln (IL-6) (B = 0.167, p = 2.4 × 10−4), and higher ln (suPAR) (B = 0.297, p = 0.001), explaining 40.2% of variance. Sensitivity analyses adjusting for smoking, cardiovascular disease, thyroid disorder, sex, and alternative diagnostic classification did not materially alter the absence of a diagnostic-group effect. In conclusion, GDF-15 variation was better explained by transdiagnostic renal-immune a stress phenotype than by categorical diagnosis, supporting its evaluation as a marker of systemic physiological burden rather than a diagnosis-specific biomarker.

Keywords: GDF-15, IL-6, suPAR, Schizophrenia spectrum disorders, Mood disorders, Inflammation

1. Introduction

Severe mental illness is increasingly understood as a systemic condition in which immune, metabolic, and vascular pathways interact with brain function across diagnoses. Acute psychiatric in-patients represent a clinically informative population in which severe psychopathology, medication exposure, and systemic physiological stress commonly converge. People with severe mental illness (SMI) have high rates of physical comorbidity, modifiable cardiometabolic risk factors, and cardiovascular mortality, partly related to behavioral risk factors, healthcare inequalities, and adverse metabolic effects of psychotropic medication (Firth et al., 2019;Goldfarb et al., 2022). Meta-analytic evidence supports that these physical health risks contribute substantially to excess premature mortality in both schizophrenia spectrum disorders (SSD) and mood disorders (MD) (Chan et al., 2022 & 2023) in addition to marked reductions in life expectancy due to chronic medical conditions (Fountoulakis et al., 2024). Circulating biomarkers may, therefore, provide clinically relevant information about biological burden. A systematic evaluation of inflammatory changes across multiple psychiatric disorders supported the possibility of differentiating psychiatric disorders by using inflammatory biomarkers and further claimed that inflammation may have a large enough effect size that these biomarkers could be detected in relatively small sample sizes (Yuan et al., 2019). However, it remains unclear whether many biomarkers map meaningfully onto categorical diagnoses or instead reflect transdiagnostic physiological processes.

Growth differentiation factor 15 (GDF-15), also known as macrophage inhibitory cytokine-1 (MIC1), is a cytokine rising in response to cellular stress and regulated during inflammation, mitochondrial dysfunction, and tissue injury (Iglesias et al., 2023; Jena et al., 2023; Salminen, 2024; Sigvardsen et al., 2025). GDF-15 is overexpressed under a broad range of conditions, including cardiovascular disease, cancer, cachexia syndromes, and insulin sensitivity states (Breit and Tsai, 2025). GDF-15 increases with age and decreasing telomerase activity, is associated with suboptimal recovery following physical illness, high cardiovascular mortality, cancer, and poorer cognition. These features make GDF-15 a plausible candidate for capturing systemic burden in SMI, where inflammatory activation and renal-metabolic vulnerability commonly co-occur.

Evidence for altered GDF-15 in SMI remains limited but a potential transdiagnostic signal has been suggested. In a Swedish SSD out-patient cohort, Kumar et al. (2017) found higher plasma GDF-15 levels in patients compared to controls, and there were no differences in GDF-15 between psychosis diagnostic subgroups after adjustment for age and smoking. Within the psychosis cohort, GDF-15 correlated positively with age and was higher in smoking males. Importantly, in a multivariate regression model, greater psychosis severity was associated with lower GDF-15 levels. There was no significant correlation between GDF-15 and high-sensitivity C-reactive protein (hsCRP), however, suggesting that GDF-15 may capture a process partly distinct from CRP-indexed inflammation. A further study found circulating GDF-15 to be elevated in chronic, clinically stable male SSD in-patients compared to controls. GDF-15 was positively associated with Body Mass Index (BMI) and negatively with cognitive performance after adjusting for age, illness duration, smoking, education, and antipsychotic dose (Guo et al., 2024). More recently, a cross-sectional case-control study from Brazil examined GDF-15 variation transdiagnostically across out-patients with SMI and healthy controls, and showed that GDF-15 was elevated in all psychiatric groups, suggesting that this marker reflects systemic stress load common to SMI across diagnoses rather than representing a signal specific to, for example, schizophrenia, which may add to physical health deterioration over the course of mental disorders. In this study, multivariate models identified diagnostic group and the number of psychiatric hospitalizations as significant main effects on GDF-15 levels. In contrast, age, sex, illness duration, and age of onset were not significant predictors of GDF-15 levels (Costanzi et al., 2025).

Interpretation of GDF-15 in psychiatric cohorts nevertheless requires careful control for biological drivers. GDF-15 covaries strongly with age and renal physiology, and inpatient populations may differ meaningfully on renal indices and general medical burden. GDF-15 seems to behave as a cellular stress cytokine in kidney tissue, and GDF-15 in serum or urine may reflect intrarenal tubular stress and injury. During renal injury, the rise in GDF-15 is often compensatory and cannot fully prevent injury despite being protective in principle. As GFR declines, circulating GDF-15 levels rise, whereas urinary levels may behave differently. Thus, serum and urine GDF-15 may capture different biological processes (retention or systemic versus local tubular production). Overall, it appears that GDF-15 acts as an integrative stress biomarker, tightly intertwined with kidney pathophysiology, tracking risk and severity across renal disorders. However, its pleiotropy means elevations are not kidney-specific and may reflect systemic stress biology in parallel with renal impairment (Delrue et al., 2023).

GDF-15 may, therefore, align more closely with inflammatory activity than with psychiatric diagnosis per se. Indeed, other inflammatory biomarkers may clarify the biological context of GDF-15, including Interleukin-6 (IL-6), hsCRP, and soluble urokinase plasminogen activator receptor (suPAR), each representing a distinct inflammatory dimension. IL-6 is one of the most consistently implicated cytokines in SSD and MD, has been associated with acute inflammatory activation, and has meta-analytic support in SMI (Misiak et al., 2021). hsCRP is a clinically established acute-phase inflammatory marker, although it is strongly influenced by adiposity, infection, and other acute medical factors. Unlike IL-6 and hsCRP, suPAR is less dependent on short-term circadian variation, fasting status, and transient acute phase responses (Reisinger et al., 2021). It has been linked to endothelial dysfunction, immune cell activation, chronic disease risk, accelerated ageing and mortality (Belvederi et al., 2025). suPAR is a relatively stable biomarker of systemic chronic inflammation (SCI), a condition representing a progressive, persistent, health-damaging, low-grade inflammation contributing to immunosenescence and the development and progression of many diseases (Marsland, 2021; Rasmussen et al., 2021).

Given this biological understanding, we briefly report findings from clinical studies highlighting the transdiagnostic correlates of suPAR in psychiatric populations. In a study from Denmark, markedly higher suPAR levels were found in patients with schizophrenia compared to controls. Authors concluded that suPAR may reflect inflammatory burden and possible somatic disease risk in schizophrenia (Nielsen et al., 2015). Furthermore, using a longitudinal twin cohort, Trotta et al. (2021) reported that adolescents with both childhood victimization and psychotic experiences demonstrated the highest suPAR values, suggesting that suPAR may index stress-related inflammatory burden rather than psychosis alone (Trotta et al., 2021). Another study found that suPAR was associated with depressive symptoms in females with schizophrenia, supporting possible sex-dependent immune dysregulation (Bigseth et al., 2021). In the Avon Longitudinal Study of Parents and Children (ALSPAC) cohort, Byrne et al. (2022) identified transdiagnostic inflammatory subgroups, with the higher-inflammation cluster showing marked elevations in soluble Tumor Necrosis Factor (TNF) receptors, suPAR, Cluster of Differentiation-93 (CD93), and Interleukin-2-Receptor-Alpha Chain (IL-2RA) and a poorer role functioning, with more participants not in employment, education, or training (Byrne et al., 2022). In addition, when inflammatory plasma markers, including cytokines, CRP, adhesion molecules, alpha-2-macroglobulin, and suPAR were examined in the same cohort, psychotic disorder was associated with higher IL-6 and suPAR, after adjustment for sex, BMI, smoking, cannabis use, and employment status (Mongan et al., 2023). Lastly, a systematic review and meta-analysis on the role of suPAR in psychiatric disorders concluded that suPAR may index chronic inflammatory dysregulation across psychiatric disorders, requiring longitudinal validation (Murphy et al., 2024).

Taken together, suPAR seems to provide complementary information to IL-6 and hsCRP regarding systemic inflammatory burden relevant to acute psychiatric admission. Given that GDF-15 is also induced by inflammatory and cellular-stress pathways, co-assessment of suPAR and GDF-15 may help contextualise a broader systemic stress phenotype.

This study of acute psychiatric in-patients with SSD and MD aimed to investigate the role of GDF-15 and other complementary biomarkers, including IL-6, hsCRP, and suPAR, across these two diagnoses after accounting for covariates. The hypothesis was that any unadjusted GDF-15 differences between SSD and MD would be attenuated after accounting for transdiagnostic biological determinants, and that higher IL-6, reflecting upstream cytokine signaling, higher hsCRP, highlighting downstream acute-phase response, and higher suPAR, representing sustained innate immune activation, would each independently associate with higher GDF-15.

2. Methods

2.1. Study design, setting, and participants

One hundred and ninety acutely admitted in-patients to the 3rd Academic Department of Psychiatry at AHEPA General University Hospital of Thessaloniki were recruited from September 2024 to August 2025. Eligibility criteria for inclusion in this cross-sectional study were a schizophrenia spectrum disorder (SSD) or a mood disorder (MD) diagnosis according to the structured DSM-5-TR diagnostic interview. SSD and MD were selected because they represented the two major severe mental illness diagnostic groupings treated in the acute in-patient service and allowed assessment of whether GDF-15 variation was diagnosis-specific or transdiagnostic. Diagnoses were subsequently mapped to ICD-10 diagnostic bands for analysis. SSD corresponded to F20-F29, and mood disorders corresponded to F30-F39. Exclusion criteria included pregnancy, malignancy, exacerbation of chronic inflammatory or autoimmune disease, severe renal failure, or use of steroids and/or other immunomodulatory drugs. They were assessed through routine emergency department and psychiatric admission assessment, review of hospital medical records, and patient or collateral history. Pregnancy testing was performed when clinically indicated according to standard hospital practice. Healthy controls were excluded due to significant differences in age and medical comorbidity. Preliminary comparisons (not shown) indicated substantial differences in age and medical comorbidity, which could have introduced major confounding in the interpretation of GDF-15 and inflammatory biomarker concentrations. The study was approved by the Scientific Board of AHEPA General University Hospital of Thessaloniki (Approval No. 9202, 01.03.2024).

2.2. Clinical and anthropometric variables

Age, sex, smoking status, legal status, and psychotropic medication exposure on admission, known somatic illness, and duration of index hospitalization were recorded. Body Mass Index (BMI, kg/m2) and waist circumference (cm) were also available. For patients receiving psychotropic medication on admission, the antipsychotic dose was expressed as olanzapine equivalents (Leucht et al., 2016), the antidepressant dose as fluoxetine equivalents (Hayasaka et al., 2015), and the benzodiazepine dose as diazepam equivalents (Borrelli et al., 2022). For each psychotropic class, binary exposure variables were also derived.

Standardized symptom severity scales, such as PANSS, YMRS, or HAM-D, were not systematically administered as part of the acute admission protocol. Acute clinical status was therefore characterized pragmatically using acute in-patient admission, legal status on admission, psychotropic medication exposure on admission, and duration of the index hospitalization.

2.3. Laboratory measures and sampling

Blood samples were collected on admission in the emergency department and within 12 h of presentation. Samples were immediately sent to the hospital laboratories for analysis, and they were not stored. Serum was used for GDF-15, IL-6, and hsCRP, and plasma was used for suPAR. Renal indices included serum creatinine (mg/dL), eGFR (mL/min/1.73 m2), and urea (mg/dl).

Serum growth differentiation factor-15 (GDF-15, measuring range: 400-20000 pg/mL) concentration was determined by electrochemiluminescence (ECLIA), using a sandwich principle, with the immunological analyzer e801 of the automated system COBAS 8000 (Roche Diagnostics, Mannheim, Germany). Samples for suPAR were collected in K2-EDTA anticoagulant tubes, centrifuged at 3000 x g for 1-10 min to separate plasma, and analysed as soon as possible according to the manufacturer's instructions. Plasma suPAR concentrations (Measuring range: 1.8-16.0 ng/mL) levels were quantified by a particle-enhanced turbidimetric immunoassay (PETIA) (suPARnostic TurbiLatex Reagents, Virogates, Birkerod, Denmark) on the biochemical analyzer c702 of the COBAS 8000 automated platform (Roche Diagnostics, Mannheim, Germany). Interleukin-6 (IL-6; measuring range: 1.5-5000 pg/mL) was quantified by electrochemiluminescence (ECLIA) on the COBAS e 402 immunoassay analyzer, using a sandwich principle. Lastly, high-sensitivity C-reactive protein (hsCRP, normal values < 3 mg/L) was quantified in serum samples using an immunoturbidimetric method on the DxC 700 AU automated analyzer (Beckman Coulter Ireland Inc., Clare, Ireland).

2.4. Statistical analysis

The IBM SPSS Statistics version 29.0 was used for analysis of the raw data. In Table 1, continuous variables are reported as mean (SD) and categorical variables as n/N (%). For continuous variables, between-group comparisons (SSD vs MD) used independent-samples t-tests, or Welch's t-test when homogeneity of variance was violated, whereas for categorical variables, either χ2 tests or Fisher's exact test, where appropriate, were used. Biomarkers were right-skewed, so they were natural log-transformed for the regression analyses (lnGDF15, lnIL-6, lnsuPAR, lnhsCRP). Regression coefficients were interpreted on the log-transformed outcome scale. All tests were two-tailed with α = 0.05. Bonferroni adjustment was applied across the 31 univariable comparisons in Table 1, yielding a corrected significance threshold of p = 0.05/31 = 1.6 × 10−3.

Table 1.

Sample characteristics by diagnosis group.

Variable, N (%) SSD, 126 (66.3) MD, 64 (33.7) p-value
Age in years, mean (SD) [N] 46.36 (14.41) [125] 50.66 (14.17) [64] 0.053
Sex (male), n/N (%) 84/126 (66.7) 31/64 (48.4) 0.019
BMI (kg/m2), mean (SD) [N] 26.57 (5.98) [106] 26.83 (6.76) [57] 0.799
Waist circumference (cm), mean (SD) [N] 90.81 (17.12) [98] 90.17 (18.31) [53] 0.832
Smoking (yes), n/N (%) 69/114 (60.5) 33/60 (55.0) 0.519
Known somatic illness (yes), n/N (%) 49/116 (42.2) 26/60 (43.3) 0.890
Cardiovascular Disease (yes), n/N (%) 19/117 (16.2) 19/60 (31.7) 0.021
Chronic Inflammatory Disease (yes), n/N (%) 8/117 (6.8) 6/61 (9.8) 0.560
Thyroid Disease (yes), n/N (%) 10/115 (8.7) 7/60 (11.7) 0.594
Duration of Psychiatric Illness (years), mean (SD) [N] 20.72 (14.11) [115] 17.95 (13.81) [60] 0.227
Number of previous hospitalizations, mean (SD) [N] 4.7 (6.74) [115] 1.53 (2.17) [60] 2.8x10−4ab
Voluntary admission, n/N (%) 42/125 (33.6) 24/62 (38.7) 0.518
Duration of hospitalization (days), mean (SD) [N] 17.59 (16.03) [126] 13.89 (13.86) [64] 0.118
Medication on admission (yes), n/N (%) 55/125 (44.0) 34/63 (54.0) 0.218
Any antipsychotic on admission, n/N (%) 46/126 (36.5) 25/64 (39.1) 0.753
Antipsychotic dose (olanzapine equivalents), mean (SD) [N] 15.27 (68.10) [126] 3.82 (6.43) [64] 0.381†
Any antidepressant on admission, n/N (%) 19/126 (15.1) 24/64 (37.5) 8.3x10−4b
Antidepressant dose (fluoxetine equivalents), mean (SD) [N] 9.15 (28.52) [126] 19.02 (30.11) [64] 6.6x10−4ab
Any benzodiazepine on admission, n/N (%) 30/126 (23.8) 13/64 (20.3) 0.714
Benzodiazepine dose (diazepam equivalents), mean (SD) [N] 5.31 (12.35) [126] 3.52 (9.37) [64] 0.308
Urea (mg/dl), mean (SD) [N] 28.10 (13.33) [126] 32.34 (12.43) [64] 0.035
Creatinine (mg/dL), mean (SD) [N] 0.89 (0.20) [126] 0.89 (0.18) [64] 0.863
eGFR (mL/min/1.73m2), mean (SD) [N] 103.94 (15.99) [100] 98.42 (17.09) [57] 0.044
GDF-15 (pg/ml), median (IQR) [N] 921.00 (682.00) [122] 1060.00 (874.00) [61] 0.051a
lnGDF-15, median (IQR) [N] 6.83 (0.69) [122] 6.97 (0.74) [61] 0.051a
IL-6 (pg/ml), mean (SD) [N] 12.83 (27.62) [116] 15.40 (34.79) [60] 0.594
lnIL-6, mean (SD) [N] 1.92 (0.92) [116] 0.92 (0.95) [60] 0.375
hsCRP (mg/dl), mean (SD) [N] 1.02 (2.84) [114] 0.93 (1.62) [61] 0.816
lnhsCRP, mean (SD) [N] −1.13 (1.51) [114] −1.08 (1.51) [61] 0.822
suPAR (μg/ml), mean (SD) [N] 3.96 (2.11) [85] 5.70 (8.00) [43] 0.226a
lnsuPAR, mean (SD) [N] 1.28 (0.41) [85] 1.45 (0.62) [43] 0.226a
Open in a new tab

BMI: Body Mass Index; eGFR: estimated glomerular filtration rate; GDF-15: Growth Differentiation Factor-15; hsCRP: high-sensitivity C-reactive protein; IL-6: Interleukin-6; ln: natural logarithm; MD: mood disorders; SD: standard deviation; SSD: schizophrenia spectrum disorders; suPAR: soluble urokinase plasminogen activator receptor.

Values are mean (SD) [N] for continuous variables and n/N (%) for categorical variables. Denominators reflect available data. Medication-equivalent dose variables include zero values for non-exposed patients. Between-group comparisons used independent-samples t-tests for approximately normally distributed continuous variables, Mann-Whitney U tests for markedly skewed or zero-inflated variables, and χ2 or Fisher's exact tests for categorical variables, as appropriate. The Bonferroni-corrected significance threshold was p = 0.05/31 = 1.6x10−3. GDF-15 and ln(GDF-15) showed a non-significant trend toward higher values in MD than SSD patients on Mann-Whitney U testing, but neither comparison survived Bonferroni correction.

a

p-value from Mann-Whitney U test.

b

Survives Bonferroni correction.

2.4.1. Handling of missing data

Missing data were handled using available-case analysis for descriptive and univariable between-group comparisons and complete-case analysis for multivariable regression models. Accordingly, denominators vary across descriptive analyses and are reported for each variable in Table 1. For the primary regression analysis, complete-case status was defined by availability of ln (GDF-15), diagnostic group, age, urea, ln (IL-6), and ln (suPAR). As the complete-case regression sample was smaller than the full analytic cohort, the pattern of missingness was examined descriptively, with particular attention to whether missingness was differentially distributed by diagnostic group. The effective sample size was reported for each regression model, and complete-case status as well as ln (suPAR) availability were compared between diagnostic groups. Little's MCAR test was not reported since missingness was not treated as a single global pattern across the entire dataset; instead, missingness was evaluated in relation to the variables entering the primary regression model and the key diagnostic contrast. This targeted assessment directly addressed the main potential source of bias for the revised analyses.

2.4.2. Regression analysis

Multivariable linear regression was used to examine independent predictors of ln (GDF-15). The primary model was specified a priori to test whether diagnostic category was independently associated with ln (GDF-15) after adjustment for age, renal/metabolic status, and inflammatory markers. The primary model included diagnostic group, age, urea, ln (IL-6), and ln (suPAR). Diagnostic group was coded as SSD = 0 and MD = 1 using the strict ICD-10 F20-F29 versus F30-F39 classification. Urea was retained as the renal/metabolic predictor in the primary model since renal indices were clinically related and inclusion of multiple renal indices could introduce redundancy. No automated stepwise, forward, or backward variable-selection procedures were used. Smoking was examined in a sensitivity model because previous work has linked smoking with GDF-15. Additional sensitivity models adjusted for cardiovascular disease, thyroid disorder, and sex. An alternative diagnostic model grouped F25 schizoaffective disorder with F30-F39 (MD), and a fully adjusted exploratory model included diagnosis, age, urea, ln (IL-6), ln (suPAR), sex, smoking, cardiovascular disease, and thyroid disorder. Regression results were reported as unstandardized coefficients, standard errors, standardized β coefficients, 95% confidence intervals, and p-values (Table 2).

Table 2.

Multivariable linear regression and sensitivity models predicting ln (GDF-15).

Model Predictor B SE β 95% CI for B p-value
1 Primary Model
N = 110 adj.R2 = 0.402
F (5,104) = 15.642 p = 1.85x10−11
 Diagnosis: MD vs SSD −0.012 0.097 −0.010 −0.204 to 0.180 0.900
Age 0.013 0.003 0.314 0.006 to 0.020 3.0x10−4
Urea 0.010 0.004 0.185 0.001 to 0.018 0.027
ln(IL-6) 0.167 0.044 0.293 0.080 to 0.253 2.4x10−4
ln(suPAR)
 0.297
 0.090
 0.254
 0.120 to 0.475
 0.001
2 Primary Model + Smoking
N = 98 adj.R2 = 0.386
F (6,91) = 11.156 p = 2.68x10−9
 Diagnosis: MD vs SSD 0.017 0.107 0.014 −0.196 to 0.230 0.874
Age 0.013 0.004 0.320 0.006 to 0.021 6.2x10−4
Urea 0.012 0.005 0.214 0.002 to 0.022 0.015
ln(IL-6) 0.159 0.046 0.285 0.068 to 0.251 8.6x10−4
ln(suPAR) 0.258 0.097 0.221 0.065 to 0.452 0.009
Smoking
 0.072
 0.095
 0.062
 −0.117 to 0.262
 0.450
3 Primary Model + Cardiovascular Disease
N = 102 adj.R2 = 0.396
F (6,95) = 12.017 p = 5.29x10−10
 Diagnosis: MD vs SSD −0.007 0.103 −0.006 −0.212 to 0.198 0.943
Age 0.012 0.004 0.300 0.005 to 0.020 0.001
Urea 0.010 0.005 0.183 0.001 to 0.020 0.038
ln(IL-6) 0.151 0.045 0.268 0.061 to 0.241 0.001
ln(suPAR) 0.294 0.093 0.252 0.110 to 0.478 0.002
Cardiovascular Disease
 0.127
 0.112
 0.097
 −0.096 to 0.349
 0.261
4 Primary Model + Thyroid Disorder
N = 101 adj.R2 = 0.399
F (6,94) = 12.063 p = 5.21x10−10
 Diagnosis: MD vs SSD 0.005 0.104 0.004 −0.201 to 0.211 0.961
Age 0.013 0.004 0.318 0.006 to 0.020 4.2x10−4
Urea 0.010 0.005 0.176 0.000 to 0.020 0.047
ln(IL-6) 0.154 0.045 0.274 0.064 to 0.244 9.8x10−4
ln(suPAR) 0.292 0.093 0.251 0.108 to 0.477 0.002
Thyroid Disorder
 0.199
 0.151
 0.107
 −0.101 to 0.499
 0.192
5 Primary Model + Male Sex
N = 110 adj.R2 = 0.422
F (6,103) = 14.264 p = 8.99x10−12
 Diagnosis: MD vs SSD 0.047 0.099 0.039 −0.150 to 0.243 0.638
Age 0.016 0.004 0.394 0.009 to 0.024 3.2x10−5
Urea 0.008 0.004 0.154 −0.001 to 0.016 0.066
ln(IL-6) 0.155 0.043 0.273 0.069 to 0.241 5.3x10−4
ln(suPAR) 0.334 0.090 0.285 0.156 to 0.512 3.3x10−4
Male sex
 0.216
 0.100
 0.187
 0.017 to 0.416
 0.034
6 Alternative Diagnosis Model
N = 110 adj.R2 = 0.402
F (5,104) = 15.653 p = 1.82x10−11
 Alternative Diagnosis: F25 + F30-F39 vs non-affective F20-F29 0.020 0.095 0.017 −0.167 to 0.208 0.829
Age 0.013 0.003 0.308 0.006 to 0.020 4.4x10−4
Urea 0.009 0.004 0.179 0.001 to 0.018 0.033
ln(IL-6) 0.167 0.044 0.293 0.080 to 0.253 2.3x10−4
ln(suPAR)
 0.294
 0.089
 0.251
 0.117 to 0.471
 0.001
7 Fully-adjusted Exploratory Model
N = 98 adj.R2 = 0.401
F (9,88) = 8.212 p = 9.12x10−9 		Diagnosis: MD vs SSD 0.064 0.109 0.052 −0.152 to 0.280 0.560
Age 0.015 0.004 0.365 0.007 to 0.023 5.5x10−4
Urea 0.007 0.005 0.130 −0.003 to 0.018 0.166
ln(IL-6) 0.144 0.046 0.258 0.052 to 0.236 0.002
ln(suPAR) 0.310 0.100 0.264 0.110 to 0.509 0.003
Male sex 0.192 0.117 0.163 −0.042 to 0.425 0.106
Smoking 0.027 0.097 0.023 −0.165 to 0.218 0.784
Cardiovascular disease 0.108 0.116 0.083 −0.122 to 0.339 0.352
Thyroid disorder 0.239 0.165 0.124 −0.088 to 0.566 0.150
Open in a new tab

B: unstandardized regression coefficient; β: standardized coefficient. CI: confidence interval; CVD: cardiovascular disease; GDF-15: Growth Differentiation Factor-15; ICD-10: International Classification of Diseases, Tenth Revision; IL-6: interleukin-6; MD: mood disorders; SE: standard error; SSD: schizophrenia spectrum disorders; suPAR: soluble urokinase plasminogen activator receptor.

Dependent variable: ln(GDF-15). Model 1 included diagnosis, age, urea, ln(IL-6), and ln(suPAR). Models 2, 3, 4, and 5 additionally adjusted for smoking, cardiovascular disease, thyroid disorder, and sex, respectively. Model 6 used an alternative diagnostic classification in which F25 schizoaffective disorder was grouped with F30-F39/MD. Model 7 included diagnosis, age, urea, ln(IL-6), ln(suPAR), sex, smoking, cardiovascular disease, and thyroid disorder. In Models 1–5 and 7, diagnostic group was coded as SSD = 0 and MD = 1 using strict ICD-10 F20-F29 versus F30–F39 classification. In Model 6, alternative diagnosis was coded as non-affective F20-F29 = 0 and F25 + F30-F39 = 1.

Individual-level data were visualized by plotting ln (GDF-15) by diagnostic group and against age, urea, ln (IL-6), and ln (suPAR), with fitted linear trends shown for continuous predictors (Fig. 1). Adjusted associations from the primary complete-case multivariable model were additionally visualized using a forest plot of unstandardized regression coefficients and 95% confidence intervals (Fig. 2).

Fig. 1.

Fig. 1

Open in a new tab

Distribution and correlates of ln(GDF-15). Individual observations are shown for participants with available data. (A) ln (GDF-15) by diagnostic group using strict ICD-10 classification, with SSD defined as F20–F29 and MD as F30–F39. (B) Association between age and ln (GDF-15). (C) Association between urea and ln (GDF-15). (D) Association between ln (IL-6) and ln (GDF-15). (E) Association between ln (suPAR) and ln (GDF-15). Individual observations in panels B–E are coloured by diagnostic group, and black lines represent fitted linear trends across the full sample.

GDF-15: Growth Differentiation Factor-15; ICD-10: International Classification of Diseases, Tenth Revision; IL-6: interleukin-6; MD: mood disorders; SSD: schizophrenia spectrum disorders; suPAR: soluble urokinase plasminogen activator receptor.

Fig. 2.

Fig. 2

Open in a new tab

Multivariable predictors of ln(GDF-15) in acutely admitted psychiatric in-patients. Forest plot showing unstandardized regression coefficients and 95% confidence intervals from the primary complete-case multivariable linear regression model predicting ln (GDF-15). Predictors included diagnostic category, age, urea, ln (IL-6), and ln (suPAR). Diagnostic category was coded as MD versus SSD. Positive coefficients indicate higher ln (GDF-15) values.

GDF-15: Growth Differentiation Factor-15; IL-6: interleukin-6; MD: mood disorders; SSD: schizophrenia spectrum disorders; suPAR: soluble urokinase plasminogen activator receptor.

3. Results

3.1. Sample characteristics

The analytic cohort included N = 190 psychiatric in-patients, comprising 126 (66.3%) SSD and 64 (33.7%) MD patients (Table 1). This unequal distribution reflected the natural diagnostic composition of acute in-patient admissions during the recruitment period rather than a matched sampling design. The MD group was older [50.66 ± 14.17 vs 46.36 ± 14.41 years, t (187) = −1.950, p = 0.053, mean difference = −4.296 (95% CI: −8.642 to 0.049), d = −0.300] and had a lower proportion of males (33.7% vs 66.3%, p = 0.019). Smoking prevalence did not differ between groups (p = 0.519). Legal status on admission (p = 0.518), medication on admission (p = 0.218), and duration of hospitalization (p = 0.118) did not differ significantly between the SSD and MD groups. Known somatic illness difference was also non-significant between the two groups, but (p = 0.890), but unadjusted cardiovascular disease prevalence was more frequent among MD compared to SSD patients [19/60 (31.7%) vs 19/117 (16.2%), p = 0.021]. The number of previous hospitalizations was higher in the SSD group (4.70 ± 6.74 vs 1.53 ± 2.17, p = 2.8 × 10−4), surviving Bonferroni correction. Group characteristics are shown in Table 1.

3.2. Renal and metabolic indices

Estimated glomerular filtration rate (eGFR, in ml/min/1.73 m2) was nominally lower in the MD group compared with SSD [98.42 ± 17.09 vs 103.94 ± 15.99, t (148) = 2.028, p = 0.044, mean difference = 5.519 (95% CI: 0.143 to 10.895), d = 0.337]. Serum creatinine (mg/dl) did not differ significantly (p = 0.863). Urea (mg/dl) was higher in MD patients [32.34 ± 12.43 vs 28.10 ± 13.33, t (188) = −2.123, p = 0.035, mean difference = −4.249 (95% CI: −8.196 to −0.301), d = −0.326]. None of these renal, metabolic, or anthropometric differences survived Bonferroni correction. BMI (kg/m2) and waist circumference (cm) were comparable between groups (p = 0.799 and p = 0.832, respectively).

3.3. GDF-15 and inflammatory biomarkers

The non-parametric Mann-Whitney U test was used to test for MD vs. SSD between-group differences in GDF-15, as its distribution was skewed. Median (IQR) GDF-15 (pg/ml) tended to be higher in the MD [1060.00 (874.00)] than the SSD [921.00 (682.00)] group [mean difference = −595.73 (95% CI: −1248.90 to 57.44)], but this difference did not quite reach conventional nominal statistical significance (U = 4381.50, Z = −1.955, p = 0.051). The logarithmic value lnGDF15 showed the same pattern and differed only marginally between the two diagnostic groups with the MD group showing higher values [median (IQR) = 6.97 (0.74) for MD vs 6.83 (0.69) for SSD, mean difference = −0.21 (95% CI: −0.41 to −0.01), U = 4381.50, Z = 1.955, p = 0.051]. IL-6, suPAR, and hsCRP did not differ significantly between diagnostic groups on either the raw or log-transformed scale (Table 1).

3.4. Medication exposure

Any antidepressant exposure (dose in fluoxetine equivalents >0; Hayasaka et al., 2015) was more frequent in MD patients, and this difference survived Bonferroni correction [24/64 (37.5%) vs 19/126 (15.1%), p = 8.3 × 10−4]. Any antipsychotic (dose in olanzapine equivalents >0; Leucht et al., 2016) and benzodiazepine exposure (dose in diazepam equivalents >0; Borrelli et al., 2022) did not differ significantly between groups (p = 0.753 and p = 0.714, respectively). Dose-equivalent distributions, particularly olanzapine equivalents, were markedly skewed, reflecting a large proportion of zero values and a small number of high-dose observations. Antidepressant dose in fluoxetine equivalents was significantly higher in the MD group [19.02 ± 30.11 vs 9.15 ± 28.52, Mann-Whitney U = 4926.00, Z = 3.406, p = 6.6 × 10−4], surviving Bonferroni correction, but antipsychotic and benzodiazepine doses were similar between the two diagnostic groups. Antipsychotic dose in olanzapine equivalents and benzodiazepine dose in diazepam equivalents also did not differ significantly between SSD and MD patients (Table 1).

In summary, after Bonferroni correction for 31 comparisons (p = 0.05/31 = 1.6 × 10−3), only the number of previous hospitalizations, antidepressant exposure on admission, and antidepressant dose-equivalent burden remained significantly different between diagnostic groups.

3.5. Regression analyses

Multivariable regression analyses were performed using complete-case analysis. The primary regression model included 110 of 190 participants, with missingness driven mainly by incomplete availability of suPAR measurements. Complete-case status was not associated with diagnostic group: 73/126 SSD participants and 37/64 MD participants were included in the primary model, corresponding to nearly identical complete-case proportions of 57.9% and 57.8%, respectively. Similarly, ln (suPAR) availability was comparable between diagnostic groups, being available in 85/126 (67.5%) SSD participants and 43/64 (67.2%) MD participants. These findings indicate that the reduction in sample size for multivariable regression reflected biomarker availability, particularly suPAR availability, rather than differential missingness by diagnostic group.

The primary complete-case multivariable linear regression model predicting ln (GDF-15) explained 40.2% of the variance in ln (GDF-15) [adjusted R2 = 0.402, F (5,104) = 15.642, p = 1.85 × 10−11], as shown in Table 2. Diagnostic category was not independently associated with ln (GDF-15) after adjustment for age, urea, ln (IL-6), and ln (suPAR) (MD vs SSD: B = −0.012, SE = 0.097, β = −0.010, 95% CI -0.204 to 0.180, p = 0.900). In contrast, ln (GDF-15) was independently associated with older age (B = 0.013, SE = 0.003, β = 0.314, 95% CI 0.006 to 0.020, p = 3.0 × 10−4), higher urea (B = 0.010, SE = 0.004, β = 0.185, 95% CI 0.001 to 0.018, p = 0.027), higher ln (IL-6) (B = 0.167, SE = 0.044, β = 0.293, 95% CI 0.080 to 0.253, p = 2.4 × 10−4), and higher ln (suPAR) (B = 0.297, SE = 0.090, β = 0.254, 95% CI 0.120 to 0.475, p = 0.001).

To visualize the individual-level distribution and correlates of ln (GDF-15), we plotted ln (GDF-15) by diagnostic group and against the continuous predictors included in the primary regression model (Fig. 1). The diagnostic-group plot showed substantial overlap between SSD and MD participants, consistent with the absence of an independent diagnostic-group association in the multivariable model. In contrast, scatterplots showed positive linear trends between ln (GDF-15) and age, urea, ln (IL-6), and ln (suPAR). These visual patterns were consistent with the regression findings, in which age, urea, ln (IL-6), and ln (suPAR), but not diagnostic category, were associated with ln (GDF-15). We additionally present a forest plot of regression coefficients from the primary multivariable model (Fig. 2), summarising the adjusted associations between the principal predictors and ln (GDF-15). This figure visually highlights the central finding that age, urea, ln (IL-6), and ln (suPAR), but not diagnostic category, were independently associated with ln (GDF-15).

Sensitivity analyses yielded substantively similar findings. Diagnostic category remained non-significant after additional adjustment for smoking (Model 2), cardiovascular disease (Model 3), thyroid disorder (Model 4), and sex (Model 5), as well as in the alternative diagnostic model in which F25 schizoaffective disorder was grouped with F3x/mood disorders (Model 6) and in the fully adjusted exploratory model (Model 7; p ≥ 0.560 for diagnostic-group terms). Across models, age, ln (IL-6), and ln (suPAR) remained the most consistent independent predictors of ln (GDF-15). Interestingly, urea showed a nominal association in the primary model and several sensitivity models but was attenuated after adjustment for sex and in the fully adjusted exploratory model.

Regression diagnostics did not indicate problematic multicollinearity, with variance inflation factors below conventional thresholds across models. In the primary model, the maximum Cook's distance was 0.270, suggesting that no single observation exerted excessive influence on the model estimates.

4. Discussion

In this cross-sectional naturalistic cohort of 190 acutely admitted psychiatric in-patients, the most statistically robust differences between SSD and MD diagnostic groups were clinical rather than biomarker-based. Specifically, after correction for multiple testing, antidepressant use on admission was more frequent in MD than SSD, and MD patients had fewer prior admissions than SSD patients. In contrast, the primary biological signal emerged from multivariable modelling, as ln (GDF-15) was bestt explained by transdiagnostic physiological correlates, namely age, urea, IL-6, and suPAR. At the same time, the diagnosis did not add explanatory value once these covariates were considered.

The expanded regression results in Table 2 clarify that the absence of a diagnostic-group effect was not dependent on a single model specification. In the primary complete-case model, diagnostic category was not associated with ln (GDF-15), whereas age, urea, ln (IL-6), and ln (suPAR) were independently associated with higher ln (GDF-15). The same pattern was largely retained across sensitivity models additionally adjusting for smoking, cardiovascular disease, thyroid disorder, and sex, as well as in the alternative diagnostic model grouping F25 schizoaffective disorder with F30-F39/MD. In the fully adjusted exploratory model, diagnostic category, smoking, cardiovascular disease, thyroid disorder, and sex were not independently associated with ln (GDF-15), while age, ln (IL-6), and ln (suPAR) remained significant predictors. These findings support the interpretation that GDF-15 in this acute in-patient cohort was more strongly related to age and renal-immune physiology than to broad psychiatric diagnostic category.

MD patients were more likely to be receiving antidepressants on admission, as expected for prescribing patterns in MD, supporting the face validity of diagnostic ascertainment in routine clinical care. Group separation in the number of previous hospitalizations suggests that, within this acute in-patient service, SSD admissions reflect a recurrent in-patient course. Interestingly, both reliable between-group contrasts characterize treatment exposure and illness-course intensity rather than supporting diagnosis-specific immune biology. Further, differences in GDF-15, renal function variables, or inflammatory biomarkers between the two diagnostic groups should be interpreted as hypothesis-generating only, as they did not survive correction for multiple testing. This study failed to provide strong evidence that categorical diagnosis per se reflects a reproducible separation in circulating inflammatory biomarkers at admission in this cohort of in-patients with SMI. This transdiagnostic interpretation is also consistent with broader evidence that conventional diagnostic categories do not always map cleanly onto symptom dimensions or biological processes in severe mental illness (Tsapakis et al., 2025).

The associations between GDF-15, IL-6, and suPAR present interesting insights from both clinical and biological perspectives. Regression analysis demonstrated that, across SSD and MD, higher GDF-15 was predicted by older age, elevated urea levels, and increased IL-6 and suPAR levels. Thus, GDF-15 seems to behave as an integrative marker reflecting renal physiology and immune activation rather than diagnosis. IL-6 is a pro-inflammatory cytokine closely linked to acute-phase signaling and systemic inflammatory activation (Aliyu et al., 2022), while suPAR is interpreted as a marker of sustained innate immune and myeloid activation (Belvederi et al., 2025). Their independent contributions to GDF-15 levels suggest that GDF-15 integrates multiple dimensions of inflammatory stress biology. In contrast, hsCRP did not emerge as a key correlate once IL-6 and suPAR were modeled, consistent with the possibility that CRP may be less specific to the biological processes underlying GDF-15 in acute psychiatric presentations (Roy et al., 2025). These findings suggest that acute psychiatric admissions may be accompanied by a shared systemic physiological burden, including medical, metabolic, renal, and inflammatory stressors, whose biomarker signatures cut across conventional diagnostic boundaries.

Previous research has suggested that GDF-15 is elevated in stable schizophrenia in-patients (Guo et al., 2024) and out-patients (Kumar et al., 2017) compared to controls but shows limited diagnostic specificity between diagnostic subgroups of psychosis and different psychiatric diagnoses (Costanzi et al., 2025). Furthermore, Kumar et al. (2017) reported no GDF-15 correlation with hsCRP, consistent with our observation that neither diagnosis nor hsCRP explained ln (GDF-15) once physiological correlates were included. It is likely that CRP marks a distinct inflammatory process from the GDF-15-captured process as it is not central in multivariable interpretation.

Finally, our results converge with Costanzi et al. (2025) on a transdiagnostic framing, but not on the course of illness characteristics. Costanzi et al. (2025) studied a Brazilian outpatient cohort of stable patients with schizophrenia, bipolar disorder, and major depressive disorder, plus healthy controls, and measured serum GDF-15 using sandwich ELISA after sample storage at −80°C. Their models included diagnosis, age, sex, age at illness onset, illness duration, and number of psychiatric hospitalizations, and they identified diagnosis and number of psychiatric hospitalizations as significant effects on GDF-15. They also linked GDF-15 to neuroprogression and somatoprogression. In contrast, our study recruited acutely admitted Greek psychiatric in-patients with SSD or MD, used ECLIA on a Roche platform in non-stored samples, and included renal and inflammatory covariates, particularly urea, IL-6, and suPAR. It could be argued, however, that the two studies are complementary rather than contradictory. GDF-15 may index broader somatic and neuroprogressive burden in stable cohorts, while in acute admission it appears more closely related to renal-immune stress physiology than to diagnosis.

The acute admission context is therefore central to interpretation. Early sampling within 12 h of presentation standardized the timing of biomarker assessment relative to admission and limited variability introduced by delayed inpatient sampling or treatment exposure. At the same time, this window captures a biologically dynamic state. Agitation, sleep deprivation, reduced oral intake, dehydration, HPA-axis and sympathetic activation, intercurrent infection, and acute medication exposure could influence GDF-15, IL-6, suPAR, and renal indices simultaneously. Accordingly, the observed associations should not be interpreted as evidence of a single causal pathway, but rather as a cross-sectional renal-immune stress profile at acute psychiatric admission. More specifically, GDF-15 is a cellular and mitochondrial stress-responsive cytokine and has been associated with acute hospitalization risk and changes in CRP and renal, hepatic, and cardiac biomarkers (Tavenier et al., 2021; Huang et al., 2025). IL-6 is known to increase after acute psychosocial stress (Marsland et al., 2017). In contrast, suPAR appears more stable than classical acute-phase markers and may be less acutely labile than IL-6 or hsCRP, but elevated suPAR may still reflect underlying immune activation or reduced physiological reserve (Rasmussen et al., 2021; Belvederi et al., 2025). Hence, we suggest that the observed GDF-15 - IL-6 - suPAR associations could be interpreted as evidence of a renal-immune stress phenotype at admission, not as proof of a causal pathway or stable trait-like biomarker profile.

Taken together, our results highlight the need for prioritizing detailed medical assessment for psychiatric in-patients with increased GDF-15 levels on admission, rather than treating GDF-15 as a diagnosis-specific biomarker. Outcome-linked longitudinal studies are, however, required to examine the diagnostic utility and actionable thresholds for GDF-15.

4.1. Strengths and limitations

Strengths of this study include the pragmatic recruitment of an acutely admitted psychiatric in-patient cohort, standardized early biomarker sampling, and concurrent assessment of inflammatory, renal, biochemical, clinical, medication, and demographic variables. Sampling within 12 h of emergency presentation standardized the temporal relationship between clinical presentation and biomarker measurement, reduced heterogeneity from delayed sampling or variable exposure to in-patient treatment, and captured the early biological state accompanying acute psychiatric admission. This timing is particularly relevant for stress-responsive and inflammation-related biomarkers such as GDF-15, IL-6, and suPAR. The inclusion of renal indices and inflammatory markers further enabled multivariable modelling of GDF-15 as part of a broader systemic physiological profile rather than as a diagnosis-specific marker alone.

Several limitations should be considered. First, the cross-sectional design precludes causal inference and does not allow us to determine whether higher GDF-15 reflects stable physiological vulnerability, acute-state effects, or both. Although sampling within 12 h of admission standardized the timing of biomarker assessment, acute admission is a biologically dynamic state in which agitation, sleep loss, dehydration, intercurrent infection, stress-related HPA-axis and sympathetic activation, and emergency medication exposure may influence GDF-15, IL-6, suPAR, and renal indices simultaneously. Second, we did not include a healthy control group or a stable severe mental illness outpatient group; therefore, we cannot determine whether GDF-15 was elevated relative to general population norms or whether the observed renal–immune profile persists after clinical stabilization. Third, standardized symptom severity scales were not systematically administered, precluding analyses of whether GDF-15 or inflammatory markers were associated with acute psychopathology severity independently of diagnosis. Fourth, the naturalistic SSD and MD group sizes were unequal, reflecting the admission profile of the service, and this may have reduced power to detect smaller between-diagnosis differences. Finally, multivariable regression analyses were based on complete cases, resulting in a smaller effective sample size than the full analytic cohort, primarily because of incomplete suPAR availability. However, complete-case status and suPAR availability were not differentially distributed by diagnostic group, making it unlikely that missingness explains the absence of an independent diagnostic-group association with ln (GDF-15). The regression findings should nevertheless be interpreted as complete-case estimates.

5. Conclusion

In acutely admitted psychiatric in-patients with SSD and MD, the clearest diagnostic contrasts were clinical rather than biomarker-based, involving antidepressant exposure and previous hospitalizations. In contrast, ln (GDF-15) was better explained by age, urea, ln (IL-6), and ln (suPAR) than by categorical diagnosis. These findings are hypothesis-generating and suggest that GDF-15 may help characterize a transdiagnostic renal–immune stress phenotype at acute psychiatric admission, rather than serving as a diagnosis-specific biomarker. Clinically, elevated GDF-15 in this context may indicate the need for closer assessment of systemic physiological burden, including renal function, inflammatory activation, cardiometabolic risk, infection, hydration status, and other medical comorbidities. Longitudinal studies with repeated sampling after clinical stabilization, healthy and stable psychiatric comparison groups, standardized symptom severity measures, and outcome-linked follow-up are needed to determine whether GDF-15 has prognostic utility or clinically actionable thresholds in psychiatric in-patient care.

Funding sources

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

CRediT authorship contribution statement

Evangelia Maria Tsapakis: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Supervision, Writing – original draft, Writing – review & editing. Eirini Keskilidou: Data curation, Methodology, Writing – review & editing. Apostolos Karapatis: Data curation, Project administration, Writing – review & editing. Theocharis Kyziridis: Investigation, Resources, Writing – review & editing. Prokopios Prokopiou: Investigation, Resources, Writing – review & editing. Paraskevi Karalazou: Methodology, Writing – review & editing. Aikaterini Thisiadou: Investigation, Methodology, Writing – review & editing. Kali Makedou: Conceptualization, Investigation, Methodology, Resources, Writing – review & editing. Konstantinos N. Fountoulakis: Conceptualization, Investigation, Methodology, Resources, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

Data will be made available on request.

References

  1. Aliyu M., Zohora F.T., Anka A.U., Ali K., Maleknia S., Saffarioun M., Azizi G. Interleukin-6 cytokine: an overview of the immune regulation, immune dysregulation, and therapeutic approach. Int. Immunopharmacol. 2022;111 doi: 10.1016/j.intimp.2022.109130. [DOI] [PubMed] [Google Scholar]
  2. Belvederi F., Leggeri S., Urbani A., Baroni S. suPAR as a biomarker of support in different clinical settings. Clinica chimica acta; international journal of clinical chemistry. 2025;573 doi: 10.1016/j.cca.2025.120303. [DOI] [PubMed] [Google Scholar]
  3. Bigseth T.T., Engh J.A., Egeland J., Andersen E., Andreassen O.A., Bang-Kittilsen G., Falk R.S., Holmen T.L., Lindberg M., Mordal J., Nielsen J., Steen N.E., Ueland T., Vang T., Fredriksen M. Exploring low grade inflammation by soluble urokinase plasminogen activator receptor levels in schizophrenia: a sex-dependent association with depressive symptoms. BMC Psychiatry. 2021;21(1):527. doi: 10.1186/s12888-021-03522-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Borrelli E.P., Bratberg J., Hallowell B.D., Greaney M.L., Kogut S.J. Application of a diazepam milligram equivalency algorithm to assess benzodiazepine dose intensity in Rhode Island in 2018. Journal of managed care & specialty pharmacy. 2022;28(1):58–68. doi: 10.18553/jmcp.2022.28.1.58. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Breit S.N., Tsai V.W. Metabolic messenger: growth differentiation factor 15. Nat. Metab. 2025;7(9):1732–1744. doi: 10.1038/s42255-025-01353-3. [DOI] [PubMed] [Google Scholar]
  6. Byrne J.F., Healy C., Mongan D., Susai S.R., Zammit S., Fӧcking M., Cannon M., Cotter D.R. Transdiagnostic inflammatory subgroups among psychiatric disorders and their relevance to role functioning: a nested case-control study of the ALSPAC cohort. Transl. Psychiatry. 2022;12(1):377. doi: 10.1038/s41398-022-02142-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Chan J.K.N., Tong C.H.Y., Wong C.S.M., Chen E.Y.H., Chang W.C. Life expectancy and years of potential life lost in bipolar disorder: systematic review and meta-analysis. Br. J. Psychiatr. : J. Ment. Sci. 2022;221(3):567–576. doi: 10.1192/bjp.2022.19. [DOI] [PubMed] [Google Scholar]
  8. Costanzi M., Goularte J.F., da Silva L.H.T., Siebert M., de Carvalho I.D.M., Ziani P.R., Deckmann I., Santos-Terra J., Gottfried C., Fleck M.P., Caldieraro M.A., Rosa A.R., Gama C.S. The effect of the number of psychiatric hospitalizations and diagnosis in serum levels of CCL11 and GDF15 in individuals with stable schizophrenia, bipolar disorder, and major depressive disorder: the compass between neuro and somatoprogression. Progress in neuro-psychopharmacology & biological psychiatry. 2025;141 doi: 10.1016/j.pnpbp.2025.111446. [DOI] [PubMed] [Google Scholar]
  9. Delrue C., Speeckaert R., Delanghe J.R., Speeckaert M.M. Growth differentiation factor 15 (GDF-15) in kidney diseases. Adv. Clin. Chem. 2023;114:1–46. doi: 10.1016/bs.acc.2023.02.003. [DOI] [PubMed] [Google Scholar]
  10. Firth Joseph, Siddiqi Najma, Koyanagi Ai, Siskind Dan, Rosenbaum Simon, Galletly Cherrie, Allan Stephanie, Caneo Constanza, Carney Rebekah, Carvalho Andre F, Chatterton Mary Lou, Correll Christoph U, Curtis Jackie, Gaughran Fiona, Heald Adrian, Hoare Erin, Jackson Sarah E, Kisely Steve, Lovell Karina, Maj Mario, McGorry Patrick D, Mihalopoulos Cathrine, Myles Hannah, O'Donoghue Brian, Pillinger Toby, Sarris Jerome, Schuch Felipe B, Shiers David, Smith Lee, Solmi Marco, Suetani Shuichi, Taylor Johanna, Teasdale Scott B, Thornicroft Graham, Torous John, Usherwood Tim, Vancampfort Davy, Veronese Nicola, Ward Philip B, Yung Alison R, Killackey Eoin, Stubbs Brendon. The Lancet Psychiatry Commission: a blueprint for protecting physical health in people with mental illness. The Lancet Psychiatry. 2019;6(8):675–712. doi: 10.1016/s2215-0366(19)30132-4. https://linkinghub.elsevier.com/retrieve/pii/S2215036619301324 [DOI] [PubMed] [Google Scholar]
  11. Fountoulakis K.N., Karakatsoulis G.N., Abraham S., Adorjan K., Ahmed H.U., Alarcón R.D., Arai K., Auwal S.S., Berk M., Bjedov S., Bobes J., Bobes-Bascaran T., Bourgin-Duchesnay J., Bredicean C.A., Bukelskis L., Burkadze A., Cabrera Abud I.I., Castilla-Puentes R., Cetkovich M., Colon-Rivera H., et al. Somatic multicomorbidity and disability in patients with psychiatric disorders in comparison to the general population: a quasi-epidemiological investigation in 54,826 subjects from 40 countries (COMET-G study) CNS Spectr. 2024;29(2):126–149. doi: 10.1017/S1092852924000026. [DOI] [PubMed] [Google Scholar]
  12. Goldfarb Michael, De Hert Marc, Detraux Johan, Di Palo Katherine, Munir Haroon, Music Sanela, Piña Ileana, Ringen Petter Andreas. Severe Mental Illness and Cardiovascular Disease. Journal of the American College of Cardiology. 2022;80(9):918–933. doi: 10.1016/j.jacc.2022.06.017. https://linkinghub.elsevier.com/retrieve/pii/S0735109722054213 [DOI] [PubMed] [Google Scholar]
  13. Guo T., Chen L., Sun W., Yang H., Li J., Zhang X., Chen P. Increased GDF-15 in chronic male patients with schizophrenia: correlation with body mass index and cognitive impairment. Schizophrenia (Heidelberg, Germany) 2024;10(1):117. doi: 10.1038/s41537-024-00541-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Hayasaka Y., Purgato M., Magni L.R., Ogawa Y., Takeshima N., Cipriani A., Barbui C., Leucht S., Furukawa T.A. Dose equivalents of antidepressants: Evidence-based recommendations from randomized controlled trials. J. Affect. Disord. 2015;180:179–184. doi: 10.1016/j.jad.2015.03.021. [DOI] [PubMed] [Google Scholar]
  15. Huang Q., Monzel A.S., Rausser S., Haahr R., Indik C.E., Savin M.J., Bobba-Alves N., Liu C.C., Devine J., Thompson E., Kelly C., Kurade M., Michelson J., Li S., Engelstad K., Lauriola V., Marsland A.L., Kaufman B.A., St-Onge M.P., Sloan R., et al. The energetic stress cytokine GDF15 is elevated in the context of chronic and acute psychosocial stress. bioRxiv : the preprint server for biology. 2025 doi: 10.1101/2024.04.19.590241. 2024.04.19. [DOI] [Google Scholar]
  16. Iglesias P., Silvestre R.A., Díez J.J. Growth differentiation factor 15 (GDF-15) in endocrinology. Endocrine. 2023;81(3):419–431. doi: 10.1007/s12020-023-03377-9. [DOI] [PubMed] [Google Scholar]
  17. Kumar P., Millischer V., Villaescusa J.C., Nilsson I.A.K., Östenson C.G., Schalling M., Ösby U., Lavebratt C. Plasma GDF15 level is elevated in psychosis and inversely correlated with severity. Sci. Rep. 2017;7(1):7906. doi: 10.1038/s41598-017-07503-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Leucht S., Samara M., Heres S., Davis J.M. Dose equivalents for antipsychotic drugs: the DDD method. Schizophr. Bull. 2016;42(Suppl. 1):S90–S94. doi: 10.1093/schbul/sbv167. Suppl 1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Marsland A.L. suPAR: a newer biomarker of systemic chronic inflammation. Brain Behav. Immun. 2021;98:263–264. doi: 10.1016/j.bbi.2021.08.224. [DOI] [PubMed] [Google Scholar]
  20. Marsland A.L., Walsh C., Lockwood K., John-Henderson N.A. The effects of acute psychological stress on circulating and stimulated inflammatory markers: a systematic review and meta-analysis. Brain Behav. Immun. 2017;64:208–219. doi: 10.1016/j.bbi.2017.01.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Misiak B., Bartoli F., Carrà G., Stańczykiewicz B., Gładka A., Frydecka D., Samochowiec J., Jarosz K., Hadryś T., Miller B.J. Immune-inflammatory markers and psychosis risk: a systematic review and meta-analysis. Psychoneuroendocrinology. 2021;127 doi: 10.1016/j.psyneuen.2021.105200. [DOI] [PubMed] [Google Scholar]
  22. Mongan D., Raj Susai S., Föcking M., Byrne J.F., Zammit S., Cannon M., Cotter D.R. Associations between plasma inflammatory markers and psychotic disorder, depressive disorder and generalised anxiety disorder in early adulthood: a nested case-control study. Brain Behav. Immun. 2023;111:90–100. doi: 10.1016/j.bbi.2023.03.025. [DOI] [PubMed] [Google Scholar]
  23. Murphy J., Zierotin A., Mongan D., Healy C., Susai S.R., O'Donoghue B., Clarke M., O'Connor K., Cannon M., Cotter D.R. Associations between soluble urokinase plasminogen activator receptor (suPAR) concentration and psychiatric disorders - a systematic review and meta-analysis. Brain Behav. Immun. 2024;120:327–338. doi: 10.1016/j.bbi.2024.06.003. [DOI] [PubMed] [Google Scholar]
  24. Nielsen J., Røge R., Pristed S.G., Viuff A.G., Ullum H., Thørner L.W., Werge T., Vang T. Soluble urokinase-type plasminogen activator receptor levels in patients with schizophrenia. Schizophr. Bull. 2015;41(3):764–771. doi: 10.1093/schbul/sbu118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Rasmussen L.J.H., Petersen J.E.V., Eugen-Olsen J. Soluble urokinase plasminogen activator receptor (suPAR) as a biomarker of systemic chronic inflammation. Front. Immunol. 2021;12 doi: 10.3389/fimmu.2021.780641. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Reisinger A.C., Niedrist T., Posch F., Hatzl S., Hackl G., Prattes J., Schilcher G., Meißl A.M., Raggam R.B., Herrmann M., Eller P. Soluble urokinase plasminogen activator receptor (suPAR) predicts critical illness and kidney failure in patients admitted to the intensive care unit. Sci. Rep. 2021;11(1) doi: 10.1038/s41598-021-96352-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Roy A., Zeller J., Nero T.L., Klepetko J., Eisenhardt S.U., Parker M.W., McFadyen J.D., Peter K. C-reactive protein: the nexus between inflammation and protein misfolding diseases. Front. Immunol. 2025;16 doi: 10.3389/fimmu.2025.1612703. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Salminen A. GDF15/MIC-1: a stress-induced immunosuppressive factor which promotes the aging process. Biogerontology. 2024;26(1):19. doi: 10.1007/s10522-024-10164-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Sigvardsen C.M., Richter M.M., Engelbeen S., Kleinert M., Richter E.A. GDF15 is still a mystery hormone. Trends Endocrinol. Metabol.: TEM (Trends Endocrinol. Metab.) 2025;36(6):591–601. doi: 10.1016/j.tem.2024.09.002. [DOI] [PubMed] [Google Scholar]
  30. Tavenier J., Andersen O., Nehlin J.O., Petersen J. Longitudinal course of GDF15 levels before acute hospitalization and death in the general population. GeroScience. 2021;43(4):1835–1849. doi: 10.1007/s11357-021-00359-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Trotta A., Arseneault L., Danese A., Mondelli V., Rasmussen L.J.H., Fisher H.L. Associations between childhood victimization, inflammatory biomarkers and psychotic phenomena in adolescence: a longitudinal cohort study. Brain Behav. Immun. 2021;98:74–85. doi: 10.1016/j.bbi.2021.08.209. [DOI] [PubMed] [Google Scholar]
  32. Tsapakis E., Paitaridou D., Koummati M., Chovardas K., Fylaki S., Karakatsoulis G., Fountoulakis K. Manic symptoms in schizophrenia spectrum disorders. Brain Medicine. 2025;1(2):53–59. doi: 10.61373/bm025r.0005. [DOI] [Google Scholar]
  33. Yuan N., Chen Y., Xia Y., Dai J., Liu C. Inflammation-related biomarkers in major psychiatric disorders: a cross-disorder assessment of reproducibility and specificity in 43 meta-analyses. Transl. Psychiatry. 2019;9(1):233. doi: 10.1038/s41398-019-0570-y. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Data will be made available on request.

Articles from Brain, Behavior, & Immunity - Health are provided here courtesy of Elsevier

RESOURCES