Metabolic hormone and adipokine alterations in major depressive disorder in relation to the acute-phase inflammatory response and early-life adversity

Abstract

Background

Major depressive disorder (MDD) involves dysregulated neuroimmune, metabolic, and inflammatory pathways. This study characterized metabolic hormone and adipokine profiles in Chinese MDD patients stratified for metabolic syndrome (MetS), and examined associations with depression severity (OSOD), suicidal ideation (SI), illness recurrence (ROI), and physiosomatic symptoms.

Methods

We enrolled 125 MDD inpatients (age 18–70 years) and 40 healthy controls (age 20–65 years). Fasting serum insulin, glucose, glucagon, GIP, GLP‑1, leptin, secretin, PAI‑1, resistin, ghrelin, and adiponectin were measured. The acute‑phase inflammatory (API) response was assessed using albumin, transferrin, and monomeric CRP. Group comparisons used ANOVA or general linear models (adjusted for age, BMI, MetS) with false discovery rate correction. Associations were tested with Pearson correlations, stepwise multiple regression, and binary logistic regression. Discriminatory performance was evaluated by ROC‑AUC.

Results

MDD showed significantly lower insulin, glucagon, and PAI‑1, along with a higher API index (all adjusted). A composite GAP index (ghrelin, adiponectin, PAI‑1) correlated negatively with OSOD, SI, ROI, physiosomatic symptoms, and adverse childhood experiences (ACEs). A model combining GAP index, API index, and ACEs discriminated MDD from controls with AUC = 0.864 and 80% accuracy.

Conclusion

Severe MDD in this Chinese inpatient sample is characterized by suppressed anabolic hormones and lower adipokines coupled with mild chronic inflammation, independent of MetS. This hormonal‑immune‑metabolic signature is integral to MDD pathophysiology.

Plain language summary

This study examined metabolic hormones and inflammatory markers in Chinese patients with severe depression. Compared to healthy controls, depressed patients had lower levels of several key hormones (insulin, glucagon, and PAI-1) but higher levels of inflammation markers. A combination of three hormones—ghrelin, adiponectin, and PAI-1 (called the GAP index)—was strongly linked to depression severity, suicidal thoughts, illness recurrence, and physical symptoms. Patients who reported difficult childhood experiences also showed more severe illness. A model combining the GAP index, inflammation markers, and childhood adversity accurately distinguished patients with depression from healthy individuals. These findings suggest that severe depression involves a distinct pattern of hormone changes driven by chronic low-grade inflammation. This work may help guide more personalized treatment approaches for depression in the future.

Chen, Luo et al. examine metabolic hormones and inflammatory markers in Chinese patients with major depressive disorder. They find a distinct pattern of suppressed anabolic hormones and elevated inflammation, which combined with early life adversity accurately distinguishes patients from controls.

Introduction

Major depressive disorder (MDD) is increasingly reconceptualized as a systemic disorder involving dysregulation of neuroimmune, metabolic, and oxidative stress (NIMETOX) pathways¹. A major neuro-immune marker in MDD is the presence of an acute-phase inflammatory (API) response as indicated by lower serum albumin and transferrin (two negative acute-phase proteins) and increased monomeric C-reactive protein (mCRP), a positive acute-phase response². Combining these three markers into a single composite score (the API index) increases accuracy for MDD². Increased levels of pro-inflammatory cytokines in MDD, such as interleukin (IL)−1, IL-6, and tumor necrosis factor-alpha (TNF-α), induce the API response³.

Growing evidence indicates that MDD is consistently accompanied by a pro-atherogenic lipid signature—lower high-density lipoprotein cholesterol (HDL), apolipoprotein (Apo)A1, and a diminished index of reverse cholesterol transport—observed across multiple populations^4,5. Our results show that in Chinese patients without MetS/obesity, these deficits persist after adjusting for the API response, confirming lipid dysregulation is intrinsic to MDD⁶.

Within the NIMETOX framework, metabolic disorders, including metabolic syndrome (MetS), represent a pathological link between peripheral NIMETOX pathways and brain dysfunction^1,7. MetS is characterized by insulin resistance (IR) and increased atherogenicity, comprising a cluster of metabolic abnormalities including central obesity, dyslipidemia, hypertension, and impaired glucose metabolism that collectively promote cardiovascular disease risk⁸. Conversely, depressive states may exacerbate metabolic dysfunction through neuroendocrine and behavioral pathways, suggesting a bidirectional relationship⁹.

However, empirical evidence regarding IR in MDD is more heterogeneous than increased atherogenicity. A meta-analysis by Fernandes et al. demonstrates that IR is increased in MDD, particularly in Western studies¹⁰. However, these findings lack adjustment for MetS status, highlighting the need for more nuanced investigations that disentangle primary metabolic disturbances in MDD from those secondary to comorbid MetS. Thus, after adjusting for MetS and body mass index (BMI), such differences were not always evident^8,11. While large-scale studies, including in Chinese cohorts, show that elevated IR markers like the triglyceride-glucose (TyG) index predict depression risk¹², a study by Luo et al. revealed that Chinese patients with MDD—after controlling for MetS—had lower fasting insulin and IR index than controls¹³. This report identified MetS as a critical effect modifier: in its absence, mild inflammation in MDD may induce a context-dependent enhancement of insulin sensitivity; in its presence, pre-primed inflammation exacerbates IR¹³. This underscores the need for metabolically stratified analyzes rather than blanket assertions of a positive MDD-IR correlation.

The NIMETOX disturbance in MDD extends beyond classic hormones to a broader “immune-metabolic-endocrine axis,” involving metabolic hormones and adipokines that communicate between metabolic tissues, the immune system, and the brain^1,14. These hormonal mediators are pivotal in linking metabolic dysregulation to core clinical outcomes. Adiponectin, an insulin-sensitizing and anti-inflammatory adipokine, exerts its effects via AMPK activation and suppression of pro-inflammatory cytokine production in macrophages and microglia; its reduction in MDD may therefore contribute to heightened neuroinflammation and insulin resistance^15,16. Ghrelin, an orexigenic hormone, modulates the hypothalamic-pituitary-adrenal (HPA) axis, enhances hippocampal neurogenesis, and exerts anti-inflammatory effects through vagal pathways and by suppressing IL-1β and TNF-α; its altered levels in MDD are linked to changes in appetite and stress-response dysregulation^17,18. Plasminogen Activator Inhibitor-1 (PAI-1), while classically linked to fibrinolysis, also serves as a pro-inflammatory adipokine; its paradoxical reduction in some MDD patients may reflect a severe catabolic state driven by chronic inflammation¹⁹. Other hormones, including glucagon, GLP-1, and leptin, further contribute to this dysregulated network, affecting energy homeostasis and neuroplasticity²⁰. Collectively, this dysregulated network suggests a distinct metabolic-endocrine subtype of depression²¹.

Clinically, MDD is multifaceted, with bifactor modeling revealing that symptoms load onto a general factor of overall severity of depression (OSOD) and a physiosomatic factor^22,23. Recurrence of illness (ROI)—encompassing lifetime episodes and suicidal behaviors—is a key feature of MDD trajectory^24,25 and is strongly predicted by adverse childhood experiences (ACEs)^26,27. Mechanistically, ACEs program long-term alterations in neuroendocrine, immune, and metabolic systems through persistent sensitization of the HPA axis and induction of chronic low-grade inflammation, thereby establishing a vulnerability phenotype characterized by heightened inflammatory responses to subsequent stressors and lasting metabolic dysregulation^28–30. This ACE-ROI axis is linked to progressive sensitization of NIMETOX pathways, including inflammation and oxidative stress, and mediates the effects of ACEs on worsening OSOD, suicidal ideation, and physiosomatic symptoms^26,27.

Despite this, the specific interrelationships between hormonal-metabolic markers (e.g., ghrelin, PAI-1, adiponectin) and these key clinical constructs remain poorly delineated, especially in Chinese populations. Current knowledge relies heavily on Western data, overlooking ethnic variations and differences in body composition (e.g., visceral adiposity) and lifestyle³¹. Chinese populations exhibit distinct metabolic characteristics compared to Western cohorts, including lower BMI thresholds for metabolic risk, different patterns of obesity and visceral adiposity, and unique dietary habits (e.g., high carbohydrate intake)³¹. Moreover, the prevalence of MetS in Chengdu, Sichuan, China³² is lower than in Western Europe³³ and the USA³⁴. The differences in obesity are even more striking, namely 5–7% (BMI > 30) or 8-12 (BMI > 28) in Southwest China, including Sichuan³⁵, versus 23% in Western Europe³⁶, and 41.92% in the USA³⁷. These differences may influence the relationship between MDD, metabolic hormones, and inflammation. Consequently, research mapping the relationships between IR, metabolic hormones, inflammation, and multidimensional clinical features in Chinese MDD is urgently needed.

Hence, we conduct a study to investigate the alterations in metabolic hormones and adipokines among Chinese MDD patients, and to examine their associations with the core clinical dimensions of MDD, namely OSOD, current SI, ROI, and physiosomatic symptoms. In addition, we explore the associations between hormonal/metabolic markers and the API response in MDD.

Results

Sociodemographic and basic clinical characteristics

Table 1 presents the sociodemographic and basic clinical profiles of patients and controls. No statistically significant differences were observed between the two groups in age, sex, BMI, WC, years of education, prevalence of MetS, or smoking history. Patients with MDD showed higher OSOD, ROI, SI, physiosomatic, and ACE scores than controls. General linear model (GLM) analysis revealed that the use of antidepressants, mood stabilizers, atypical antipsychotics, or benzodiazepines did not significantly affect the levels of insulin, glucagon, ghrelin, PAI-1, adiponectin, or the API index after False Discovery Rate (FDR) correction (all p > 0.05). These findings indicate that the observed hormonal and inflammatory alterations are unlikely to be attributable to medication use.

Table 1.

Clinical, socio-demographic, and biochemical data of patients with major depressive disorder (MDD) and healthy controls (HC)

Variables	HC (n = 40)	MDD (n = 125)	F/χ²	df	p
Age (years)	37.1 (13.8)	35.7 (12.1)	0.37	1/163	0.542
Gender (m/f)	13/27	38/87	0.06	1	0.802
BMI (kg/m²)	23.52 (4.07)	22.30 (3.6)	3.58	1/163	0.06
Waist circumference (cm)	79.38 (11.73)	78.27 (11.32)	0.28	1/163	0.595
Ranking MetS	1.590 (1.46)	1.42 (1.26)	0.5	1/161	0.479
Education years (years)	13.88 (4.34)	13.54 (3.33)	0.26	1/163	0.613
Metabolic syndrome (MetS) (No/Yes)	29/10	102/22	1.17	1	0.355
History of smoking (No/Yes)	37/3	101/24	3.03	1	0.091
OSOD	−1.518 (0.237)	0.528 (0.491)	640.21	1/153	<0.001
Four ACEs	−0.785 (0.170)	0.150 (0.110)	22.87	1/155	<0.001
Physiosomatic	−1.316 (0.120)	0.513 (0.091)	158.22	1/139	<0.001
Current SI	−0.687 (0.163)	0.222 (0.117)	15.419	1/145	<0.001
ROI	−1.032 (0.140)	0.274 (0.101)	31.793	1/145	<0.001

All results are shown as mean (SD).

BMI body mass index, OSOD overall severity of depression, ACEs adverse childhood experiences, Current SI current suicidal ideation, ROI reoccurrence of illness index.

Metabolic hormone alterations in MDD

Table 2 shows the significant alterations in hormonal and metabolic profiles observed in MDD patients after adjusting for age and BMI (sex was not significant). Key findings include a significant reduction in fasting insulin, glucagon, and PAI-1 levels. Concurrently, the API index was markedly elevated in the MDD group. No significant intergroup differences were found for GIP, GLP-1, leptin, secretin, resistin, or adiponectin.

Table 2.

Differences in hormone profiles between major depressive disorder (MDD) and healthy controls (HC)

Variables	HC (n = 40)	MDD (n = 125)	F/χ²	df	P	Partial Eta squared
FPG (mmol/L)	5.563 (0.086)	5.374 (0.056)	3.841	1/155	0.109	0.023
Insulin (mU/L)	8.792 (0.586)	6.559 (0.380)	10.967	1/155	0.005	0.066
API index	−1.437 (0.346)	0.498 (0.225)	23.560	1/155	<0.001	0.132
Ghrelin (unmeasurable/measurable)	32/7	119/6	7.041	1	0.038	–
GIP (pg/ml)	70.193 (6.678)	61.319 (4.336)	1.333	1/155	0.288	0.009
GLP-1 (pg/ml)	32.769 (7.417)	17.153 (4.816)	3.345	1/155	0.110	0.021
Glucagon (pg/ml)	49.168 (3.761)	37.546 (2.468)	9.556	1/155	0.008	0.058
Leptin (ng/ml)	5.288 (0.858)	5.111 (0.557)	0.032	1/155	0.858	<0.001
Secretin (pg/ml)	56.078 (8.500)	36.212 (5.519)	4.122	1/155	0.094	0.026
Resistin (ng/ml)	2.615 (0.347)	1.895 (0.225)	3.252	1/155	0.110	0.021
PAI-1 (ng/ml)	3.329 (0.465)	1.795 (0.305)	8.163	1/131	0.015	0.059
Adiponectin (ng/ml)	10.718 (1.479)	9.315 (0.961)	0.679	1/155	0.440	0.004

All results are shown as estimated marginal means (SE), after adjusting for age and BMI.

FPG fasting plasma glucose, API acute phase inflammatory, GLP-1 glucagon-like peptide-1, PAI-1 plasminogen activator inhibitor-1.

Adipokine alterations in MDD

To assess the accuracy of the aforementioned biomarkers, a series of binary logistic regression models was constructed (Table 3, and Supplementary Table 2). The discriminatory performance of the models improved progressively as variables from different pathophysiological domains were integrated: a baseline model containing only hormonal markers (ghrelin, PAI-1, adiponectin) yielded an area under the curve (AUC) of 0.741. Based on this model, we constructed a z-unit-based adipokine composite score as z PAI-1 + z adiponectin + ghrelin (0 vs 1), labeled the GAP index (the first letter of the markers). The inclusion of a metabolic marker (FPG) did not significantly improve performance; the addition of the inflammatory marker (API index) increased the AUC to 0.776, and further incorporation of sex and ACEs resulted in an AUC of 0.818. The optimal diagnostic model was constructed by integrating the GAP index, API index, and ACEs (Model 5). This model demonstrated adequate discriminatory power: AUC = 0.864 (SE = 0.031), with an overall accuracy of 80.0% (sensitivity 79.8%, specificity 74.4%). The corresponding receiver operating characteristic (ROC) curve is shown in Fig. 1.

Table 3.

Results of binary logistic regression analysis examining the discrimination of major depressive disorder (MDD) and healthy controls (HC)

							95% C.I. for Exp (B)
Dichotomies		Explanatory variables	β	S.E.	Wald (df = 1)	p	Exp(B)	Lower	Upper
MDD versus HC (all subjects)	Model 1	Ghrelin	−1.426	0.655	4.747	0.029	0.240	0.067	0.867
		PAI-1	−0.705	0.229	9.499	0.002	0.494	0.316	0.774
		Adiponectin	−0.401	0.210	3.625	0.057	0.670	0.444	1.012
	Model 2	Ghrelin	−1.631	0.663	6.047	0.014	0.196	0.053	0.718
		PAI-1	−0.721	0.230	9.835	0.002	0.486	0.310	0.763
		FPG	−0.595	0.268	4.939	0.026	0.552	0.327	0.932
	Model 3	Ghrelin	−1.569	0.765	4.210	0.040	0.208	0.047	0.932
		PAI-1	−0.765	0.259	8.703	0.003	0.465	0.280	0.773
		API index	0.647	0.143	20.359	<0.001	1.910	1.442	2.529
	Model 4	Ghrelin	−1.548	0.820	3.566	0.059	0.213	0.043	1.060
		PAI-1	−0.748	0.286	6.836	0.009	0.473	0.270	0.829
		API index	0.527	0.158	11.160	0.001	1.695	1.244	2.309
		sex	0.726	0.323	5.037	0.025	2.066	1.096	3.893
		Four ACEs	0.895	0.329	7.409	0.006	2.448	1.285	4.665
	Model 5	GAP index	−0.828	0.281	8.691	0.003	0.437	0.252	0.758
		API index	0.500	0.146	11.721	0.001	1.649	1.238	2.195
		Four ACEs	1.150	0.315	13.290	<0.001	3.158	1.702	5.860

PAI-1 plasminogen activator inhibitor-1, API acute phase inflammatory, ACEs adverse childhood experiences, GAP Ghrelin+ Adiponectin+ PAI-1.

Fig. 1.

The receiving operating curve discriminating major depressive disorder from controls using GAP index, API index, and Four ACEs as explanatory variables.

Inflammatory markers and the API index

Pearson correlation analysis in Table 4 shows the relationships between key variables. The API index showed significant negative correlations with PAI-1, adiponectin, and the GAP composite index. Concurrently, PAI-1, adiponectin, and the GAP index were significantly positively correlated with FPG, insulin, and the IR index (z FPG + z INS).

Table 4.

Intercorrelation matrix (Pearson’s correlation coefficients) between hormone profile, insulin biomarkers, and metabolic indicators

Variables	Four_ACEs	API index	FPG	Insulin	z FPG + z INS
Ghrelin	−0.053	−0.118	0.042	0.012	0.032
GIP	−0.099	0.044	0.144	0.182*	0.189*
GLP-1	−0.111	−0.027	0.114	0.024	0.080
Glucagon	0.038	−0.089	0.141	0.147	0.167*
Leptin	0.025	−0.023	0.136	0.277**	0.240**
Secretin	−0.018	0.033	0.168*	−0.064	0.061
PAI-1	−0.204**	−0.189*	0.298**	0.217**	0.299**
Adiponectin	−0.217**	−0.319**	0.126	0.216**	0.198*
Resistin	−0.037	−0.170*	0.131	−0.022	0.063
GAP index	−0.270**	−0.340**	0.274**	0.274**	0.318**

ACEs adverse childhood experiences, API acute phase inflammatory, FPG fasting plasma glucose, INS insulin, GIP glucose-dependent insulinotropic polypeptide, GLP-1 glucagon-like peptide-1, PAI-1 plasminogen activator inhibitor-1, GAP zGhrelin+ zAdiponectin+ zPAI-1.

** Correlation is significant at the 0.01 level; * Correlation is significant at the 0.05 level.

Associations with early-life adversity and clinical outcomes

Table 5 presents correlations between hormonal/metabolic indicators and clinical symptoms. Ghrelin, PAI-1, adiponectin, FPG, and the IR index were all significantly negatively correlated with OSOD, physiosomatic symptoms, and ROI. Among these, the GAP composite index showed the strongest negative correlations with all clinical dimensions.

Table 5.

Intercorrelation matrix (Pearson’s correlation coefficients) between hormone profile and severity scales

Variables	Current SI	OSOD	Physiosomatic symptoms	ROI	Weight loss
Ghrelin	−0.053	−0.211*	−0.13	−0.210**	−0.212**
GIP	−0.099	−0.114	−0.104	−0.139	−0.041
GLP-1	−0.111	−0.083	−0.008	−0.127	−0.097
Glucagon	0.038	−0.026	−0.003	−0.059	−0.019
Leptin	0.025	−0.097	−0.058	−0.049	−0.193*
Secretin	−0.018	−0.070	−0.001	−0.129	−0.081
PAI-1	−0.204**	−0.316**	−0.246**	−0.199*	−0.185*
Adiponectin	−0.217**	−0.220**	−0.220**	−0.159*	−0.128
Resistin	−0.037	−0.194*	−0.172*	−0.234**	−0.093
Insulin	−0.110	−0.200*	−0.187*	−0.201*	−0.118
FPG	−0.216**	−0.227**	−0.185*	−0.277**	−0.186*
z FPG + z INS	−0.189*	−0.246**	−0.214**	−0.277**	−0.176*
GAP index	−0.270**	−0.378**	−0.318**	−0.266**	−0.238**

Current SI current suicidal ideation, OSOD overall severity of depression, API acute phase inflammatory, FPG fasting plasma glucose, INS insulin, GIP glucose-dependent insulinotropic polypeptide, GLP-1 glucagon-like peptide-1, PAI-1 plasminogen activator inhibitor-1, GAP Ghrelin+ Adiponectin+ PAI-1, ROI reoccurrence of illness index.

** Correlation is significant at the 0.01 level; * Correlation is significant at the 0.05 level

Further multiple regression analysis (Table 6) quantified the contribution of these predictors. ACEs and the API index were the most consistent and significant positive predictors for OSOD, physiosomatic symptoms, and illness recurrence. Additionally, specific hormonal markers had independent predictive value: for example, ghrelin and PAI-1 were negative predictors of OSOD, while resistin and GLP-1 were negative predictors of current SI. These models collectively explained between 30.1% and 40.2% of the variance in clinical symptom severity.

Table 6.

Results of multiple regression analysis with clinical rating scale scores as dependent variables

Dependent variables	Explanatory variables	Coefficient statistics			Model statistics
		β	t	p	R2	F	df	p
#1, OSOD	Model				0.402	23.819	4/146	<0.001
	Four ACEs	0.356	5.154	<0.001
	API index	0.298	4.268	<0.001
	PAI-1	−0.148	−2.533	0.012
	Ghrelin	−0.518	−2.258	0.025
#2, Current SI	Model				0.301	15.934	4/152	<0.001
	Four ACEs	0.464	6.699	<0.001
	Resistin	−0.158	−2.295	0.023
	GLP-1	−0.157	−2.255	0.026
	Leptin	−0.147	−2.130	0.035
#3, physiosomatic	Model				0.313	21.675	3/146	<0.001
	Four ACEs	0.294	3.835	<0.001
	API index	0.297	4.071	<0.001
	Gender	0.181	2.459	0.015
#4, Melancholia	Model				0.358	13.466	4/152	<0.001
	Four ACEs	0.358	5.113	<0.001
	API index	0.251	3.566	<0.001
	Gherlin	−0.179	−2.696	0.008
	PAI-1	−0.148	−2.166	0.032
#5, ROI	Model				0.362	22.532	4/148	<0.001
	Four ACEs	0.494	7.455	<0.001
	Resistin	−0.178	−2.646	0.009
	Gherlin	−0.145	−2.190	0.030
	FPG	−0.138	−2.055	0.042

OSOD overall severity of depression, ACEs adverse childhood experiences, API acute phase inflammatory, PAI-1 plasminogen activator inhibitor-1, GLP-1 glucagon-like peptide-1, Current SI current suicidal ideation, ROI reoccurrence of illness index.

Discussion

This study delineates a distinct neuroendocrine-metabolic profile in Chinese MDD patients, characterized by lower fasting insulin, glucagon, and PAI-1 levels, alongside an elevated API index.

This profile of reduced insulin and PAI-1, in the context of heightened inflammation, appears counterintuitive when viewed against the conventional paradigm linking MDD to IR and elevated pro-inflammatory adipokines, as often reported in Western cohorts¹⁰. A critical contextual factor is the relatively low prevalence of MetS and obesity in our Chengdu sample, consistent with regional epidemiological data³⁸ and our previous report^6,13. In populations with high MetS prevalence, metabolic dysregulation (e.g., IR, elevated PAI-1) often co-occurs with and may be amplified by MDD, potentially obscuring the primary hormonal alterations intrinsic to the mood disorder^5,39.

We observed that the suppressed hormonal profile was inversely associated with the elevated API index, a composite measure reflecting a chronic, smoldering inflammatory response^3,6. This finding underscores chronic mild inflammation as a pivotal mediator of the observed endocrine disturbances.

The inverse association between the API index and PAI-1/adiponectin aligns with established biology whereby systemic inflammation actively suppresses the synthesis and secretion of these metabolic hormones. Pro-inflammatory cytokines, such as IL-6 and TNF-α—which are frequently elevated in MDD—are potent inhibitors of adiponectin gene expression and secretion from adipose tissue^1,40. Similarly, while PAI-1 is often viewed as an inflammatory marker itself, its production in visceral adipose tissue and liver can be differentially regulated; sustained, low-grade inflammation as seen in our cohort may disrupt normal regulatory feedback, leading to the observed reduction^19,41. The significant correlation between lower PAI-1/adiponectin and higher IR index further supports a model where inflammation-driven hormonal suppression contributes to a catabolic or maladaptive metabolic state, even in the absence of classical hyperinsulinemia.

This hormonal suppression extends beyond adipokines. The reduction in fasting insulin and glucagon may also be viewed through the lens of a chronic acute-phase response. The liver, the primary site for synthesizing negative acute-phase reactants (albumin, transferrin) and a key target of inflammatory signaling, may experience a reprioritization of protein synthesis during inflammation^42,43. Cytokines like TNF-α can directly inhibit hepatic synthesis of albumin and transferrin⁴⁴, and this catabolic milieu may concurrently dampen pancreatic endocrine function or enhance insulin clearance, contributing to lower circulating insulin levels¹³. This state of “inflammatory hypometabolism” is consistent with earlier observations in MDD linking lower visceral protein levels (albumin, transferrin) to anorexia, weight loss, and a malnutrition-inflammation axis^43,45. The reduction in anabolic hormones (e.g., insulin, ghrelin) observed in our cohort may further reflect a chronic, smoldering inflammatory state that drives a catabolic milieu. This might be exacerbated by the anorectic effects of pro-inflammatory cytokines (e.g., IL-1, IL-6, TNF-α), which are known to suppress appetite and contribute to weight loss—a common clinical feature in our MDD patients.

Our findings extend this concept, suggesting that the chronic inflammatory process in MDD orchestrates a coordinated downregulation of anabolic hormones and hemostatic factors. Thus, the observed hormonal profile is not merely a comorbidity but appears integral to the pathophysiology of MDD in this population, directly linked to the API response. This reinforces the concept of MDD as a disorder of NIMETOX pathways, where immune activation exerts downstream disruptive effects on metabolic and endocrine homeostasis^1,29.

Our third major finding shows significant negative correlations between key hormones (ghrelin, PAI-1, adiponectin), IR indices, and the core clinical constructs examined here. The composite GAP index emerged as the strongest hormonal correlate^1,29.

The predictive value of these hormonal markers in multivariate models substantiates their clinical relevance. Ghrelin and PAI-1 independently predicted OSOD and melancholia severity, while resistin and GLP-1 were negative predictors of current SI. Notably, the GAP index—which captures the combined signal of ghrelin, adiponectin, and PAI-1—consistently showed the strongest associations with clinical outcomes, suggesting that these three markers converge to reflect a shared underlying pathophysiological state.

Critically, our multivariate regression models positioned ACEs and the API index as the most consistent and potent positive predictors across all clinical domains. This supports a pathogenic model wherein ACEs might establish a lifelong vulnerability^26,30. Within this sensitized framework, subsequent stressors may trigger a chronic smoldering inflammatory response (high API), which in turn suppresses protective hormonal secretion (low GAP). The combined contribution of ACEs, inflammation, and hormonal deficiency explained up to 40% of the variance in clinical severity, underscoring that these domains collectively drive the clinical phenotype of MDD rather than operating independently. This triple hit—ACEs × Inflammation × Hormonal Deficiency—might drive the phenome of MDD, including overall severity, somatic burden, suicidal ideation, and the tendency towards recurrence^24,46. Our findings thus integrate the hormonal dysregulation into the previously established “ACE-ROI-inflammation” axis, illustrating how metabolic hormones serve as both mediators and modulators of this core pathological pathway³⁰.

Building upon the distinct profile identified in this study—suppressed anabolic hormones (low GAP index), elevated smoldering inflammation (high API index), and preserved insulin sensitivity—we propose this constellation delineates a specific “hormonal-immune-metabolic” (HIM) subtype of MDD. This subtype is characterized by a catabolic hormonal state driven by chronic low-grade inflammation, rather than the hypermetabolic, insulin-resistant profile commonly reported in Western cohorts with high MetS prevalence.

Our data provide empirical support for this subtyping framework. First, the GAP index, API index, and ACEs collectively distinguished MDD from controls with high accuracy, demonstrating that these three domains capture distinct but complementary aspects of pathophysiology. Second, within MDD patients, the API index showed significant negative correlations with PAI-1, adiponectin, and the GAP index, supporting the model wherein inflammation drives hormonal suppression. Third, the GAP index correlated positively with insulin and FPG, indicating that preserved or enhanced insulin sensitivity accompanies this hormonal profile—a pattern distinct from the insulin resistance typically associated with obesity-driven inflammation.

The contrast with profiles often documented in Western studies is enlightening. The prevalence of MetS in our sample (22.4%) and in our province (Sichuan, China) is comparatively lower than that observed in Western Europe and particularly in the USA, but the prevalence of obesity is significantly lower in Sichuan, China, than in Western nations. Moreover, here we report that the inflammatory and metabolic characteristics differ substantially. In populations with high obesity prevalence, MDD is frequently associated with elevated IR, PAI-1, and pentameric CRP—a pattern driven by the superimposition of depressive pathophysiology on a pre-existing state of adiposity-driven meta-inflammation^10,47,48. In such contexts, inflammatory pathways are pre-primed, and adipokine secretion is skewed toward a pro-inflammatory, insulin-resistant state^49,50. Conversely, in our sample with low obesity prevalence (mean BMI 22.3 kg/m²), the inflammatory tone is of a different nature: a chronic, low-grade, “smoldering” response, marked predominantly by a negative acute-phase reaction (low albumin, transferrin) and a modest rise in monomeric CRP³. This type of inflammation, potentially linked to persistent low-grade immune activation and catabolism, may create a milieu that paradoxically enhances insulin sensitivity in the short term—a hormetic effect observed in some contexts of mild inflammation⁵¹—while simultaneously suppressing the synthesis of hormones like adiponectin, PAI-1, and ghrelin.

Our findings directly support a region- and metabolism-stratified understanding of MDD pathophysiology. In this Chinese cohort with low background MetS and obesity prevalence, the MDD-associated hormonal profile—suppressed insulin, glucagon, and PAI-1 with preserved insulin sensitivity—contrasts sharply with the hyperinsulinemic, pro-inflammatory, insulin-resistant profile commonly reported in Western cohorts where obesity and MetS are more prevalent^10,48. This stratification explains the discrepant findings regarding IR and PAI-1 in MDD in the literature, highlighting that “metabolic dysregulation” in MDD is not monolithic. Recognizing this HIM subtype is therefore crucial for biomarker interpretation, disease subtyping, and personalized treatment.

The transdiagnostic relevance of this immunometabolic axis warrants consideration. Similar profiles of suppressed anabolic hormones and elevated inflammatory markers have been observed in bipolar disorder and schizophrenia, particularly among patients with histories of early-life adversity^29,39. This raises the possibility that the HIM axis represents a shared pathophysiological dimension across major psychiatric disorders, with implications for staging models and mechanism-informed treatment selection. Future research should explore whether these biomarkers can differentiate unipolar depression from bipolar disorder during depressive episodes—a critical unmet clinical need.

The integrated biomarker model developed in this study—combining the hormonal GAP index, the inflammatory API index, and a history of ACEs achieved an AUC of 0.864, demonstrating that peripheral profiling can effectively distinguish MDD patients from healthy controls. This framework opens avenues for mechanism-informed treatment stratification. For instance, patients with this specific HIM profile might be prioritized for treatments aimed at correcting the catabolic state. Conversely, patients with MDD and comorbid MetS might exhibit a different profile (high pCRP, high PAI-1, high IR) and may benefit more from insulin sensitizers or statins^52,53. Thus, the GAP and API indices could serve as objective tools to guide personalized therapeutic decisions.

Future research must validate these biomarkers in longitudinal cohorts to establish their predictive value for treatment response and disease progression. Intervention studies targeting the inflammation-hormone axis (e.g., with anti-cytokine agents or hormone-modulating drugs) in patients selected based on this profile represent a critical next step toward translating these mechanistic insights into clinical practice. From a clinical translation perspective, while the current diagnostic model demonstrates high accuracy in distinguishing MDD from healthy controls, its utility may extend beyond this binary discrimination. Early diagnosis of depression in primary care settings remains challenging due to the overlap with general somatic complaints; the GAP-API-ACEs model could serve as an objective adjunctive tool to aid in early detection. Beyond early diagnosis, future studies should also explore whether this hormonal-immune-metabolic signature can differentiate unipolar depression from bipolar disorder during a depressive episode—a critical unmet clinical need. Moreover, integrating peripheral blood biomarkers with other modalities, such as neuroimaging, gut microbiome composition, and metabolomics, would likely yield greater predictive accuracy and provide deeper mechanistic insights into the brain-gut-immune axis, moving toward a truly multi-aspect, precision psychiatry approach.

Limitations

This study has several limitations that warrant consideration. The cross-sectional design precludes causal inferences, and the single-center recruitment from southwestern China may limit generalizability. Although we controlled for major confounders, residual influences from unmeasured variables (e.g., diet, physical activity, and medication histories) cannot be entirely ruled out. While psychotropic medication use was statistically controlled, potential long-term modulation of endocrine axes cannot be discounted. The absence of a family history of metabolic diseases represents a residual confounding factor, and single fasting blood samples may not capture dynamic secretory patterns or circadian rhythms critical for hormones such as ghrelin and cortisol. Emerging evidence suggests that patients with atypical depression may also display aberrations in NIMETOX pathways^54,55. Nevertheless, we were unable to include patients with atypical depression, although our MDD patients were consecutively admitted to the hospital. Atypical MDD patients generally have a lower score on the HAMD⁵⁶, which may explain the lower prevalence of this subtype in MDD inpatients (selected based on high symptom severity and functional collapse), e.g., between 7 and 15%^57,58. Additionally, our study focused on higher-order clinical constructs (OSOD, physiosomatic symptoms, ROI, current SI) derived from bifactor modeling, which provide a parsimonious summary of the complex clinical phenomenon of MDD. Future studies should examine these individual symptom dimensions to determine whether the hormonal-immune-metabolic profile identified here is preferentially associated with particular symptom clusters, which could inform more targeted pathophysiological models and treatment approaches. Longitudinal studies with more frequent biosampling are needed to validate and extend our findings.

Conclusions

In summary, this study delineates a distinct hormonal-metabolic signature in Chinese patients with MDD, characterized by suppressed levels of key metabolic hormones (notably PAI-1, adiponectin, and ghrelin) within a context of chronic, smoldering inflammatory response, as indexed by an elevated API score. Critically, this profile is strongly associated with core clinical dimensions of the disorder, including overall severity, physiosomatic burden, and recurrence risk. Our findings extend the NIMETOX model of depression by integrating hormonal and adipokine components, revealing a pathophysiological axis where ACEs, persistent inflammation, and hormonal suppression converge to drive the clinical phenotype.

These results advocate for a stratified, context-dependent understanding of metabolic dysfunction in MDD. The identified HIM subtype may be particularly relevant in populations with low MetS prevalence. The integrative biomarker model (GAP-API-ACEs) demonstrates promising diagnostic and prognostic utility, paving the way for mechanism-informed subtyping and personalized treatment strategies that aim to modulate the inflammation-hormone axis in a targeted subset of patients.

Methods

Study design and participants

This cross-sectional case-control study enrolled 165 participants, including 125 inpatients meeting DSM-5 criteria for MDD and 40 age-, sex-, BMI-, and education-matched healthy controls. Participants were recruited from the Psychiatric Center of Sichuan Provincial People’s Hospital in Chengdu, China. MDD diagnosis was confirmed using the Mini-International Neuropsychiatric Interview (MINI), with a severity threshold of >18 on the 21-item Hamilton Depression Rating Scale (HAMD-21). We used a structured interview according to the Structured Interview for DSM-5 (SCID-5) to make the diagnosis of MDD⁵⁹.

Exclusion criteria for both groups comprised medical disorders, such as diabetes type 1, immune and autoimmune disorders, including inflammatory bowel disease, rheumatoid arthritis, neurological and neurodegenerative disorders, such as Parkinson’s and Alzheimer’s disease, multiple sclerosis), or psychiatric disorders, such as bipolar disorder, schizophrenia, substance use disorders, autism spectrum disorders), infectious disease, immunomodulatory treatments, treatments with therapeutical dosages of antioxidants or omega-3 polyunsaturated fatty acids²³. Metabolic syndrome (MetS) was not an exclusion criterion; instead, its presence was assessed and statistically controlled for in all analyzes to isolate the effects of MDD independent of MetS. We also excluded pregnant and lactating women. Controls were additionally excluded for MDD and a family history of mood disorders, suicide, or psychosis. A flow diagram detailing participant recruitment, inclusion/exclusion procedures, and the sequence of assessments is provided in Supplementary Fig. 1.

The study was approved by the ethics committee of Sichuan Provincial People’s Hospital ([Ethics (Research) 2024-203]), and all participants provided written informed consent.

Clinical and behavioral assessments

Trained clinicians conducted semi-structured interviews to collect demographic and clinical data. Supplementary Table 1 shows the instruments we employed to assess the clinical indices as described before²³. Based on those measurements, composite clinical constructs were generated as previously described to construct OSOD, current suicidal ideation (Current SI), ROI, and physiosomatic symptoms. OSOD was constructed as a general factor extracted via bifactor modeling from all symptom domains assessed using the HAMD-21, including affective, cognitive, vegetative, and physiosomatic items, as previously described²³. Physiosomatic symptoms were operationalized as a specific group factor derived from three sources: the Somatic Symptom Scale-8 (SSS-8), items assessing chronic fatigue syndrome (CFS)-like symptoms, and the physiosomatic subdomain of the HAMD-21. ROI, a continuous composite score reflecting the recurrence of MDD episodes and suicidal behaviors, was computed based on the number of lifetime depressive episodes, lifetime suicide attempts, and lifetime suicidal ideation using the Columbia-Suicide Severity Rating Scale (C-SSRS), as previously described²⁴. All regression models involving ROI included age as a covariate to account for this potential confounding effect. The frequency of suicidal ideation and attempts was assessed using the C-SSRS, which evaluated the frequency of suicidal behaviors over the past month. ACEs were quantified using the Childhood Trauma Questionnaire-Short Form⁶⁰.

Anthropometric and metabolic syndrome assessment

BMI was calculated using the measured height and weight. Waist circumference (WC) was measured at the midpoint between the iliac crest and the lowest rib. MetS was defined according to the 2009 Joint Interim Statement criteria⁶¹, requiring at least three of five components: elevated WC, triglycerides, blood pressure, fasting glucose, or reduced HDL-cholesterol. Participants with MetS were included in the study, and MetS status was entered as a covariate in all relevant analyzes to control for its potential confounding effects.

Assays

Fasting venous blood was collected in the morning (between 06:30 and 08:00 h) using EDTA-containing tubes. Following centrifugation, plasma was aliquoted and stored at –80 °C until analysis. Fasting plasma glucose (FPG) was measured using the glucose oxidase method on a fully automated biochemistry analyzer (ADVIA 2400, Siemens Healthcare Diagnostics Inc.). Insulin was assayed using a chemiluminescence immunoassay (Atellica IM insulin assay kit, Siemens Healthcare Diagnostics Inc.) on an Atellica IM analyzer. All biomarker concentrations are reported in standard units: FPG in mmol/L, insulin in mU/L, ghrelin, GIP, GLP-1, glucagon, secretin, resistin, PAI-1, and adiponectin in pg/mL or ng/mL. To assess insulin resistance, we computed a composite indicator (zFPG + zINS) by summing the standardized z-scores of FPG and insulin levels⁶². The API response was assessed by measuring plasma levels of albumin, transferrin (Tf), and monomeric C‑reactive protein (mCRP). The mCRP values were adjusted for the effects of age, sex, and BMI, and the residualized mCRP values were used in subsequent analyzes⁴¹. The API index was then calculated using the formula: z mCRP – z albumin – z transferrin.

Metabolic hormones (Ghrelin, GIP, GLP‑1, Glucagon, Leptin, Secretin) and adipokines (Resistin, PAI‑1, Adiponectin) were quantified using multiplex magnetic bead‑based immunoassays (MILLIPLEX® Human Metabolic Hormone Panel V3 and Human Adipokine Magnetic Bead Panel 1, respectively) based on Luminex® xMAP® technology. Assays were performed according to the manufacturer’s protocols, including preparation of standards and controls, plate washing, and signal acquisition on a Luminex instrument. The intra-assay coefficients of variation for all analytes were <10%, respectively. Data were analyzed using a 5‑parameter logistic curve‑fitting method to determine analyte concentrations.

Statistical analysis

Sample size estimation was performed a priori using G*Power 3.1.9.4. For the primary multiple regression analyzes, a minimum sample size of 51 participants was required based on an anticipated effect size of 0.33 (equivalent to approximately 25% explained variance), a two-tailed alpha of 0.05, a statistical power of 0.80, and a maximum of seven predictors. The current study, comprising 125 MDD patients and 40 healthy controls, exceeds this threshold, confirming adequate statistical power.

Statistical analyzes were performed using SPSS 30.0 (IBM Corp., Armonk, NY, USA). Continuous data are presented as mean ± standard deviation, and categorical data as frequencies (percentages). Group differences (MDD vs. controls) in continuous variables were tested using Analysis of Variance (ANOVA) or the GLM. GLM was employed to compare hormonal and adipokine biomarkers between groups while adjusting for the effects of age, sex, BMI, and MetS status, with results presented as estimated marginal means. Associations between categorical variables were assessed using the chi-square test. For all group comparisons, the FDR procedure was applied to adjust p-values for multiple testing.

Associations between continuous variables were examined using Pearson’s correlation analysis. To identify predictors of clinical outcomes (OSOD, ROI, Current SI, physiosomatic symptoms), multiple linear regression analyzes were conducted. Independent variables, including ACEs, hormonal biomarkers, and inflammatory indices, were entered into the model using stepwise selection. Model assumptions, including linearity, homoscedasticity, independence of residuals, and absence of multicollinearity, were checked and met. Binary logistic regression analysis was performed to construct diagnostic models differentiating MDD patients from healthy controls. Significant variables were then entered into the multivariate models using the forward conditional method. The model’s discriminatory power was evaluated by the AUC of the ROC curve. Model fit was assessed with the Hosmer-Lemeshow test, and the overall classification accuracy, sensitivity, and specificity were reported. Odds ratios (OR) with 95% confidence intervals (CI) were calculated for each significant predictor. The diagnostic performance of key biomarker combinations was further evaluated using ROC analysis. A two-tailed alpha level of 0.05 was considered statistically significant for all tests unless otherwise specified.

Ethical approval

This study was conducted in accordance with the Declaration of Helsinki and the relevant guidelines and regulations of China. The study protocol was approved by the Ethics Committee of Sichuan Provincial People’s Hospital (Approval No. [Ethics (Research) 2024-203]). Written informed consent was obtained from all participants before their inclusion in the study.

Author contributions

M.M. and Y.Q.Z. designed the study; Y.Q.Z., T.C.C., Y.Y.L. M.D.L., and F.A.A. performed the assays; T.C.C. wrote the first draft; M.M., Y.Q.Z. and M.K. revised the manuscript; all other authors edited the manuscript; M.Q.N. recruited patients. All authors have read and agreed to the published version of the manuscript.

Peer review

Peer review information

Communications Medicine thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This research was funded by the Chengdu Science and Technology Project (Grant No.: 2025-ZJ000-00044-WZ), Sichuan Science and Technology Program “PIANJI” Project (Grant No.: 2025HJPJ0004), and Sichuan Science and Technology Program (Grant No.2025ZNSFSC1567).

Data availability

All numerical source data underlying Fig. 1 and Tables 1–6 are provided in the Supplementary Data. This file includes all data points used to create the graphs and charts in the main figures. Additional requests should be directed to the principal investigator and corresponding author, Michael Maes.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s43856-026-01686-4.