Nonlinear association between body roundness index and depressive symptoms in women of reproductive age: The mediating role of systemic inflammation

Abstract

Background

Previous studies reported a linear association between Body Roundness Index (BRI) and depressive symptoms in general adults, but the relationship in women of reproductive age remains unclear. This study aimed to investigate the association between Body Roundness Index (BRI) and depressive symptoms and to examine the statistical mediating role of inflammatory markers in this association.

Methods

This cross-sectional study analyzed nationally representative data from 7451 women aged 18−45 years in the National Health and Nutrition Examination Survey (NHANES) 2005−2018. BRI was calculated from height and waist circumference. Depressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9). Inflammatory markers included white blood cell count (WBC), Systemic Immune-Inflammation Index (SII), Systemic Inflammation Response Index (SIRI), platelet-to-lymphocyte ratio (PLR), neutrophil-to-lymphocyte ratio (NLR), and monocyte-to-lymphocyte ratio (MLR). Weighted linear regression, threshold effect analysis, and mediation analysis were performed.

Results

BRI was nonlinearly associated with PHQ-9 score, with an exploratory turning point at approximately BRI = 3.85 (P for likelihood ratio = 0.020). Above the threshold, BRI was significantly associated with depressive symptoms (β = 0.20, 95% CI: 0.15, 0.25; P < 0.0001). The statistical association between BRI and PHQ-9 was partially accounted for by WBC (indirect association proportion: 9.3%, 95% CI: 1.1%, 18.6%; P = 0.024); the other five markers showed no significant mediation.

Conclusions

Among U.S. women of reproductive age, BRI was nonlinearly associated with depressive symptoms, with a significant positive association only above a turning point. The statistical pattern through WBC is consistent with a hypothesized immune-to-brain pathway linking visceral adiposity to mood; however, given the cross-sectional design, these findings do not establish causality.

Highlights

• •BRI and depressive symptoms showed a nonlinear relationship in women of reproductive age.
• •A turning point was identified at BRI ≈3.85, above which the positive association was significant.
• •WBC statistically accounted for 9.3% of the BRI–depression association.
• •SII, SIRI, PLR, NLR, and MLR showed no significant statistical mediation.

1. Introduction

Depressive disorders represent a leading cause of disability worldwide, with women of reproductive age being disproportionately affected (Dai et al., 2025; Yang et al., 2025). Depressive symptoms are particularly prevalent during the reproductive years, a period characterized by substantial physiological, psychosocial, and hormonal transitions that may increase vulnerability (Freeman et al., 2004). Persistent depressive symptoms during this life stage may impair quality of life and are associated with potential adverse consequences for maternal and offspring health (Slomian et al., 2019).

Obesity is a well-established risk factor for depressive symptoms, but traditional measures like Body Mass Index (BMI) have limitations in capturing the distribution of body fat (Luppino et al., 2010; Müller et al., 2016). The Body Roundness Index (BRI), a novel geometric index incorporating height and waist circumference, provides a more accurate assessment of body fat percentage and visceral adipose tissue (Rico-Martín et al., 2020; Thomas et al., 2013). As a simple and non-invasive measure derived from routinely collected anthropometric parameters, BRI may provide a practical approach for characterizing body fat distribution that complements traditional indicators such as BMI. From a clinical perspective, if such an association is consistently observed, BRI may help identify reproductive-age women with a higher burden of depressive symptoms who may benefit from further psychological assessment, particularly when considered alongside other metabolic and clinical information. Visceral fat, in particular, is metabolically active and contributes to systemic inflammation, which has been hypothesized to play a role in the development of depressive symptoms (Milaneschi et al., 2019). Previous studies have reported a positive linear association between BRI and depressive symptoms in general adult populations (Zhang et al., 2024; Zhou et al., 2026). However, these studies did not focus specifically on women of reproductive age, a group with distinct physiological and psychosocial profiles. The reproductive years are marked by cyclical fluctuations in estrogen and progesterone, which not only influence adipose tissue distribution and inflammatory responses but also directly affect mood-regulating neurotransmitter systems (Kundakovic and Rocks, 2022; Pillerová et al., 2022). Additionally, psychosocial stressors prevalent during this life stage, such as caregiving burden and work–life balance challenges, may interact with metabolic disturbances to increase vulnerability to depressive symptoms (Schweizer-Schubert et al., 2020). These hormonal, metabolic, and psychosocial characteristics may modify the strength or pattern of the association between visceral adiposity and depressive symptoms, potentially resulting in different exposure–outcome relationships compared with broader adult populations. Therefore, investigating the relationship between BRI and depressive symptoms specifically among women of reproductive age may provide additional insights beyond findings from general adult populations.

Inflammatory markers are often elevated in individuals with obesity and depressive symptoms and have been hypothesized to serve as a potential link between them (Foley et al., 2023). Elevated white blood cell count (WBC) and other inflammatory indices have been independently associated with both visceral adiposity and depressive symptoms (Rengasamy et al., 2022; van den Munckhof et al., 2024). Several recent studies have further advanced this field. Rengasamy et al. demonstrated that WBC and C-reactive protein (CRP) partially mediated the relationship between physical illness and depression severity in a large community sample (Rengasamy et al., 2022). A previous study by Xie found that lymphocyte-based ratios (platelet-to-lymphocyte ratio, neutrophil-to-lymphocyte ratio, and monocyte-to-lymphocyte ratio) were associated with postpartum depressive symptoms, with platelet-to-lymphocyte ratio showing a nonlinear relationship (Xie, 2026). Another study by Xie et al. further identified that multiple immune-inflammatory biomarkers (neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, and systemic immune-inflammation index) during the postpartum period were independent risk factors for postpartum depression (Xie et al., 2025). However, the potential mediating role of inflammatory markers in the BRI and depressive symptoms relationship, especially in women of reproductive age, has not been explored. Among these markers, WBC is a direct, stable measure of overall systemic inflammatory burden (Foley et al., 2023; Foy et al., 2025), making it a suitable candidate for examining the inflammatory pathway between visceral adiposity and depressive symptoms in this population.

Previous studies have primarily evaluated linear associations between BRI and depressive symptoms in general adult populations. However, whether this association exhibits a nonlinear pattern, particularly among women of reproductive age whose hormonal and metabolic characteristics may influence the relationship between adiposity, inflammation, and mood, remains unclear. Furthermore, the extent to which different inflammation-related biomarkers statistically account for the association between BRI and depressive symptoms in this specific population has not been systematically investigated. This study aimed to investigate the association between BRI and depressive symptoms in U.S. women of reproductive age, and to examine whether multiple inflammatory markers mediate this association. Specifically, we hypothesized that higher BRI would be associated with higher levels of depressive symptoms in women of reproductive age and explored whether this association exhibited a nonlinear pattern. Furthermore, we hypothesized that inflammatory markers would partially mediate this association in the statistical sense.

2. Materials and methods

2.1. Study population

Data were derived from the National Health and Nutrition Examination Survey (NHANES) cycles 2005–2018. NHANES is a nationally representative, cross-sectional survey conducted by the National Center for Health Statistics (NCHS) using a complex, multistage probability sampling design. The survey protocol was approved by the NCHS Research Ethics Review Board, and all participants provided written informed consent.

In this analysis, we initially selected women aged 18–45 years from the NHANES 2005–2018 cycles (Qi et al., 2024). To minimize potential confounding from pregnancy-related physiological changes in body composition, inflammatory status, and depressive symptoms, pregnant women and women within 12 months postpartum were excluded (n = 1080). We further excluded participants with missing data on BRI or PHQ-9 scores. The final analytic sample comprised 7451 women (Fig. 1).

Fig. 1.

Flowchart of participant selection. NHANES, National Health and Nutrition Examination Survey.

2.2. Assessment of body roundness index

Body Roundness Index was calculated based on height (cm) and waist circumference (WC, cm) using the formula proposed by Thomas et al., (2013). Data on height and waist circumference were obtained from the participants’ examination files. To ensure accuracy, body measurement data were collected by trained health technicians in Mobile Examination Centers (MECs). BRI was calculated as follows:

$$\text{BRI}=364.2-365.5\times \sqrt{1-{\left(\frac{\text{WC}/2\mathrm{\pi }}{0.5\times \text{Height}}\right)}^2}$$

2.3. Assessment of depressive symptoms

Depressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9), a validated screening tool for depression in primary care (Kroenke et al., 2001). The PHQ-9 consists of nine items corresponding to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria for major depressive disorder, each scored from 0 (“not at all”) to 3 (“nearly every day”). Total scores range from 0 to 27, with higher scores indicating greater symptom severity. In this study, PHQ-9 total score was used as a continuous outcome (Qiao et al., 2025). In sensitivity analyses, clinically relevant depressive symptoms were additionally defined as a PHQ-9 score ≥10, a commonly used threshold indicating clinically significant depressive symptoms, and were analyzed as a binary outcome.

2.4. Assessment of inflammatory markers

Six inflammatory markers were derived from complete blood count (CBC) data: white blood cell (WBC) count, Systemic Immune-Inflammation Index (SII), Systemic Inflammation Response Index (SIRI), platelet-to-lymphocyte ratio (PLR), neutrophil-to-lymphocyte ratio (NLR), and monocyte-to-lymphocyte ratio (MLR). WBC count was obtained directly from the CBC analysis. SII was calculated as (platelet count × neutrophil count)/lymphocyte count (Huang, 2023). SIRI was calculated as (neutrophil count × monocyte count)/lymphocyte count (Huang, 2023). PLR was calculated as platelet count/lymphocyte count (Firment and Hulin, 2024). NLR was calculated as neutrophil count/lymphocyte count (Firment and Hulin, 2024). MLR was calculated as monocyte count/lymphocyte count (Meng et al., 2018).

Complete blood count (CBC) parameters were obtained from laboratory measurements as part of the NHANES examination data. Blood specimens were collected from participants at MECs via standard venipuncture. All blood cell counts are reported in 1000 cells/μL.

2.5. Assessment of covariates

Based on prior literature (Pan et al., 2023; Zhang et al., 2024; Zhou et al., 2026), we selected a comprehensive set of potential confounders, including: age (continuous, years), race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Mexican American, Other Hispanic, Other Race), marital status (married/living with partner, widowed/divorced/separated, never married), poverty-to-income ratio (PIR; categorized as poor, nearly poor, middle income, high income), education level (below high school, high school, above high school), smoking status (never, former, now), alcohol use (never, former, mild, moderate, heavy), total physical activity (<600 metabolic equivalent of task (MET)-min/week, ≥600 MET-min/week), and hyperlipidemia (yes/no).

In sensitivity analyses, body mass index (BMI) and dietary quality were additionally considered to evaluate whether traditional adiposity measures or dietary factors influenced the observed association between BRI and depressive symptoms. BMI was calculated as weight in kilograms divided by height in meters squared (kg/m²). Dietary quality was assessed using the Healthy Eating Index-2015 (HEI-2015), a validated index measuring adherence to the 2015–2020 Dietary Guidelines for Americans (Krebs-Smith et al., 2018). The HEI-2015 total score was derived from NHANES 24-h dietary recall data and calculated as the average score from Day 1 and Day 2 dietary recalls, ranging from 0 to 100, with higher scores indicating better overall diet quality.

2.6. Statistical analysis

All analyses accounted for the complex survey design by using NHANES examination sample weights, strata, and primary sampling units, as recommended by the NCHS. Baseline characteristics were compared across BRI tertiles using weighted linear regression for continuous variables and weighted chi-square tests for categorical variables. Missing covariate data were addressed with dummy variables.

Weighted multivariable linear regression models were fitted to estimate β coefficients and 95% confidence intervals (CIs) for the association between BRI (both continuous and tertiles) and PHQ-9 total score. Three models were constructed: Model 1 adjusted for no covariates; Model 2 adjusted for age and race/ethnicity; Model 3 further adjusted for all covariates listed above.

To further examine the dose–response relationship between BRI and PHQ-9 score, a generalized additive model (GAM) with smoothed curve fitting was first employed to explore potential nonlinearity. When a nonlinear pattern was suggested, a two-piecewise linear regression model was subsequently applied to assess the threshold effect of BRI on PHQ-9 total score. The turning point was identified using a two-step recursive method that selected the inflection point yielding the maximum model likelihood, as previously described (Yu et al. 2013, 2018). Briefly, candidate turning points were initially evaluated at 5% increments from the 5th to 95th percentile of the BRI distribution, and the search interval was subsequently narrowed through recursive optimization until the value providing the highest likelihood was identified. The log-likelihood ratio test was used to compare the two-piecewise linear regression model with the conventional single-line linear model to determine whether the threshold model provided a significantly better fit. Bootstrap resampling (1000 iterations) was used to estimate the 95% CI for the turning point.

Mediation analysis was conducted using nonparametric bootstrap methods with 1000 resamples to estimate the total, direct, and indirect statistical associations, as well as the proportion of the total association accounted for by the mediator (Xu et al., 2024). Separate mediation models were run for each inflammatory marker: WBC count, SII, SIRI, PLR, NLR, and MLR. The mediation models were adjusted for the same covariates as Model 3. Notably, the results of the mediation analyses should be interpreted cautiously. Mediation analysis inherently assumes temporal ordering of exposure, mediator, and outcome—an assumption that cannot be verified in cross-sectional data. Therefore, these findings should be viewed as exploratory statistical associations rather than causal mediation effects or evidence supporting biological pathways.

To evaluate the robustness of the primary findings, several sensitivity analyses were conducted. First, subgroup analyses were performed stratified by age (18–29 years vs. 30–45 years), race/ethnicity, marital status, poverty-to-income ratio, education level, smoking status, alcohol use, physical activity, and hyperlipidemia. Interaction terms were tested by adding cross-product terms to the fully adjusted models. Second, depressive symptoms defined by a PHQ-9 score ≥10 were used as an alternative binary outcome in sensitivity analyses. Survey-weighted logistic regression models were applied to evaluate the association between BRI and clinically relevant depressive symptoms. In addition, a two-piecewise logistic regression model was performed using the turning point identified in the primary analysis to assess whether the association pattern above and below the predefined threshold was consistent with the primary analysis. Third, to assess the stability of the identified turning point, we performed an internal validation by stratifying the sample into two subsets based on NHANES survey period (2005–2010 and 2011–2018) and repeated the threshold effect analysis in each subset using the same turning point identified in the primary analysis, to evaluate whether the pattern of association was consistent across time periods. Fourth, BMI was examined as a traditional anthropometric measure, and additional models including BMI alone and both BRI and BMI simultaneously were constructed using the same covariates as the primary model to evaluate whether BRI remained associated with depressive symptoms after accounting for BMI. Fifth, to evaluate whether overall diet quality influenced the observed association, we additionally adjusted for the HEI-2015 total score.

All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant. All analyses were performed with R (version 4.2.0) and EmpowerStats (https://www.empowerstats.net/, X&Y Solutions, Inc., Boston, MA).

3. Results

3.1. Baseline characteristics

A total of 7451 women aged 18–45 years were included in the study. The weighted mean age was 32.07 ± 0.16 years, and the mean PHQ-9 score was 3.53 ± 0.06. Baseline characteristics stratified by BRI tertiles are presented in Table 1. Women in the highest BRI tertile (reflecting greater central obesity) were older, had higher WBC counts, and higher PHQ-9 scores compared with those in the lowest tertile (all P < 0.0001).

Table 1.

Weighted demographic and clinical characteristics of the study population.

Characteristics	Total	BRI tertile			P-value
		Low (1.19–3.63)	Middle (3.63–5.70)	High (5.70–20.00)
Participantsa	7451	2484	2483	2484
Age (years)	32.07 ± 0.16	29.45 ± 0.25	33.14 ± 0.23	33.86 ± 0.19	<0.0001
PHQ-9 total score	3.53 ± 0.06	3.01 ± 0.09	3.27 ± 0.12	4.39 ± 0.13	<0.0001
WBC (1000 cells/μL)	7.58 ± 0.04	6.93 ± 0.05	7.47 ± 0.06	8.40 ± 0.06	<0.0001
SII	573.73 ± 5.23	527.25 ± 6.98	570.96 ± 8.51	627.82 ± 8.87	<0.0001
SIRI	1.14 ± 0.01	1.10 ± 0.02	1.12 ± 0.02	1.21 ± 0.02	<0.0001
PLR	127.76 ± 0.73	129.21 ± 1.15	129.23 ± 1.18	124.59 ± 1.21	0.0102
NLR	2.10 ± 0.02	2.07 ± 0.02	2.09 ± 0.03	2.15 ± 0.02	0.0110
MLR	0.25 ± 0.01	0.26 ± 0.01	0.25 ± 0.01	0.23 ± 0.01	<0.0001
Race/ethnicity (%)					<0.0001
Non-Hispanic White	60.42	67.23	59.02	54.31
Non-Hispanic Black	13.39	10.30	12.08	18.20
Mexican American	10.57	5.63	12.14	14.42
Other Hispanic	6.88	5.88	7.67	7.17
Other Race	8.74	10.96	9.08	5.90
Marital status (%)					<0.0001
Married/Living with Partner	58.80	55.04	62.32	59.02
Widowed/Divorced/Separated	11.45	7.98	12.82	13.65
Never married	29.75	36.98	24.85	27.33
Poverty income ratio (%)					<0.0001
Poor	19.18	16.12	17.69	24.12
Nearly poor	21.14	18.42	20.04	25.28
Middle income	29.07	28.00	29.94	29.35
High income	30.61	37.46	32.33	21.25
Education level (%)					<0.0001
Below high school	3.32	1.02	4.51	4.61
High school	30.87	27.19	28.68	37.27
Above high school	65.82	71.79	66.81	58.12
Smoking status (%)					0.0046
Never	65.03	68.30	64.61	61.98
Former	13.31	12.76	13.29	13.93
Now	21.66	18.95	22.10	24.09
Alcohol use (%)					<0.0001
Never	13.02	11.69	13.42	14.06
Former	7.30	4.15	7.60	10.41
Mild	25.31	27.39	23.67	24.77
Moderate	26.18	28.65	27.11	22.49
Heavy	28.19	28.12	28.20	28.26
Physical activity (MET/week) (%)					0.0351
<600	23.80	21.42	24.78	25.69
≥600	76.20	78.58	75.22	74.31
Hyperlipidemia (%)					<0.0001
No	46.53	66.36	45.27	25.77
Yes	53.47	33.64	54.73	74.23

BRI, body roundness index; MLR, Monocyte-to-Lymphocyte Ratio; NLR, Neutrophil-to-Lymphocyte Ratio; PHQ-9, the Patient Health Questionnaire-9; PLR, Platelet-to-Lymphocyte Ratio; SII, Systemic Immune-Inflammation Index; SIRI, Systemic Inflammation Response Index; WBC, White Blood Cell.

Continuous variables are presented as weighted mean ± SE, and P-values were derived from weighted linear regression. Categorical variables are presented as weighted percentage, and P-values were derived from weighted chi-square tests.

Among the 7451 participants, the amount of missing values for the covariates were 730 (9.8%) for marital status, 533 (7.1%) for poverty income ratio, 3 (0.1%) for education level, 583 (7.8%) for smoking status, 543 (7.3%) for alcohol use, and 1560 (20.9%) for total physical activity.

^aUnweighted number of observations in dataset.

3.2. Association between BRI and depressive symptoms

Weighted multivariable linear regression showed a positive association between BRI and PHQ-9 scores (Table 2). In the fully adjusted model (Model 3), each one-unit increase in BRI was associated with a 0.18-point increase in PHQ-9 score (β = 0.18, 95% CI: 0.12, 0.23; P < 0.0001). When BRI was analyzed by tertiles, women in the highest tertile had significantly higher PHQ-9 scores compared to those in the lowest tertile (β = 0.99, 95% CI: 0.64, 1.35; P < 0.0001), with a significant trend across tertiles (P for trend < 0.0001). Compared with women in the lowest BRI tertile, those in the highest tertile had nearly 1-point higher PHQ-9 scores, indicating a greater depressive symptom burden at the population level. Although the absolute difference in PHQ-9 scores was modest at the individual level, the higher symptom burden observed among women with higher BRI suggests a potentially meaningful population-level association.

Table 2.

Weighted linear regression results for association between BRI and PHQ-9 total score.

	Model 1		Model 2		Model 3
	β (95% CI)	P-value	β (95% CI)	P-value	β (95% CI)	P-value
BRI	0.24 (0.19, 0.29)	<0.0001	0.25 (0.20, 0.30)	<0.0001	0.18 (0.12, 0.23)	<0.0001
BRI tertile
Low (1.19–3.63)	Ref.		Ref.		Ref.
Middle (3.63–5.70)	0.26 (−0.04, 0.55)	0.0885	0.32 (0.01, 0.62)	0.0426	0.11 (−0.19, 0.42)	0.4732
High (5.70–20.00)	1.38 (1.07, 1.70)	<0.0001	1.45 (1.12, 1.77)	<0.0001	0.99 (0.64, 1.35)	<0.0001
P for trend		<0.0001		<0.0001		<0.0001

BRI, body roundness index; PHQ-9, the Patient Health Questionnaire-9.

Model 1 was adjusted for no covariates.

Model 2 was adjusted for age and ethnicity.

Model 3 was adjusted for age, ethnicity, marital status, poverty income ratio, education level, smoking status, alcohol use, hyperlipidemia, and physical activity.

3.3. Nonlinear association and threshold effect

A nonlinear relationship was observed between BRI and PHQ-9 total score (Fig. 2). Threshold effect analysis further confirmed this nonlinear pattern (Table 3). The turning point was identified at approximately BRI = 3.85 (95% CI for the turning point: 3.38, 4.08). Below this threshold, BRI was not significantly associated with PHQ-9 score (β = −0.06, 95% CI: −0.25, 0.14; P = 0.5711). Above the threshold, the association was significant (β = 0.20, 95% CI: 0.15, 0.25; P < 0.0001). The log-likelihood ratio test indicated that the two-piecewise model fit significantly better than the standard linear model (P = 0.020). The observed nonlinear association suggests that the relationship between BRI and depressive symptoms may not be uniform across the BRI range. Women with higher central adiposity, particularly those above the identified threshold, may represent a subgroup with a relatively greater burden of depressive symptoms. However, this turning point should be interpreted as an exploratory statistical finding rather than a clinically established cutoff.

Fig. 2.

Dose-response relationship between BRI and PHQ-9 total score. The black vertical tick marks along the horizontal axis represent the distribution of BRI values. The solid red line represents the smooth curve fitting between BRI and PHQ-9 total score, and the blue bands represent the 95% CI of the fitted curve. The model was adjusted for age, ethnicity, marital status, poverty income ratio, education level, smoking status, alcohol use, hyperlipidemia, and physical activity. The turning point identified by two-piecewise linear regression analysis (BRI ≈ 3.85) is shown and should be interpreted as an exploratory statistical finding rather than a clinical cutoff. BRI, body roundness index; PHQ-9, the Patient Health Questionnaire-9. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Table 3.

Threshold effect analysis of BRI on PHQ-9 total score with NHANES Sampling Weights.

BRI	β (95% CI)	P-value
Fitting by standard linear model	0.17 (0.11, 0.22)	<0.0001
Fitting by two-piecewise linear model
turning point	3.85
≤3.85	−0.06 (−0.25, 0.14)	0.5711
>3.85	0.20 (0.15, 0.25)	<0.0001
95% CI for turning point	3.38, 4.08
Log-likelihood ratio	0.020

BRI, body roundness index; PHQ-9, the Patient Health Questionnaire-9; NHANES, National Health and Nutrition Examination Survey.

The model was adjusted for age, ethnicity, marital status, poverty income ratio, education level, smoking status, alcohol use, hyperlipidemia, and physical activity.

β and 95% CI: from a weighted two-piecewise linear regression; CIs via weighted maximum likelihood (normal approximation).

3.4. Comparative mediation analysis of inflammatory markers

Mediation analysis was conducted to examine the extent to which the association between BRI and PHQ-9 score could be statistically accounted for by inflammatory markers. After adjusting for all covariates, the statistical association between BRI and PHQ-9 was significantly accounted for by white blood cell count in the mediation model (Fig. 3). The indirect effect through WBC was 0.05 (95% CI: 0.01, 0.10; P = 0.024), accounting for 9.3% (95% CI: 1.1%, 18.6%) of the total effect. Other inflammatory markers, including SII, SIRI, PLR, NLR, and MLR, did not show significant mediation effects (all P > 0.05; see Table S1 for detailed estimates). The direct effect of BRI on PHQ-9 score remained significant after adjusting for WBC (β = 0.50, 95% CI: 0.35, 0.66; P < 0.0001), indicating that WBC partially accounted for the observed association in this statistical model. Overall, among the six inflammatory markers tested, only WBC showed a statistically significant indirect association; the remaining five markers did not, suggesting only partial support for the hypothesis that systemic inflammation accounts for the BRI–depressive symptoms relationship in these cross-sectional models. Because mediation analysis assumes a temporal sequence that cannot be verified with cross-sectional data, the results should be interpreted as statistical associations consistent with mediation rather than as evidence of causal pathways.

Fig. 3.

The mediating effects of WBC on the relationship between BRI and PHQ-9 total score. Mediation estimates represent statistical associations consistent with mediation and do not imply causality due to the cross-sectional study design. Models were adjusted for age, ethnicity, marital status, poverty income ratio, education level, smoking status, alcohol use, hyperlipidemia, and physical activity. BRI, body roundness index; PHQ-9, the Patient Health Questionnaire-9; WBC, White Blood Cell.

3.5. Sensitivity analyses

Subgroup analyses were performed to assess whether the association between BRI and PHQ-9 score varied across different population characteristics (Fig. 4). The association was generally consistent across most subgroups. Given the potential biological differences across the reproductive age span, we further examined age-stratified associations. The association between BRI and PHQ-9 score appeared stronger among women aged 30–45 years (β = 0.18, 95% CI: 0.12, 0.24; P < 0.0001) than among those aged 18–29 years (β = 0.08, 95% CI: 0.00, 0.17; P = 0.0410), although the interaction did not reach conventional statistical significance (P for interaction = 0.0507). No significant interactions were observed for marital status, poverty income ratio, education, smoking, alcohol use, hyperlipidemia, or physical activity (all P for interaction > 0.05). A marginally significant interaction was observed for race/ethnicity (P for interaction = 0.0486); the association was positive in all racial/ethnic groups except Non-Hispanic Black women, where it was not significant (β = 0.05, P = 0.2838). Given the exploratory nature of the subgroup analyses and the potential impact of multiple comparisons, the observed interactions should be interpreted cautiously and require validation in future studies.

Fig. 4.

Forest plot of subgroup analyses for the association between BRI and PHQ-9 total score. Age, ethnicity, marital status, poverty income ratio, education level, smoking status, alcohol use, hyperlipidemia, and physical activity were adjusted, except the stratifying variable itself. BRI, body roundness index; PHQ-9, the Patient Health Questionnaire-9.

Using clinically relevant depressive symptoms defined by PHQ-9 score ≥10 as an alternative outcome, the association between BRI and depressive symptoms remained consistent with the primary analysis. Higher BRI was associated with increased odds of clinically relevant depressive symptoms (OR = 1.08, 95% CI: 1.04, 1.11; P = 0.0001), and women in the highest BRI tertile had significantly higher odds compared with those in the lowest tertile (OR = 1.73, 95% CI: 1.32, 2.26; P = 0.0001). Furthermore, the two-piecewise logistic regression model using the predefined turning point (BRI = 3.85) showed a positive association above the threshold, consistent with the primary analysis, further supporting the robustness of the observed association across different definitions of depressive symptoms (Table S2).

To evaluate the consistency of the identified nonlinear pattern, we stratified the sample by NHANES survey period (2005–2010 and 2011–2018). Using the same turning point identified in the primary analysis (BRI = 3.85), the association pattern was consistent across both subsets: no significant association below the threshold, and a significant positive association above the threshold (2005–2010: β = 0.20, 95% CI: 0.11, 0.28; P < 0.0001; 2011–2018: β = 0.21, 95% CI: 0.14, 0.27; P < 0.0001) (Tables S3 and S4). However, the log-likelihood ratio tests did not reach statistical significance in either subset, likely reflecting reduced statistical power due to smaller sample sizes.

When comparing BRI with BMI, BMI was positively associated with PHQ-9 total scores in the BMI-only model (β = 0.05, 95% CI: 0.03, 0.07; P < 0.0001). However, when BRI and BMI were simultaneously included in the same model, BRI remained significantly associated with PHQ-9 scores (β = 0.29, 95% CI: 0.14, 0.43; P < 0.001), whereas the association for BMI was attenuated and no longer statistically significant (P = 0.127). The results of these analyses are presented in Table S5.

After additionally adjusting for the HEI-2015 total score, the association between BRI and PHQ-9 scores remained essentially unchanged (continuous BRI: β = 0.18, 95% CI: 0.12, 0.23; P < 0.0001; highest vs. lowest BRI tertile: β = 0.97, 95% CI: 0.61, 1.33; P < 0.0001; Table S6). These findings indicate that the observed association was robust to adjustment for overall diet quality.

4. Discussion

In this large, nationally representative sample of 7451 U.S. women of reproductive age, we observed a nonlinear association between BRI and depressive symptoms. Threshold effect analysis revealed a turning point at approximately BRI = 3.85, below which no significant association was found, whereas above this value, higher BRI was associated with increased PHQ-9 scores. Furthermore, we explored potential inflammatory mechanisms and found that white blood cell count was identified as a significant statistical mediator, partially accounting for 9.3% of the total observed association. In contrast, other inflammatory indices (SII, SIRI, PLR, NLR, MLR) showed no significant statistical mediation. This pattern provides only partial support for the inflammatory hypothesis, and if these statistical associations reflect underlying biological processes—which cannot be confirmed due to the cross-sectional design—the relevant inflammatory component may be limited to overall leukocyte burden rather than representing a broad systemic inflammatory response.

Previous studies have established a close relationship between BRI and adverse mental health outcomes. Zhang et al. reported a positive linear association between BRI and depressive symptoms in general adults (Zhang et al., 2024). Zhou et al. further found that SII partially mediated this association, with a mediation proportion of approximately 2.55% (Zhou et al., 2026). Ye et al. observed a nonlinear threshold pattern between BRI and suicidal ideation, with a turning point at approximately 6.7 (Ye et al., 2025). Our study shares some similarities with these findings: we also observed a positive association between BRI and depressive symptoms. However, our study differs in several important aspects. First, unlike the linear pattern reported by Zhang and Zhou, we observed a nonlinear threshold association among women of reproductive age, with a turning point at approximately BRI = 3.85. Furthermore, in sensitivity analyses, BRI remained significantly associated with depressive symptoms after further adjustment for BMI, whereas the association of BMI was attenuated and no longer statistically significant, suggesting that BRI may capture additional aspects of body shape and central adiposity beyond general body size. This difference may relate to the specific hormonal and metabolic characteristics of women of reproductive age, which could influence the association between visceral adiposity and mood. We therefore recommend interpreting the findings as associations rather than causal effects: the magnitude of the association between BRI and depressive symptoms may not be uniform; in this cohort there appears to be a threshold effect where no clear adverse association is observed at lower BRI levels, but the association strengthens above that threshold. Second, we found that WBC showed a pattern consistent with partial statistical mediation of the BRI–PHQ-9 association (indirect effect proportion ≈ 9.3%), while other composite inflammatory indices did not show significant mediation. Of note, the absolute magnitude of the indirect statistical association attributed to WBC was modest (9.3%), with a wide 95% CI (1.1%–18.6%) whose lower bound was close to zero. While this proportion was larger than the 2.55% reported for SII by Zhou et al. in a general adult sample (Zhou et al., 2026), it should be interpreted cautiously. A proportion of 9.3% implies that, even if a causal pathway exists, the majority of the statistical association between BRI and PHQ-9 scores (>90%) would be accounted for by other factors not captured by WBC, such as dysregulation of the hypothalamic–pituitary–adrenal (HPA) axis, health behaviors, or psychosocial mechanisms. Nonetheless, the choice of inflammatory marker likely contributes to the difference: WBC is a direct measure of total leukocyte burden, which captures the chronic low-grade inflammation driven by visceral adiposity, whereas composite indices like SII may be more sensitive to acute stress or infection and could be influenced by the broader age range and sex distribution in previous studies (Zhou et al., 2026). These findings suggest that total WBC may be a more informative statistical mediator than derived inflammatory ratios in the study of BRI and depressive symptoms in women of reproductive age.

The observed statistical association between BRI and depressive symptoms is biologically plausible. It should be emphasized that the following biological interpretations are hypothetical and were not directly assessed in the present cross-sectional study. Visceral adipose tissue—which BRI specifically captures—is metabolically active and secretes pro-inflammatory adipokines such as interleukin-6 and tumor necrosis factor-α (Monsalve et al., 2025; Oracz et al., 2025). Previous studies have hypothesized that these inflammatory signals may influence the HPA axis, neurotransmitter systems, and neuroinflammatory processes, thereby potentially contributing to depressive symptoms (Hole et al., 2025; Sălcudean et al., 2025). The observed threshold pattern may be consistent with the possibility that adiposity-related biological changes become more relevant beyond a certain level of visceral fat accumulation. Below this level, the biological burden associated with excess adiposity may be insufficient to manifest as measurable differences in depressive symptoms; above it, these changes may become more evident. This nonlinear pattern is compatible with the concept of a biological “tipping point,” although the underlying mechanisms remain hypothetical and require confirmation in longitudinal studies (Xie et al., 2025; Xie, 2026). Furthermore, the finding that WBC—but not composite indices—showed a significant indirect statistical association in cross-sectional models may generate hypotheses for future mechanistic studies. However, the assumptions required for mediation analysis, including correct temporal ordering and absence of unmeasured confounding, cannot be verified in the present study. WBC is a direct measure of total leukocyte burden, reflecting the overall inflammatory load. In contrast, composite indices such as SII, SIRI, PLR, NLR, and MLR incorporate platelet or lymphocyte counts, which may fluctuate with acute stress or infection and do not consistently reflect the chronic, low-grade inflammation associated with visceral adiposity in this otherwise healthy reproductive-aged population (Liu et al., 2024; Rengasamy et al., 2022; van den Munckhof et al., 2024). The lack of statistical mediation by composite inflammatory indices in these cross-sectional models may be explained by their dependence on specific leukocyte subtypes. For instance, SII and SIRI involve neutrophil and monocyte counts, while PLR, NLR, and MLR are ratios that can be influenced by changes in lymphocyte counts. In the context of visceral obesity-related inflammation, previous studies have suggested that total leukocyte burden may increase without necessarily altering the balance between leukocyte subtypes, suggesting that total WBC may represent a relatively stable correlate compared with derived ratios in this setting (Liu et al., 2024; van den Munckhof et al., 2024). Total leukocyte burden has been shown to capture the chronic inflammatory state associated with visceral fat, supporting its potential relevance as a marker in this context (van den Munckhof et al., 2024). Consistent with Rengasamy et al.'s finding in the context of physical illness, our study extends the observed statistical mediating role of WBC to the context of visceral adiposity, suggesting that total leukocyte burden is a consistent inflammatory correlate of depressive symptoms. Together, these findings support the hypothesis that visceral adiposity is associated with depressive symptoms and suggest that leukocyte-related inflammatory processes may represent one potential pathway underlying this association. However, these findings should be interpreted as hypothesis-generating, as the cross-sectional design prevents conclusions regarding causality or temporal sequence (Liu et al., 2024; Rengasamy et al., 2022) However, the cross-sectional design cannot establish the directionality or temporality of these associations.

These findings have several important implications, though they must be interpreted within the limitations of a cross-sectional study. First, by focusing on women of reproductive age—a group with unique hormonal and metabolic characteristics—we address a previously understudied population at high risk for both obesity and depressive symptoms. Although women aged 18–45 years encompass diverse biological stages, our subgroup analysis did not identify statistically significant age-related effect modification. However, the association appeared stronger among women aged 30–45 years than among those aged 18–29 years, although the interaction did not reach conventional statistical significance. This pattern may reflect age-related differences in metabolic burden, cumulative exposure to visceral adiposity, hormonal fluctuations, or psychosocial stressors across the reproductive life span. Nevertheless, given the borderline interaction P value and the exploratory nature of subgroup analyses, these findings should be interpreted cautiously and require validation in studies specifically designed to evaluate age-related effect modification. In addition, the potential clinical and public health implications of the observed association should be considered. Although the absolute difference in PHQ-9 scores associated with higher BRI levels was modest at the individual level, even relatively small shifts in depressive symptom burden may have public health relevance when observed across a large population. In particular, the identification of women with higher BRI and greater depressive symptom burden may help inform targeted assessment and preventive strategies. Second, the identification of a turning point suggests that the association between BRI and depressive symptoms may not be uniform across the BRI range. The consistency of the association pattern when applying the same turning point in two temporally independent NHANES subsets provides some support for the robustness of this statistical finding. This nonlinear pattern highlights the potential heterogeneity of the BRI–depressive symptoms association rather than assuming a simple linear relationship. Notably, this turning point should be interpreted as a change point in the statistical association rather than a biological threshold indicating the onset of increased depressive symptoms. Until externally validated in independent cohorts, this value should be considered an exploratory, hypothesis-generating finding rather than a clinically meaningful cutoff. Third, our comparative mediation analysis suggests that total leukocyte burden (WBC) is a more relevant statistical mediator than composite indices in this population, assuming the temporal order implied by the model. Clinically, BRI may help identify women who could be at higher risk for depressive symptoms. From a public health perspective, integrating BRI into routine checkups for women of reproductive age could enable early risk detection and align with the emerging field of “metabolic psychiatry.”

Several strengths of this study warrant mention. First, we used a large, nationally representative sample of U.S. women of reproductive age, which enhances the generalizability of our findings. Second, BRI was calculated using standardized anthropometric measurements, and depressive symptoms were assessed with the well-validated PHQ-9 questionnaire. Third, our analytical approach was robust: we explicitly tested for and identified a nonlinear, threshold-dependent relationship using two-piecewise linear regression and a likelihood ratio test. Fourth, we performed a comparative mediation analysis on six inflammatory markers in cross-sectional models, which allowed us to identify WBC as a significant statistical mediator while other composite indices showed no significant indirect statistical associations. These methodological strengths increase confidence in our findings.

Some limitations of this study merit consideration. First, the cross-sectional design precludes causal inference; bidirectional relationships between visceral adiposity, inflammation, and depressive symptoms cannot be ruled out. Second, depressive symptoms were assessed using the self-reported PHQ-9 questionnaire rather than a structured clinical interview, which may be subject to recall bias and social desirability bias. Nonetheless, the PHQ-9 is a well-validated screening tool with high sensitivity and specificity for detecting depressive symptoms in large epidemiological studies. Third, WBC and other inflammatory markers were measured only once, which may not fully capture chronic, long-term inflammatory status and may have attenuated the estimated indirect statistical associations. However, single measurements of blood cell counts are commonly used in large epidemiological studies and have been shown to be relatively stable over time. Fourth, residual confounding by unmeasured factors (e.g., history of depression, medication use, social support, etc.) cannot be excluded. Although sensitivity analyses adjusting for overall diet quality using the HEI-2015 score did not materially alter the findings, residual confounding by specific dietary components not fully captured by the HEI-2015 composite score cannot be excluded. Fifth, mediation analyses rely on assumptions regarding temporal ordering and absence of unmeasured confounding, which cannot be fully verified in cross-sectional data; therefore, the observed indirect associations should be interpreted cautiously. The 95% CI lower bound for the indirect effect was already close to zero, suggesting that even modest unmeasured confounding could attenuate this statistical association. Notably, when stratified by survey period, the direction and approximate magnitude of the WBC indirect effect were consistent across subsets, providing some qualitative support for robustness. Sixth, although our analyses were guided by predefined hypotheses, the study involved multiple statistical evaluations, including subgroup interaction and mediation analyses, which may increase the possibility of type I error. Therefore, findings from these exploratory analyses should be interpreted cautiously and require confirmation in future studies. Seventh, CRP, another well-established inflammatory marker, was not included because its measurement cycles did not align with our study period; consequently, the indirect statistical association through WBC could not be compared with that through CRP. Nonetheless, WBC has been shown to be a stable and reliable marker of systemic inflammation in large epidemiological studies. Additionally, pregnant women and women within 12 months postpartum were excluded to minimize potential confounding from pregnancy-related physiological changes in adiposity, inflammation, and depressive symptoms. Therefore, the findings may not be directly generalizable to pregnant or early postpartum populations. Despite these limitations, the present findings provide an epidemiological basis for future longitudinal studies to clarify the temporal relationships among BRI, inflammatory markers, and depressive symptoms, as well as intervention studies to evaluate whether targeting visceral adiposity may influence depressive symptom burden.

In conclusion, in this nationally representative study of U.S. women of reproductive age, we observed a nonlinear, threshold-dependent association between BRI and depressive symptoms, with a significant positive association emerging only after BRI exceeded a certain point. In mediation analysis, white blood cell count statistically accounted for 9.3% of the observed association, whereas other inflammatory markers (SII, SIRI, PLR, NLR, MLR) did not significantly account for the association. These findings are consistent with the hypothesis that excess visceral adiposity is associated with depressive symptoms. Among the inflammatory markers examined, only total leukocyte count showed a modest indirect statistical association, suggesting a potential contribution of leukocyte-related inflammatory processes to the observed association. These results generate testable hypotheses for future longitudinal studies to examine the temporal direction of the observed associations. Intervention trials are warranted to determine whether reducing visceral adiposity or targeting systemic inflammation can improve depressive symptoms among women with high BRI.

CRediT authorship contribution statement

Li Zhang: Writing – original draft, Methodology, Formal analysis, Conceptualization. Xianmei Ma: Writing – original draft, Visualization, Formal analysis, Data curation. Gaoyan Shi: Writing – original draft, Software, Data curation. Yanhua Liu: Writing – review & editing, Supervision, Project administration, Conceptualization.

Ethics approval and consent to participate

The ethical protocol for NHANES was approved by the National Center for Health Statistics (NCHS) Ethics Review Board (https://www.cdc.gov/nchs/nhanes/about/erb.html). Written informed consent was obtained from all participants prior to data collection. All personal identifiers in the NHANES dataset have been removed, and the data were anonymized before being made publicly available. Since this study is based on a secondary analysis of publicly available anonymized NHANES data, no additional ethical approval was required.

Funding

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

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.

Appendix A. Supplementary data

The following is the Supplementary data to this article.

Data availability

All NHANES data for this study are publicly available online (https://www.cdc.gov/nchs/nhanes/).