Exploring the mediating roles of diet and physical activity in the association between different obesity phenotypes and risk of depression: A cox survival analysis approach

Abstract

Although obesity has been linked to an increased risk of depression, the risk associated with different obesity phenotypes remain unclear, as does the role of lifestyle behaviors. Participants from the UK Biobank who were free of depression at the baseline and had complete data on body composition and metabolism-related measures were selected and followed up. Multivariable Cox models were employed to assess the risk of developing depression according to obesity phenotypes. Mediation analyses were conducted to evaluate the potential mediating effects of diet and physical activity in this association. Among 391,781 participants, a total of 20,027 incident depression was recorded after a median follow-up period of 13.39 years. Our study revealed that different obesity phenotypes increased the risk of depression to varying degrees. Compared with healthy non-obese individuals, those with BMI ≥30 kg/m² only had the lowest risk of depression (HR: 1.15, 95 %CI: 1.04, 1.28), and then followed by central obesity (HR: 1.60, 95 %CI: 1.51, 1.70), metabolically unhealthy obesity (HR: 1.63, 95 %CI: 1.56, 1.70). Sarcopenic obesity results in the highest risk of depression among these four phenotypes (HR: 1.86, 95 %CI: 1.40, 2.46). Both physical activity and diet mediated the effect of the four obesity phenotypes on the risk of depression. In the relationship between BMI ≥30 only and depression, physical activity mediated for 2.49 % and diet mediated for 1.60 %. Between metabolically unhealthy obesity and depression, physical activity mediated for 7.08 % and diet mediated for 1.86 %. This set of data was 5.29 % and 1.47 % in the relationship between central obesity and depression, and 6.22 % and 1.25 % in the relationship between sarcopenic obesity and depression. Subgroup analyses revealed that the female with obesity (including four types) had a higher risk of depression compared to the male, whereas individuals aged 60 years or older and those who were former or current smokers with obesity (except sarcopenic obesity) exhibited a lower risk of depression. These findings provide a basis for the prevention of obesity-depression comorbidity in obese patients.

Highlights

• •Obesity were associated with risk of developing depression no matter which phenotypes.
• •Obesity with only BMI ≥30 kg/m² increase risk of depression lightly, and sarcopenic obesity people showed the highest risk.
• •Diet and physical activity play a mediating role in the relationship between obesity and depression.
• •Age, sex and smoking have an interaction effect on the incidence of depression caused by different obesity phenotypes.

1. Introduction

Recent statistics indicate that among the 25 leading Level 3 causes, age-standardized disability adjusted life year (DALY) rates increased most substantially for anxiety disorders (16.7 %) and depression (16.4 %) [1]. Without timely and appropriate intervention, mild and moderate depression may progress to severe depression, thereby increasing the risks of disability and suicide. The etiology of depression is complex, and various chronic diseases, such as obesity, type 2 diabetes, cardiovascular diseases (CVD), and cognitive impairment, are believed to be associated with the onset and progression of depression [[2], [3], [4], [5]].

Obesity has emerged as a critical global public health challenge. Over the past four decades, the worldwide prevalence of overweight and obesity has nearly tripled [6]. Beyond the physical discomfort associated with obesity, it also serves as a significant risk factor for numerous chronic diseases, including cardiovascular disease, hypertension, diabetes, stroke and mental health issues such as anxiety and depression [5,7]. While the global prevalence of obesity has tripled over recent decades, there has also been a parallel rise in depression prevalence. This simultaneous upward trend has attracted significant scholarly attention to the potential association between obesity and depression. From an emotional and psychological perspective, individuals with obesity are more likely to experience body dissatisfaction or engage in restrictive dieting, which may contribute to heightened vulnerability to mental health issues and emotional distress [8,9].

Epidemiological evidence linking obesity and depression has been gradually accumulating recent years, but the results are inconsistent. Several cross-sectional studies have demonstrated an association between obesity or overweight and the prevalence of depression, not only in adults [10], but also in adolescents [11]. A meta-analysis further indicated that baseline obesity increases the risk of subsequent depression [5]. One prospective study reported that, among patients with depression, obesity was more strongly associated with specific symptoms (such as low mood, loss of interest, and decreased energy) than with the overall severity of depression [12]. However, another longitudinal study conducted on the Chinese population revealed that obese men exhibited a reduced risk of developing depression [13]. A mendelian randomization study by Hung et al. also found no evidence supporting a causal relationship between higher body mass index (BMI) and major depressive disorder [14].

It is worth noting that most previous studies have used BMI as the sole indicator of obesity without considering more nuanced aspects of body composition or metabolic health. In particular, the weight component of BMI cannot distinguish between muscle mass and fat mass. Research indicates that muscle loss or internal substance changes in muscle cells can lead to changes in depression risk [15]. Muscle and adipose tissue also possess endocrine function. Therefore, different muscle and fat distribution will cause different results on endocrine, thereby affecting the occurrence and development of depression. Failure to account for these factors may lead to phenomena such as obesity paradox, obscuring the true association between obesity and related outcomes [12,16]. In recent years, studies have gradually explored the association between sarcopenic obesity or metabolically unhealthy obesity and depression risk [[17], [18], [19]]. However, most of these studies were cross-sectional studies and discuss the impact of a single phenotype on depression in a single population. Due to their different reference groups and the control of confounding factors, it is difficult to compare the heterogeneity of depression risk caused by different obesity phenotypes making it difficult to help obese patients prioritize intervention programs such as prioritizing weight loss, managing metabolic syndrome or strengthening muscle.

Obesity may influence depression risk through changes in emotional states mediated by alterations in diet and physical activity during subsequent weight control. Previous research on depression has accumulated substantial evidence indicating that diet and physical activity can affect the onset of depression [20,21], and play crucial mediating roles in its development [22]. However, the potential mediating effects of these obesity-related factors in the relationship between obesity and depression have been rarely explored. In fact, obese individuals are highly likely to recognize the dangers of obesity and improve their dietary habits and physical activity frequency. Therefore, the mediating role of these two factors in whether obese individuals develop depression warrant further investigation.

By utilizing the comprehensive community information and cohort follow-up data from the UK Biobank (UKB), the aim of this study is to explore the difference in the risk of depression between various obesity phenotypes. According to the presence or absence of the common comorbidities such as abnormal waist circumference, sarcopenia and metabolic syndrome, the obese participants were divided into five types: obesity with only BMI ≥30 kg/m², central obesity, metabolically unhealthy obesity, sarcopenic obesity and multiple obesity phenotypes, and we focus on the first four phenotypes. Although most previous studies have shown that obesity with comorbidities increases the risk of various chronic diseases, the aim of this study was to explore the heterogeneity of these obesity phenotypes, so as to provide priority of self-examination and intervention for obese patients. Furthermore, mediation analysis and subgroup analysis will be conducted to elucidate the mediating role of dietary habits and physical activity in the obesity-depression relationship, as well as to examine the strength of this association within diverse population groups.

2. Methods

2.1. Data source and study subjects

The UK Biobank is a large-scale, population-based prospective cohort study involving over 500,000 participants aged 40–69 years from England, Scotland, and Ireland, who were enrolled between 2006 and 2010. At baseline, basic demographic data, lifestyle factors, medical history, and genetic information were collected through touchscreen questionnaires, physical examinations, and biospecimen sampling. Additional questionnaires and assessments were administered during follow-up periods. Participant health outcomes were systematically recorded using integrated data sources, including primary care records, hospital inpatient reports, and mortality registries.

As shown in Fig. 1, of the 502,401 participants initially recruited at baseline, 5733 were underweight participants (BMI <18.5 kg/m²), and these participants were excluded from the analysis. Participants who were diagnosed with depression prior to baseline were therefore excluded from the analysis and the number of them was 3598. Additionally, participants whose information related to obesity phenotypes was missing data were excluded (n = 101,289). Based on the predefined exclusion criteria, a total of 391,781 participants were ultimately included in the study for the main analysis. When conducting the mediation analysis, 74,604 individuals had missing data on physical activity, and so did 16,153 individuals on diet. Therefore, the total number of individuals included in the mediation analysis was 317,177 (with physical activity as the mediator) and 375,628 (with diet as the mediator). The participants included were followed up for 13.39 years.

Fig. 1.

Participant flowchart.

2.2. Assessment of obesity and sarcopenia

BMI was calculated based on baseline height and weight measurements obtained by trained personnel using calibrated equipment. Height was measured with Seca 202 stadiometer (precision 0.1 cm), and weight was determined using a Tanita BC-418MA body composition analyzer (precision 0.1 kg). The BMI was calculated using the formula $BMI=weight\left(kg\right)/{height\left(m\right)}^2$. According to the World Health Organization (WHO) criteria, BMI of 30 kg/m² or higher is classified as obesity [23]. Waist circumference (WC) was manually measured in centimeters by trained personnel.

According to the criteria established by European Working Group on Sarcopenia in Older People 2 (EWGSOP2), sarcopenia is characterized by the loss of both muscle mass (adjusting Appendicular Skeletal Muscle Mass (ASM) by body size, $ASM\left(kg\right)/{height\left(m\right)}^2$) and muscle strength. Specifically, sarcopenia is defined as a muscle mass index below 7.0 kg/m² and grip strength less than 27 kg in male, and below 5.5 kg/m² with grip strength less than 16 kg in female [24]. Baseline handgrip strength was assessed using the Jamar J00105 hydraulic hand dynamometer within the UKB. The average value of measurements from both hands was used as the final handgrip strength result. Participant bioimpedance values were obtained using the Tanita BC-418MA body composition analyzer. One commonly used equation for estimating skeletal muscle mass is the Janssen equation developed by Janssen et al. [25], which in fact calculates total body skeletal muscle mass based on resistance values derived from bioelectrical impedance analysis (BIA). Later, Kyle et al. proposed an alternative formula specifically for calculating appendicular skeletal muscle mass [26]. However, Kyle's formula requires separate resistance and reactance values, whereas the BIA data collected in the UKB only provides combined impedance values without distinguishing between resistance and reactance. As a result, it is not feasible to calculate appendicular skeletal muscle mass using Kyle's formula in this study. Therefore, according to the EWGSOP2 criteria, grip strength alone was used to define "Sarcopenia Probable".

2.3. Metabolic syndrome definition

According to the criteria established by the Adult Treatment Panel III (ATP), an individual can be classified as having metabolic syndrome if at least three out of the following five indicators are met [27]: (1) absence of abdominal obesity, defined as waist circumference (WC) > 102 cm in male and >88 cm in female; (2) blood pressure ≥130/85 mmHg; (3) fasting plasma glucose (FPG) ≥ 6.1 mmol/L or with the use of antidiabetic medications; (4) triglycerides ≥1.7 mmol/L or with the use of lipid-lowering agents; and (5) high-density lipoprotein cholesterol (HDL-C) < 1.03 mmol/L in male and <1.30 mmol/L in female.

2.4. Definition of exposure and outcome

Exposures were categorized into different obesity phenotypes, which were defined based on the following criteria: (1) Healthy non-obese: 18.5 kg/m² ≤ BMI <30 kg/m², and having no abnormal WC, sarcopenia, nor metabolic syndrome; (2) Unhealthy non-obese: 18.5 kg/m² ≤ BMI <30 kg/m², and having abnormal WC, sarcopenia, or metabolic syndrome; (3) BMI ≥30 only: BMI ≥30 kg/m² and without the above three symptoms; (4) Central obesity: BMI ≥30 kg/m² combined with WC > 102 cm in male and >88 cm in female; (5) Sarcopenic obesity: BMI ≥30 kg/m² accompanied by "Sarcopenia Probable"; (6) Metabolically unhealthy obesity: BMI ≥30 kg/m² with metabolic syndrome; (7) Multiple obesity phenotypes: having two or more obesity phenotypes.

The outcome of this study was the occurrence of depression cases during the follow-up period, which were identified using hospital inpatient reports from UKB. Incident depression was defined according to the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10), including the following codes: F32.0, F32.1, F32.2, F32.3, F32.8, F32.9, and F33. The cutoff dates for hospital admission data were October 31, 2022 in England, July 31, 2021 in Scotland, and February 28, 2018 in Wales. Accordingly, the follow-up period for each participant was calculated from the baseline date to the date of the first depressive event, death, loss to follow-up, or the last available follow-up date, whichever occurred first.

2.5. Measurement of covariates and mediation variables

Following variables were selected as covariates to control for potential confounding factors (collected at baseline): (1) Sex: Male or Female; (2) Age: age at baseline recruitment; (3) Race: categorized as White or Non-white; (4) Education qualifications: No qualifications, Not college/university degree, and College or university degree; (5) Smoking status: Non-smokers, Ex-smokers, and Current smokers; (6) Drinking status: Never, Occasionally, 1–2 times per week, 3–4 times per week, and Daily or almost daily; (7) Family history of depression: presence or absence of a family history of depression; (8) History of diabetes: presence or absence of diabetes; (9) History of hypertension: presence or absence of hypertension.

The type, frequency, and quantity of food consumed by participants were collected and defined using a touchscreen food frequency questionnaire (FFQ). The frequency categories for meat and fish consumption were recoded as follows: "Never" = 0, "Less than once a week" = 0.5, "Once a week" = 1, "2–4 times a week" = 3, "5–6 times a week" = 5.5, and "Once a day or more" = 7. The frequencies of beef, lamb/mutton, and pork consumption were summed to determine the overall frequency of unprocessed red meat intake. A healthy diet was defined based on a Healthy Diet Score (HDS), which was calculated using the following dietary components [28]: at least four tablespoons of vegetables per day, at least three servings of fruit per day, fish consumption at least twice per week, unprocessed red meat consumption no more than twice per week, and processed meat consumption no more than twice per week. Each favorable dietary component contributed one point, resulting in a total score ranging from 0 to 5. Participants' dietary patterns were categorized into three groups: poor dietary pattern (score 0 or 1), medium dietary pattern (score 2 or 3), and ideal dietary pattern (score 4 or 5) [29,30].

Physical activity was assessed using Metabolic Equivalent Task (MET) scores derived from the International Physical Activity Questionnaire (IPAQ) administered via touchscreen [31]. Physical activity levels were classified into three intensity categories: low, moderate, and high.

2.6. Statistical analysis

Participants were categorized into seven groups based on the BMI and the presence or absence of abnormal WC, sarcopenia, and metabolic syndrome. Quantitative data with normal distribution were summarized using mean ± standard deviation, while median and interquartile range were used for data that did not conform to a normal distribution. Qualitative variables were described using frequency and percentage. The Wilcoxon rank-sum test was employed to compare skewed continuous variables between two groups. Differences in categorical variables were assessed using either the Chi-square test or Fisher's exact test, depending on the sample size. Missing values in quantitative and qualitative data were imputed using sex-specific medians and modes, respectively. Notably, missing data on family history of depression, which had a relatively high missing rate (13.6 %), was handled using a missing indicators approach.

The Kaplan-Meier method was used to estimate the cumulative incidence of depression across different obesity phenotype groups. Log-rank tests were conducted to compare the incidence of the endpoint event among groups, and the corresponding cumulative risk curves were plotted.

The Cox proportional hazards model was employed to examine the association between various obesity phenotypes and the risk of depression. The proportional hazards assumption was evaluated using the survival curve method. Using the healthy non-obese phenotype as the reference group and considering follow-up time, the hazard ratios (HRs) and corresponding 95 % confidence intervals (95 % CIs) for depression were estimated across different obesity phenotypes. Multivariable models were constructed by sequentially adjusting for potential confounding variables. Model 1 was unadjusted; Model 2 was adjusted for sex and age; and Model 3 was further adjusted for sex, age, race, education qualifications, smoking status, drinking status, family history of depression, history of diabetes, and history of hypertension.

Mediation analyses were conducted using the "CMAverse" R package to evaluate the potential mediating effects of dietary habits and physical activity on the associations identified in the Cox proportional hazards models. These analyses were similarly adjusted for the aforementioned covariates. Parametric bootstrapping (n = 400 replications) was applied to estimate the 95 % CIs and p-values.

Sex, age, smoking status, and drinking status were utilized as subgroup variables to examine their potential effect modification on the association between different obesity phenotypes and the incidence of major depressive disorder. In each subgroup analysis, the grouping factors were excluded from the model, while the remaining adjusted covariates remained unchanged.

To assess the robustness of the association between different obesity phenotypes and the risk of depression, the following sensitivity analyses were conducted: (1) participants who developed depression within one year after the initiation of follow-up were excluded; (2) the analysis was restricted to White individuals; and (3) participants with self-reported depression at baseline but without a documented healthcare diagnosis were excluded.

All data cleaning procedures and statistical analyses were conducted using R 4.4.3 software. A two-sided p-value of less than 0.05 was considered statistically significant.

3. Results

3.1. Demographic characteristics of study population

A total of 391,781 participants were enrolled in this study. The baseline characteristics of individuals with different obesity phenotypes are shown in Table 1. In the whole study population, 209,968 were female and 181,813 were male. Except BMI ≥30 only group (males: 61.0 %) and metabolically unhealthy obesity group (males: 53.2 %), females were predominant in the other groups. The median age for the female was 58.00 years (50.00, 63.00), and for the male, it was 58.00 years (50.00, 64.00). In the sarcopenic obesity group, the proportion of non-smokers was the highest (62.3 %), while in the metabolically unhealthy obesity group, this proportion was the lowest (48.5 %). Participants who drank once a week or more were predominant in healthy non-obese group (74.7 %), and the proportion of this was lowest in multiple obesity phenotypes group (47.6 %).

Table 1.

Baseline characters in different obesity phenotypes.

	Healthy non-obese n = 199,197	Unhealthy non-obese n = 97,203	BMI ≥30 only n = 9236	Central obesity n = 19,065	Metabolically unhealthy obesity n = 56,734	Sarcopenic obesity n = 653	Multiple obesity phenotypes n = 9693
Sex (%)
Female	109,686 (55.1)	50,986 (52.5)	3601 (39.0)	12,550 (65.8)	26,535 (46.8)	368 (56.4)	6242 (64.4)
Male	89,511 (45.0)	46,217 (47.6)	5635 (61.0)	6515 (34.2)	30,199 (53.2)	285 (43.6)	3451 (35.6)
Age (median [IQR])	56.0 [48.0, 62.0]	61.0 [55.0, 65.0]	54.0 [47.0, 61.0]	55.0 [49.0, 61.0]	59.0 [52.0, 64.0]	59.0 [53.0, 64.0]	61.0 [56.0, 65.0]
Race (%)
Non-white	9246 (4.6)	6203 (6.4)	714 (7.7)	1441 (7.6)	2715 (4.8)	73 (11.2)	846 (8.7)
White	189,951 (95.4)	91,000 (93.6)	8522 (92.3)	17,624 (92.4)	54,019 (95.2)	580 (88.8)	8847 (91.3)
Education qualifications(%)
No qualifications	23,772 (11.9)	21,186 (21.8)	1438 (15.6)	3190 (16.7)	12,610 (22.2)	182 (27.9)	3366 (34.7)
Not college/university degree	99,603 (50.00)	48,567 (49.96)	5215 (56.4)	10,507 (55.1)	30,550 (53.9)	321 (49.1)	4680 (48.3)
College or university degree	75,822 (38.1)	27,450 (28.2)	2583 (28.0)	5368 (28.2)	13,574 (23.9)	150 (23.0)	1647 (17.0)
Smoking status (%)
Non-smokers	116,162 (58.3)	50,184 (51.6)	5270 (57.1)	10,653 (55.9)	27,544 (48.5)	407 (62.3)	5057 (52.2)
Ex-smokers	62,687 (31.5)	36,209 (37.3)	3114 (33.7)	6845 (35.9)	23,486 (41.4)	201 (30.8)	3753 (38.7)
Current smokers	20,348 (10.2)	10,810 (11.1)	852 (9.2)	1567 (8.2)	5704 (10.1)	45 (6.9)	883 (9.1)
Drinking status (%)
Never	11,989 (6.0)	9155 (9.4)	608 (6.6)	1487 (7.8)	5488 (9.6)	95 (14.5)	1804 (18.6)
Occasionally	38,517 (19.3)	22119 (22.8)	2140 (23.2)	5222 (27.4)	16689 (29.4)	201 (30.8)	3277 (33.8)
1–2 times a week	51,805 (26.0)	24408 (25.1)	2641 (28.5)	5201 (27.3)	14894 (26.3)	171 (26.2)	2274 (23.5)
3–4 times a week	51,682 (26.0)	20920 (21.5)	2259 (24.5)	3939 (20.7)	10754 (19.0)	107 (16.4)	1301 (13.4)
Daily or almost daily	45,204 (22.7)	20601 (21.2)	1588 (17.2)	3216 (16.8)	8909 (15.7)	79 (12.1)	1037 (10.7)
Family history (%)
No	158,632 (79.6)	74,322 (76.5)	7158 (77.5)	14,580 (76.5)	42,884 (75.6)	469 (71.8)	6832 (70.5)
Yes	18,074 (9.1)	8308 (8.5)	753 (8.1)	1756 (9.2)	4686 (8.3)	59 (9.1)	881 (9.1)
miss	22,491 (11.3)	14,573 (15.0)	1325 (14.4)	2729 (14.3)	9164 (16.1)	125 (19.1)	1980 (20.4)
Diabetes (%)
No	197,747 (99.3)	89,672 (92.2)	9108 (98.6)	18,835 (98.8)	48,445 (85.4)	625 (95.7)	7879 (81.3)
Yes	1450 (0.7)	7531 (7.8)	128 (1.4)	230 (1.2)	8289 (14.6)	28 (4.3)	1814 (18.7)
Hypertension (%)
No	171,439 (86.1)	61,780 (63.6)	7073 (76.6)	14,154 (74.2)	28,645 (50.5)	473 (72.4)	4600 (47.5)
Yes	27,758 (13.9)	35,423 (36.4)	2163 (23.4)	4911 (25.8)	28,089 (49.5)	180 (27.6)	5093 (52.5)

Note.

Definition for each group.

Healthy non-obese: 18.5 kg/m² ≤ BMI <30 kg/m², and having no abnormal WC, sarcopenia, nor metabolic syndrome.

Unhealthy non-obese: 18.5 kg/m² ≤ BMI <30 kg/m², and having abnormal WC, sarcopenia, or metabolic syndrome.

BMI ≥30 only: BMI ≥30 kg/m² and without the above three symptoms.

Central obesity: BMI ≥30 kg/m² combined with WC > 102 cm in male and >88 cm in female.

Sarcopenic obesity: BMI ≥30 kg/m² accompanied by "Sarcopenia Probable".

Metabolically unhealthy obesity: BMI ≥30 kg/m² with metabolic syndrome.

Multiple obesity phenotypes: having two or more obesity phenotypes.

3.2. Correlation between different obesity phenotypes and the risk of depression

After a median follow-up duration of 13.39 years, a total of 20,027 new cases of depression were identified in the overall study population (Table 2). After taking multiple covariates into account, compared with healthy non-obese individuals, BMI ≥30 only would increase the risk of depression by 15 % (HR: 1.15, 95 % CI: 1.04, 1.28), while central obesity would increase the risk by 60 % (HR: 1.60, 95 % CI: 1.51, 1.70). The HR for the metabolically unhealthy obesity group was 1.63 (95 % CI: 1.56, 1.70), while sarcopenic obesity would increase the risk of depression by 86 % (HR: 1.86, 95 % CI: 1.40, 2.46). It can be seen that obesity characterized only by BMI over 30 kg/m² has the lowest impact on the risk of depression. If obesity is combined with other symptoms, it will increase the risk of depression, especially when combined with sarcopenia, which has the greatest impact on obese individuals. Additionally, we also found that non-obese with comorbidities (unhealthy non-obese) would increase the risk of depression (HR: 1.43, 95 % CI: 1.38, 1.48), and the coexistence of multiple obesity phenotypes would further increase the risk of depression (HR: 2.55, 95 % CI: 2.39, 2.72). Specific information about these two groups can be found in Supplementary Materials Table S1.

Table 2.

Adjusted HRs for depression by different obesity phenotypes at baseline.

	Healthy non-obese (Reference)	Unhealthy non-obese	BMI ≥30 only	Central obesity	Metabolically unhealthy obesity	Sarcopenic obesity	Multiple obesity phenotypes
Case	7490	5633	396	1261	4017	49	1181
Person-years	2,579,770	1,218,618	119,080.3	241,103.1	705,758.4	8009.54	113,614.9
Incidence density (1/10,000 person-years)	29.03	46.22	33.25	52.30	56.92	61.18	103.95
Model 1	1.00	1.59 (1.54,1.65)	1.14 (1.03,1.27)	1.80 (1.70,1.91)	1.97 (1.89,2.04)	2.10 (1.59,2.78)	3.59 (3.38,3.82)
Model 2	1.00	1.64 (1.58,1.70)	1.22 (1.10,1.35)	1.73 (1.63,1.84)	2.05 (1.98,2.14)	2.11 (1.59,2.79)	3.52 (3.31,3.75)
Model 3	1.00	1.43 (1.38,1.48)	1.15 (1.04,1.28)	1.60 (1.51,1.70)	1.63 (1.56,1.70)	1.86 (1.40,2.46)	2.55 (2.39,2.72)

HRs: hazard ratios.

Model 1: none of covariates.

Model 2:sex and age.

Model 3:sex, age, race, education qualifications, smoking status, drinking status, family history of depression, diabetes, and hypertension.

Cumulative risk curves for depression across various obesity phenotypes were constructed and are presented in Fig. 2, along with the results of Log-rank tests. We can also observe that, with healthy non-obese as the reference group, among the four obesity phenotypes we focused on, only the population with BMI ≥30 kg/m² had the lowest increase in the cumulative risk of depression, and this was followed by central obesity, metabolically unhealthy obesity, and sarcopenic obesity.

Fig. 2.

Cumulative risk curves for the association of different obesity phenotypes with depression.

3.3. Mediation effect of diet and physical activity

As presented in Table 3 and Fig. 3, both dietary habits and physical activity were found to exert mediating effects on the association between various obesity phenotypes and the onset of depression. Taking the healthy non-obese group as the reference, in the relationship between the four obesity phenotypes and depression, the proportions mediated of physical activity were 2.49 % (BMI ≥30 only), 7.08 % (metabolically unhealthy obesity), 5.29 % (central obesity), and 6.22 % (sarcopenic obesity). The mediating effect of diet was lower than physical activity. Its mediating proportions were 1.60 % (BMI ≥30 only), 1.86 % (metabolically unhealthy obesity), 1.47 % (central obesity), 1.25 % (sarcopenic obesity).

Table 3.

Mediating effects of physical activity and diet.

Mediator	Physical activity	Diet
Obesity phenotype: BMI ≥ 30 only vs healthy non-obese
Total effect	1.12 (1.00,1.26)	1.15 (1.04,1.27)
Natural direct effect	1.12 (1.00,1.26)	1.15 (1.04,1.27)
Natural indirect effect	1.00 (1.00,1.01)	1.00 (1.00,1.00)
Proportion mediated	2.49 % (0.24 %,17.49 %)	1.60 % (0.34 %,5.55 %)
Obesity phenotype: metabolically unhealthy obesity vs healthy non-obese
Total effect	1.52 (1.46,1.60)	1.57 (1.49,1.64)
Natural direct effect	1.49 (1.42,1.56)	1.56 (1.48,1.63)
Natural indirect effect	1.02 (1.02,1.03)	1.01 (1.00,1.01)
Proportion mediated	7.08 % (4.87 %,9.43 %)	1.86 % (0.91 %,2.74 %)
Obesity phenotype: central obesity vs healthy non-obese
Total effect	1.60 (1.49,1.71)	1.61 (1.51,1.70)
Natural direct effect	1.57 (1.46,1.68)	1.60 (1.50,1.69)
Natural indirect effect	1.02 (1.01,1.03)	1.01 (1.00,1.01)
Proportion mediated	5.29 % (3.52 %,7.34 %)	1.47 % (0.58 %,2.50 %)
Obesity phenotype: sarcopenic obesity vs healthy non-obese
Total effect	1.53 (1.09,2.12)	1.74 (1.32,2.30)
Natural direct effect	1.50 (1.07,2.08)	1.73 (1.31,2.28)
Natural indirect effect	1.02 (1.01,1.04)	1.01 (1.00,1.01)
Proportion mediated	6.22 % (2.67 %,17.05 %)	1.25 % (0.41 %,3.10 %)

Note.

Covariates: sex, age, race, education qualifications, smoking status, drinking status, family history of depression, diabetes, and hypertension.

Fig. 3.

The mediating effect of physical activity and diet in the relationship between different obesity phenotypes and depression. Note: ACME: Average causal mediation effect; ADE: Average direct effect; PM: Proportion mediated; A: The mediating effect of physical activity in the relationship between BMI ≥30 only and depression; B: The mediating effect of physical activity in the relationship between metabolically unhealthy obesity and depression; C: The mediating effect of physical activity in the relationship between central obesity and depression; D: The mediating effect of physical activity in the relationship between sarcopenic obesity and depression; E: The mediating effect of diet in the relationship between BMI ≥30 only and depression; F: The mediating effect of diet in the relationship between metabolically unhealthy obesity and depression; G: The mediating effect of diet in the relationship between central obesity and depression; H: The mediating effect of diet in the relationship between sarcopenic obesity and depression.

3.4. Results of subgroup analysis

As shown in Supplementary Materials (Table S2), subgroup analysis revealed that individuals younger than 60 years old had a higher risk of obesity-associated depression compared to those aged 60 years or older, and this was present in four obesity phenotypes. Among the associations between various obesity phenotypes and the risk of depression, the risk for female was higher than that for male. Additionally, except participants with sarcopenic obesity, people who have never smoked exhibited a greater risk of depression related to obesity than former or current smokers. Alcohol consumption did not significantly modify the relationship between obesity subtypes and depression risk.

3.5. Sensitivity analysis

The results of sensitivity analysis are presented in supplementary material (Table S3 to Table S5). After adjusting for all covariates, the findings remained consistent with those from the initial main analyses, suggesting that the study's results are relatively robust.

4. Discussion

We conducted a comprehensive assessment of the impact of different obesity phenotypes on the incidence of depression using a large community-based cohort. The results indicate that among the four obesity phenotypes, the participants who only had BMI over 30 kg/m² had the lowest increase in the risk of depression. The risk of depression was similar between central obesity and metabolically unhealthy obesity, and the risk of depression was highest in participants with sarcopenic obesity. We also found that diet and physical activity mediate the association between obesity and depression.

In fact, the status of BMI ≥30 only is similar to metabolically healthy obesity without sarcopenia. A previous meta-analysis showed that metabolically healthy obese persons also had a slightly increased risk of depression compared with metabolically healthy nonobese persons, which is consistent with our findings [32]. However, according to our results, the risk of depression among persons with BMI ≥30 kg/m² only was much lower than that among the other three obesity phenotypes, which indicates that a high-fat state alone is not a major cause of depression.

In contrast, the presence of central obesity marked a significant step-up in depression risk, corroborated by comparable studies conducted in the United States and China [33,34]. This may be due to the fact that central obesity implies the accumulation of visceral fat, which can exacerbate the inflammatory state more strongly than subcutaneous fat. The inflammatory state is the key pathogenic mechanism leading to depression and anxiety. Depression is associated with increased circulating levels of proinflammatory cytokines, prostaglandins and other arachidonic acid derivatives [35]. Cytokines, chemokines, and other components of peripheral inflammation can cross the blood-brain barrier and influence various brain structures involved in the regulation of depression [7]. Furthermore, people with central obesity are more likely to develop insulin resistance. Previous studies have shown that insulin resistance not only reduces the antidepressant efficacy of insulin, but also increases the risk of depression by aggravating inflammatory states [36,37].

The risk was further amplified in the metabolically unhealthy obesity phenotype. This group is characterized by exacerbated inflammation, dyslipidemia, hyperglycemia, and hypertension. The cumulative physiological burden of these multiple metabolic disturbances may synergistically act on the brain. For instance, the combination of hypercortisolemia (from HPA axis dysregulation), increased oxidative stress, and endothelial dysfunction in metabolically unhealthy obesity creates a neurotoxic milieu that can damage brain regions critical for mood regulation, such as the hippocampus and prefrontal cortex [7,38]. From a psychological standpoint, the diagnosis and management of multiple metabolic conditions can itself be a chronic stressor, contributing to feelings of illness burden and helplessness, thereby increasing vulnerability to depression.

In a separate small-scale longitudinal study conducted in the United Kingdom and a study based on NHANES data, they both discovered sarcopenic obesity increased risk for depression [39,40]. Our study got the consistent result and found sarcopenic obesity confers the highest risk of depression among all common obesity phenotypes. The reason why sarcopenic obesity brings such a high risk of depression may be that it will cause inflammation in the body, and muscle tissue has a strong secretory function. The loss of muscle affects the release of biological factors such as Cathepsin B, which have antidepressant effects [18,41]. From a behavioral perspective, sarcopenic obesity is strongly associated with frailty, functional limitations, and reduced independence. Loss of physical capacity can lead to social isolation, reduced participation in rewarding activities, and reduced self-efficacy, all of which are potential psychosocial risk factors for depression [42].

In this study, both physical activity and dietary habits were found to mediate the relationship between obesity and depression. Huang et al. demonstrated that moderate-intensity physical activity and reduced sedentary behavior can lower the risk of depression among individuals with obesity [43]. This effect may be attributed to the sense of satisfaction derived from spontaneous physical activity, which contributes to improved psychological well-being. Additionally, physical activity stimulates the secretion of neurotransmitters that influence mood regulation [44,45]. Prior studies have found that high-fat diets may disrupt leptin and insulin signaling in the central nervous system, potentially contributing to the development of depression [7]. High-fat diets can lead to fat accumulation and systemic inflammation. Although sugar consumption may temporarily relieve emotional stress, high-sugar diets are associated with increased risks of depression and anxiety [46]. When combined with high-fat intake, high-sugar diets can induce glucolipotoxicity in pancreatic β cells, impairing insulin secretion and exacerbating metabolic dysfunction. Given that individuals with obesity often exhibit unhealthy dietary patterns, such as excessive consumption of high-fat or high-sugar foods, diet plays a crucial role in the interplay between obesity and depression [7]. However, it is worth noting that although obesity can affect depression risk by affecting physical activity frequency or dietary habits, people with depression can similarly increase the likelihood of obesity due to lack of physical activity or poor dietary habits as a result of depressive episodes [47]. This creates a two-way pathway that can be vicious. It is because of this bidirectional pathway that our study, although statistically significant, cannot represent a mediating effect on a causal link because of the lack of temporal order distinction between obesity phenotypes and mediators. From an intervention perspective, our findings show the potential of targeting physical activity and diet not only for weight management but also for mental health improvement. Integrated interventions that combine physical activity promotion with dietary counseling have shown promise in simultaneously addressing obesity and depression.

In the subgroup analysis, we found that sex, age and smoking had effect modification on the relationship between obesity and depression. In a previous study on obesity and depression, there was evidence suggesting that in female, body weight increase was associated with depression developing, while this connection did not exist in male [48]. The reason may be obese female are more likely to be stigmatized, and male depression is often not identified and diagnosed timely [48,49]. The reason for the differences in the risk of obesity and depression among people of different age groups is that middle-aged people bear more work and financial pressures in life, and they also need to deal with weight gain and emotional problems brought about by aging [49]. Contrary to the common belief that smoking increases the risk of depression, our research found that among non-smokers, the risk of depression related to obesity was actually higher. A prospective study in China also discovered that smoking reduces the risk of depression but the explanation is insufficient [50]. Therefore, the complex relationship in the smoking population still requires further discussion.

This study has several strengths, including a large sample size, extended follow-up duration, and a comprehensive assessment of the real-world associations between different obesity phenotypes and depression. Additionally, it accounts for the potential mediating roles of diet and physical activity. Nevertheless, certain limitations should be acknowledged. First, although grip strength is an internationally recognized indicator of sarcopenia, incorporating additional measures such as limb skeletal muscle mass would enhance the scientific rigor and accuracy of the assessment. Second, the study was conducted using data from the UK Biobank, where over 90 % of participants were of White ethnicity. Given the known differences in obesity phenotype distribution across racial groups, the generalizability of our findings may be limited. Moreover, both exposure and mediator assessments were derived solely from baseline data, thereby causing unable to establish temporal sequencing. Finally, although we controlled for a wide range of potential confounding variables, it remains challenging to fully account for all possible confounders and their residual effects. Therefore, in future studies, we plan to test the consistency of our findings in different ethnic groups, such as Asian populations. At the same time, the data about skeletal muscle mass should be collected in the follow-up study to make the judgment of sarcopenia more comprehensive. In addition to observational studies, interventions such as weight loss, metabolic regulation, and muscle mass enhancement can be used to explore which intervention has a more effective effect on reducing the risk of depression in obese people.

In conclusion, our study found that obesity only with BMI over 30 kg/m² has a small increase in the risk of depression, while central obesity, metabolically unhealthy obesity and sarcopenic obesity can greatly increase the risk of depression, and sarcopenic obesity patients have the highest risk of depression. Diet and physical activity were identified as mediating factors in these associations. These findings reveal the true risk of depression in individuals with complex obesity profiles and provide a basis for the prevention of obesity-depression comorbidity in obese patients.

CRediT authorship contribution statement

Yang Liu: Writing – review & editing, Writing – original draft, Formal analysis, Data curation. Yue Zheng: Writing – review & editing, Data curation. Mingfang Wang: Data curation. Juan Liao: Writing – review & editing. Lu Long: Writing – review & editing, Supervision, Data curation, Conceptualization.

Ethical approval and patient consents

The UK Biobank study was performed in accordance with the Declaration of Helsinki and had obtained ethical approval from the North West Multi-centre Research Ethics Committee (approval number: 11/NW/03820). All participants have provided their informed consent to participate.

Funding source

This work was supported by the Chengdu Municipal Science and Technology Bureau (Grant No. 2024-YF05-02077-SN) and Institute of Health New Quality Productive Forces, West China School of Public Health, Sichuan University. The funding sponsors had no role in the study design, data collection, data analysis, data interpretations, or composition of the manuscript.

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: