Sunshine Duration, Genetic Predisposition, and Incident Depression: Findings from a Prospective Cohort

Abstract

Background: Published studies have documented the association between sunshine duration and depression symptoms; however, the evidence regarding the long-term effects and potential mechanisms remains insufficient. This study aimed to examine the association between sunshine duration and incident depression and to explore potential mediating pathways. Methods: A total of 336,805 participants from the UK Biobank were included in the study. Meteorological exposures were estimated using the bilinear interpolation approach and time-weighted method. The association between sunshine duration and incident depression was examined through the time-dependent Cox proportional hazard model and generalized propensity score model. Vitamin D, calcium, immune biomarkers, an aggregated inflammation score (INFLA-score), and sleep pattern were selected as the potential mediators. Causal mediation analysis was employed to elucidate underlying mediating effects. Results: With a median follow-up of 13 years, 13,862 cases of incident depression were identified. Sunshine duration demonstrated a negative association with the incident depression. The effects were stronger among the elderly, alcohol consumers, individuals who spent less time outdoors, and those who were less physically active. Vitamin D, calcium, INFLA, neutrophils, and monocytes emerged as the top five contributors of immune biomarkers to the natural indirect effect. The combined mediating effect of top five biomarkers and sleep pattern accounted for 30% of the total effect of sunshine duration on the incident depression. Conclusion: Our study suggests that longer sunshine duration might mitigate depression through vitamin D-related metabolism, inflammation, and sleep pattern. It may serve as an effective natural antidepressant, particularly for the elderly, alcohol consumers, less outdoor spenders, and those who were less physically active.

1. Introduction

The escalating prevalence of depressive disorders imposes a substantial burden on global public health. Approximately 280 million people in the world are suffering from depression, and the number continues to rise [1]. According to the 2019 Global Burden of Disease report, the age-standardized disability-adjusted life-year rate for depressive disorders was 3440.1 per 100,000 population [2]. Therefore, there is an urgent need to identify its risk factors in order to reduce the detrimental impact.

Previous studies have documented seasonal variations in psychological changes in humans, indicating a link between weather conditions and depression [3]. Sunlight exposure has been proposed as a potential method to reduce the risk of depression [4]. A growing number of studies focus on the effect of sunlight on depression, but the findings are inconsistent. For instance, a study in China showed that sunshine duration was associated with outpatient visits for depression [5]. However, another study in the southern Netherlands found no discernible effect of sunlight on depression [6]. In addition, existing studies are mainly based on cross-sectional and ecological studies, and evidence from cohort studies with large sample sizes is still lacking [7].

The mechanism underlying the association between sunlight and depression is not fully understood. Some studies have suggested that sunlight is correlated with various biomarkers, including vitamin D, calcium, C-reactive protein [8]. Furthermore, sunshine has been reported to be associated with human behaviors such as sleep patterns, and these factors may, in turn, exert further influence on mental health [9,10,11]. However, the extent to which these biomarkers and behavioral factors may mediate the association between sunlight and depressive disorders remains unclear.

Therefore, this study aims to investigate the association between long-term exposure to sunshine duration and the risk of incident depression across different levels of genetic susceptibility, and also exploring the potential mediation effects of biochemical markers and sleep patterns.

2. Methods

2.1. Study Design and Participants

UK Biobank is a national cohort that recruited 502,940 individuals aged 40–69 years in England, Scotland, and Wales between 2006 and 2010 [12]. Participants’ demographic, socioeconomic, health, and lifestyle information was collected through questionnaires and physical measurements at baseline and followed longitudinally thereafter. Detailed information of the study is available online (https://www.ukbiobank.ac.uk/ (accessed on 4 September 2025)). Ethics approval was obtained from the North West Multicenter Research Ethics Committee, and all recruited participants provided informed consent.

In this analysis, we excluded 42,151 participants with depression at baseline to ensure that only incident cases were included. Additionally, 91,674 participants with missing data on exposure variables and covariates, 20,617 non-white participants, and 11,214 participants with a history of antidepressant medication use were also excluded. The remaining 336,805 participants were included (Figure S1). Table S2 presents the sample size utilized in each analysis.

2.2. Environmental Exposure

The monthly sunshine duration and ambient temperature were collected from the HadUK-Grid dataset, which provides high-resolution meteorological data for the UK from 1862 to 2022 [13]. Based on station observations, monthly meteorological variables were modelled on a 1 km × 1 km grid using the inverse distance-weighted interpolation method [14]. We employed the bilinear interpolation method to assess the monthly meteorological exposure of each participant with the grid data and geocoded home addresses.

Furthermore, as some participants may have changed residence during the follow-up period, we implemented a time-weighted method to calculate a more accurate exposure. First, we collected information on the participants’ home addresses and the duration of their stay at each address. The time spent at each address was then calculated on a monthly basis. Finally, the duration of residence served as a weighting factor in computing the average exposure for study subjects in each month. A comprehensive description of the exposure assessment strategy can be found in a prior publication [15].

Air pollution data were obtained from the UK’s Department for Environment, Food and Rural Affairs (DEFRA), which provides an annual mean concentration of air pollutants at a high resolution. These air pollution data were generated on 1 km × 1 km grid using an air dispersion model that incorporates diverse sources from the National Atmospheric Emissions Inventory, a combination of air-pollutant and greenhouse-gas dataset. Subsequently, the data were calibrated using automatic air quality monitoring data from DEFRA’s rural and urban network [16]. Grid data for fine particulate matter (PM_2.5), inhalable particulate matter (PM₁₀), nitrogen dioxide (NO₂), and nitrogen dioxides (NO_x) were extracted for the period 2003–2022.

Gridded data on artificial light at night was collected from the Defense Meteorological Satellite Program Operational Linescan System (DMSP-OLS) stable nighttime light (NTL) data and the Suomi National Polar-orbiting Partnership Visible Infrared Imaging Radiometer Suite (NPP-VIIRS) NTL data. The calibrated data can be accessed at https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/YGIVCD (accessed on 4 September 2025). This dataset has a resolution of 500 m × 500 m and has been reported to demonstrate good consistency, with an R² value of 0.87 [17]. The data was transformed into units of radiance (nW/cm²/sr).

We also employed the bilinear interpolation approach and a time-weighted method to calculate the annual concentration of air pollution and artificial light at night for each participant.

2.3. Measurement of Peripheral Markers

Blood biochemistry data were measured from 480,000 participants at baseline (2006–2010). Standard hematological tests were conducted on fresh whole blood within 24 h of collection. Quality control procedures are reported online (https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/biomarker_issues.pdf (accessed on 4 September 2025)). Blood cell samples were analyzed using an automated hematology analyzer, the Coulter LH750 (https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/haematology.pdf (accessed on 4 September 2025)). The UK Biobank provided thirty blood markers and thirty-one blood cell counts (Table S1).

We included the biomarkers related to vitamin D metabolism and inflammation including vitamin D, calcium, lymphocytes, monocytes, leukocytes, platelets, C-reactive protein (CRP) levels, red cell distribution width, basophils, eosinophils, the percentages of lymphocytes, the percentages of monocytes, neutrophils relative to the total white blood cell count, the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR) [18]. We further calculated the INFLA-score, a well-established marker of low-grade inflammation. It has also been reported as an indicator of pro-inflammatory processes in various biological mechanisms of the immune response. To calculate the INFLA-score, values of CRP, WBC, platelet count, and the NLR ranging from +1 to +4 were assigned to the top four deciles (7th to 10th) for all components, while those in the lowest four deciles (1st to 4th) received values from −4 to −1. The INFLA score is the total sum of the scores from these four biomarkers, each weighted equally, resulting in a range from −16 to +16. A higher score reflects a greater degree of low-grade inflammation [19].

2.4. Assessment of Sleep Characteristics

Data on sleep behaviors from the UK Biobank were self-reported and collected from the following items in the questionnaire: sleep duration (≥9 h/day, 7–8 h/day, or <7 h/day), chronotype (morning, morning than evening, evening than morning; or evening), insomnia (usually, sometimes, or never/rarely), snoring (no or yes), and daytime sleepiness (often/always, sometimes, or never/rarely) (Table S1) [20].

2.5. Genotyping and Imputation

The genotyping, imputation and quality control procedures of the UK Biobank have been described previously [21]. Briefly, both the Affymetrix UK BiLEVE Axiom and Affymetrix UK Biobank Axiom arrays were used for genotyping and imputation procedures utilized the IMPUTE4 program. Rigorous sample-based and marker-based quality control measures were applied to select high-quality genetic data. Exclusion criteria included participants with mismatches between self-reported sex and genetic sex, high missing rates, mixed ancestral backgrounds, and genetic relatedness [22].

2.6. Follow-Up and Ascertainment of Outcomes

The incidence of depressive disorders was determined through hospital records and mental health questionnaires. Hospital records utilized the International Classification of Diseases, 10th Edition (ICD-10) codes F32-33 to define depressive disorders. The Patient Health Questionnaire-9 (PHQ-9) was employed to assess depressive status during 2016–2017 (Table S1). The PHQ-9 consists of nine items, with higher total scores indicating a greater level of depression.

2.7. Covariates

We considered potential confounders based on previous literature and further selected the variables that need to be controlled through directed acyclic graph (DAG) [23] (Figure S2). The covariates included as potential confounders in our analysis were age, sex (male or female), household income (unknown, ≥52,000 £, 31,000–51,999 £, 18,000–30,999 £, or <18,000 £), Townsend deprivation index (high, moderate, or low), polypharmacy (0, 1–2, or ≥3), smoking status (never, former, or current), time spent outdoors (high, moderate, or low), residential area (urban or rural), percentage of greenspace at 300 m buffers (%; high, moderate, or low), ambient temperature (low/moderate/high), and air pollution (low/moderate/high). Further details on these variables were provided in Table S1.

2.8. Statistical Analyses

2.8.1. Association Between Sunshine Duration and Depressive Disorders

We conducted two analyses to explore the causal relationship between sunshine duration and the incidence of depressive disorders. Analysis 1 used the data from hospital records with time-dependent Cox proportional hazard model combined with genderized propensity score (GPS) model. Analysis 2 included participants with data from mental health questionnaire with Genderized linear model (GLM) combined with GPS model.

GPS represents a balancing score derived from the conditional probability density of exposure to continuous factors given the confounders. Employing GPS with a weighting method facilitates effective confounding balance by creating a pseudo-population that simulates a randomized controlled trial [24,25,26]. In consideration of machine-learning algorithms, recognized for their ability to enhance GPS estimation and prediction, we used an Extreme Gradient Boosting (XGboost) model [27]. Stabilized Inverse Probability Weights (IPWs) were computed with a specific formula, described in previous work [28,29,30]. Detailed information on GPS and IPW is shown in Supplementary methods. To further enhance the stability of IPWs, we truncated the lowest and highest 1% of weights [24]. The Absolute Correlation value (AC) served as an index for assessing covariate balance post-weighting, with a value below 0.1 considered indicative of good covariate balance [24,31].

The time-dependent Cox proportional hazard model was employed to investigate the associations between the exposure and outcome. Sunshine duration, ambient temperature, air pollution, and age were considered as time-varying covariates. We assumed that sunshine duration and ambient temperature varied from month to month, and air pollution and age varied annually. In the time-independent model, we utilized baseline data for exposure and covariates, assuming that these variables do not change over time.

For these two analyses, we constructed four models: Model 1 adjusted for age and sex. Model 2 additionally included adjustments for residential area, household income, Townsend deprivation index, smoking status and polypharmacy. Model 3 further incorporated adjustments for time spent outdoors, ambient temperature, green space, and PM_2.5. Model 4 was developed using the causal inference method, serving as our primary model.

2.8.2. Polygenic Risk Score (PRS) and Interaction

Single nucleotide polymorphisms (SNPs) with a significance less than 5 × 10⁻⁸ reported in previous Genome-Wide Association Studies (GWAS) were chosen to establish polygenic risk scores for depression in the UK population [32] (Table S4). The construction of polygenic risk score (PRS) involved the following steps:

(1)

Recoding SNPs as 0, 1, or 2 based on the risk allele;

(2)

Calculating PRS with the following formula:

$$PRS=(\sum _{i=1}^n{\beta }_i\times {SNP}_i)\times \left(n/{\sum }_{i=1}^n{\beta }_i\right)$$

where

$n$ is the total number of SNPs included in the PRS construction.

${SNP}_i$ is the number of risk alleles (0, 1, or 2) carried by the individual at the $i^{th}$ SNP.

${\beta }_i$ is the natural logarithm of the odds ratio (log(OR)) for ${SNP}_i$, derived from previously published GWAS results [32].

(3)

Standardizing the PRSs.

The PRSs were stratified into low (quintile 1), intermediate (quintile 2 to 4), and high (quintile 5) according to the quintile distribution of PRS. We evaluated the interactions to discern the combined effect of genetic risk and sunshine duration on depression. The product term of sunshine duration and PRS classes was incorporated into four models, with the p-values of the product terms indicating interaction in the multiplicative scale.

Similarly, we calculated the hazard ratios (HRs) for sunshine duration, stratified by demographic and lifestyle factors, including age, sex, household income, physical activity, smoking status, alcohol intake, time spent outdoors, residential area and artificial light at night. We also assessed the multiplicative interaction effects of the sunshine duration and the stratified factors.

2.8.3. Mediation Analysis

We investigated the mediating effects of biochemical markers and sleep pattern that underlie the association between sunshine duration and depression.

Biochemical Markers

Before analysis, we excluded data outside the 3 interquartile ranges of the median. A rank-based inverse normal transformation approach was employed to standardize units and approximate a normal distribution [33].

Causal mediation analyses were undertaken to investigate whether the biochemical marker mediates the association between sunshine duration and depression. As biochemical marker data were not measured monthly, we utilized the average monthly sunshine duration over the 3 years preceding recruitment date as the exposure. We estimated both the natural indirect effect (NIE) and the natural direct effect (NDE), where NIE indicates the causal effect and the indirect path from exposure through the mediator to outcome, and NDE indicates the causal effect and the direct path from exposure to outcome [34]. The proportion of the association mediated through the potential mediators was calculated by $log\left(NIE\right)/\left(log\left(NIE\right)+log\left(NDE\right)\right)$, with 95% confidence intervals estimated through bootstrap methods [35].

Sleep Pattern

Chronotype, insomnia, snoring, sleep duration, and excessive daytime sleepiness were incorporated to formulate a sleep score to reflect sleep pattern. Early chronotypes (morning or morning than evening), sleep 7–8 hour a day, never or rarely having insomnia, no self-reported snoring, and no frequent daytime sleepiness (never/rarely or sometimes) were recoded as 0, while the remaining items were recoded as 1 [36]. The summation of all component scores yielded the sleep score, with a higher score indicating an unhealthy sleep pattern [20]. Causal mediation analyses were further applied to assess the mediating effects of sleep pattern on the relationship between sunshine duration and depression.

2.8.4. Parallel Mediation Analysis

We selected the top five biomarkers based on their individual mediation proportions. Additionally, we conducted a parallel mediation analysis to assess the joint mediation effect of these top five biomarkers along with sleep patterns.

2.9. Stratification and Interaction Analysis

Stratified analyses were performed according to gender, age, household income, alcohol intake, smoking status, time spent outdoors, physical activity, residential area, exposure to artificial light at night, and polygenic risk score (PRS). To assess potential effect modification, we included interaction terms between each stratification variable and sunlight exposure in the regression models. The statistical significance of the interaction terms was used to evaluate potential interactions.

2.10. Sensitivity Analysis

We assessed the robustness of the results through the following sensitivity analyses: (1) calculating the E-value, an index commonly utilized to quantify the magnitude of unmeasured confounding [37]; (2) excluding participants with depressions or death within one year of follow-up; (3) conducting the mediation analysis using the method developed by Huang et al. [38]; (4) using educational level as a surrogate for household income to control for the influence of social-economic background; (6) dividing the latitude as low (tertile 1), medium (tertile 2), and high (tertile 3) groups, and further including latitude group as covariates; (7) further included the skin pigmentation (diagnosed by ICD-10 code L81) as the covariate; (8) further included season as the covariate.

All analyses were conducted by R software, version 4.4.0. p < 0.05 at two side was considered as statistical significance, and the Bonferroni method was applied to corrected the multiple testing.

3. Results

3.1. Descriptive Results

A total of 336,805 participants were included in the present analysis, with 13,862 cases of depressive disorders occurring over a median follow-up of 13 years. As shown in

Table 1. Characteristics of study participants at baseline.

Characteristics	All Participants	None-Cases	Incident Depression	p
Number of participants	336,805	322,943	13,862
Sex [n (%)]
Female	175,215 (52)	167,030 (52)	8185 (59)	<0.001
Male	161,590 (48)	155,913 (48)	5677 (41)
Age [year (mean ± SD)]	56.9 ± 8.1	56.9 ± 8.0	56.5 ± 8.4	<0.001
Household income [n (%)]
Less than 18,000 £	59,350 (18)	55,581 (17)	3769 (27)	<0.001
18,000 to 30,999 £	74,214 (22)	71,005 (22)	3209 (23)
31,000 £ to 51,999 £	78,479 (23)	75,667 (23)	2812 (20)
Greater than 52,000 £	81,309 (24)	79,209 (25)	2100 (15)
Unknown	43,453 (13)	41,481 (13)	1972 (14)
Residential area [n (%)]
Urban	284,619 (85)	272,551 (84)	12,068 (87)	<0.001
Rural	52,186 (16)	50,392 (16)	1794 (13)
Physical activity [n (%)]
Low	49,875 (15)	47,535 (15)	2340 (17)	<0.001
Moderate	115,947 (34)	111,550 (35)	4397 (32)
High	118,823 (35)	114,208 (35)	4615 (33)
Unknown	52,160 (16)	49,650 (15)	2510 (18)
Smoking status [n (%)]
Never	183,528 (55)	177,101 (55)	6427 (46)	<0.001
Former	121,071 (36)	115,900 (36)	5171 (37)
Current	32,206 (10)	29,942 (9)	2264 (16)
Polypharmacy [n (%)]
0	102,505 (30)	99,772 (31)	2733 (20)	<0.001
1–2	119,654 (36)	115,085 (36)	4569 (33)
≥3	114,646 (34)	108,086 (34)	6560 (47)
Time spent outdoors [n (%)]
Low	108,104 (32)	103,932 (32)	4172 (30)	<0.001
Moderate	82,971 (25)	79,620 (25)	3351 (24)
High	145,730 (43)	139,391 (43)	6339 (46)
Townsend deprivation index [n (%)]
Low	115,626 (34)	111,786 (35)	3840 (28)	<0.001
Moderate	113,897 (34)	109,522 (34)	4375 (32)
High	107,282 (32)	101,635 (32)	5647 (41)
Greenspace [n (%)]
Low	107,259 (32)	102,854 (32)	4405 (32)	<0.001
Moderate	110,809 (33)	105,883 (33)	4926 (36)
High	118,737 (35)	114,206 (35)	4531 (33)
Ambient temperature [n (%)]
Low	113,081 (34)	108,382 (34)	4699 (34)	0.001
Moderate	114,461 (34)	109,595 (34)	4866 (35)
High	109,263 (32)	104,966 (33)	4297 (31)
PM2.5 [n (%)]
Low	111,142 (33)	106,441 (33)	4701 (34)	<0.001
Moderate	111,148 (33)	106,329 (33)	4819 (35)
High	114,515 (34)	110,173 (34)	4342 (31)
Average monthly sunshine duration over the three years prior to recruitment (mean (SD))	129 ± 8	129 ± 8	128 ± 8	<0.001

Abbreviations: PM_2.5, fine particulate matter; SD, standard deviation.

, the mean (SD) age of all participants was 56.9 (8.1) years and 52% were females. The majority of study subjects resided in urban areas, with household incomes exceeding 31,000 £. The mean exposure level to monthly sunshine duration was 129 ± 8 h. Additionally, in comparison to participants without depression, those with depression tended to be female (59% (95% CI: 58–60%) vs. 52% (95% CI: 51–52%), younger (56.5 (95% CI: 56.4–56.6) years vs. 56.9 (95% CI: 56.8–56.9) years, had lower economic status (15% (95% CI: 15–16%) vs. 25% (95% CI: 24–25%) with incomes greater than 52,000 £), exhibited lower physical activity levels (33% (95% CI: 33–34%) vs. 35% (95% CI: 35–36%) at high levels) and with less sunshine duration (128 (95% CI: 127–128) hours vs. 129 (95% CI: 128–129) hours).

3.2. Survival Analysis on the Association Between Sunshine Duration and Depression

The AC of covariates for both unweighted and weighted populations are presented in Table S5, illustrating a notable improvement in covariate balance after GPS weighting. The exposure–response curve showed that the risk of incident depression significantly decreased with an increase in sunshine duration (Figure 1). As shown in

Table 2. The association between sunshine duration and incident depression.

Model	Analysis 1		Analysis 2
	HR (95% CI)	p	β (95% CI)	p
Total population
Model 1	0.91 (0.90, 0.92)	<0.001	−0.13 (−0.20, −0.06)	<0.001
Model 2	0.92 (0.89, 0.95)	<0.001	−0.11 (−0.18, −0.04)	0.002
Model 3	0.96 (0.92, 0.97)	<0.001	−0.10 (−0.17, −0.03)	0.003
Model 4	0.92 (0.89, 0.95)	<0.001	−0.08 (−0.15, −0.01)	0.03
Low genetic risk
Model 1	0.89 (0.83, 0.95)	<0.001	−0.18 (−0.33, −0.03)	0.02
Model 2	0.92 (0.85, 0.98)	0.01	−0.16 (−0.32, −0.01)	0.04
Model 3	0.91 (0.85, 0.98)	0.01	−0.16 (−0.31, −0.01)	0.04
Model 4	0.90 (0.85, 0.96)	0.001	−0.14 (−0.29, −0.01)	0.03
Medium genetic risk
Model 1	0.90 (0.86, 0.93)	<0.001	−0.13 (−0.22, −0.04)	0.005
Model 2	0.92 (0.89, 0.96)	<0.001	−0.11 (−0.20, −0.02)	0.02
Model 3	0.92 (0.87, 0.96)	<0.001	−0.11 (−0.20, −0.02)	0.02
Model 4	0.92 (0.87, 0.96)	<0.001	−0.08 (−0.17, −0.01)	0.04
High genetic risk
Model 1	0.89 (0.84, 0.94)	<0.001	−0.11 (−0.22, 0.01)	0.23
Model 2	0.91 (0.87, 0.97)	0.002	−0.08 (−0.15, −0.02)	0.02
Model 3	0.92 (0.87, 0.97)	0.002	−0.04 (−0.08, −0.01)	0.04
Model 4	0.92 (0.87, 0.97)	0.003	−0.04 (−0.07, −0.01)	0.01

, Hazard Ratios (HRs) per 30 h increment of monthly sunshine duration were 0.92 (95% CI: 0. 89, 0. 95), 0.90 (95% CI: 0.85, 0.96), 0.92 (95% CI: 0.87, 0.96), and 0.92 (95% CI: 0.87, 0.97) for the total, low genetic risk, medium genetic risk and high genetic risk populations in the main model (Model 4). Similarly, negative associations were observed between sunshine duration and PHQ-9 score. The PHQ-9 score decreased −0.08 (95% CI: −0.15, −0.01), −0.14 (95% CI: −0.29, −0.01), −0.08 (95% CI: −0.17, −0.01), and −0.04 (95% CI: −0.07, −0.01) points per 30 h increment of monthly sunshine duration in total, low genetic risk, medium genetic risk and high genetic risk populations.

3.3. Stratification and Interaction Analysis

We found that an increase in sunshine duration significantly reduced the risk of incident depression across various stratification groups, as detailed in

Table 3. The results of stratification and interaction analysis.

Group	Hazard Ratio (95% C)	p for Interaction
Gender
Female	0.94 (0.91, 0.96)	Ref
Male	0.95 (0.91, 0.98)	0.31
Age
Young	0.97 (0.94, 0.99)	Ref
Old	0.94 (0.92, 0.96)	0.02
Household income
Less than 18,000 £	1.02 (0.98, 1.07)	Ref
18,000 to 30,999 £	0.95 (0.91, 0.99)	0.58
31,000 £ to 51,999 £	0.90 (0.86, 0.94)	0.08
Greater than 52,000 £	0.84 (0.79, 0.89)	0.25
Alcohol intake
Never	1.03 (0.96, 1.11)	Ref
Occasional	0.97 (0.93, 1.02)	0.14
Moderate	0.90 (0.87, 0.93)	<0.001
Heavy	0.94 (0.90, 0.99)	0.02
Smoking status
Never smoked	0.94 (0.91, 0.97)	Ref
Previous smoked	0.94 (0.90, 0.97)	0.94
Current smoked	0.93 (0.88, 0.99)	0.87
Time spent outdoors
Low	0.92 (0.88, 0.95)	Ref
Moderate	0.95 (0.91, 0.99)	0.02
High	0.97 (0.96, 0.98)	0.001
Physical activity
Low	0.92 (0.87, 0.96)	Ref
Moderate	0.97 (0.94, 0.99)	0.03
High	0.92 (0.88, 0.95)	0.82
Residential area
Rural	0.94 (0.92, 0.96)	Ref
Urban	0.94 (0.88, 0.99)	0.70
Artificial light at night
Low	0.94 (0.91, 0.98)	Ref
Medium	0.90 (0.87, 0.94)	0.08
High	0.96 (0.94, 0.99)	0.19
PRS
Low	0.90 (0.85, 0.96)	Ref
Medium	0.92 (0.87, 0.96)	0.19
High	0.92 (0.87, 0.97)	0.21

. The association between sunshine duration and incident depression was particularly pronounced among individuals over 65 years of age, those with moderate to heavy alcohol consumption, those who spent low amounts of time outdoors, and those engaging in low levels of physical activity.

3.4. Mediation Analyses

3.4.1. Biochemical Markers

As illustrated in Figure 2 and Table S6, vitamin D, PLR, platelet count, and calcium were positively associated with sunshine duration in the fully adjusted model. Conversely, biomarkers such as neutrophils, lymphocytes, leukocytes, INFLA, and CRP demonstrated a negative association with sunshine duration. Additionally, calcium and vitamin D were negatively associated with the risk of depression, whereas red cell distribution width, NLR, neutrophils, leukocytes, INFLA, and CRP were positively related to the risk of incident depression. Vitamin D, PLR, neutrophils, monocytes, leukocytes, INFLA, eosinophils, basophils and calcium were identified as mediators of the effect (Figure 2 and Table S7). Among these, vitamin D, calcium, INFLA, neutrophils, and monocytes emerged as the top five contributors to the natural indirect effect.

3.4.2. Sleep Pattern

Sunshine duration exhibited a significant association with the reduced risk of unhealthy sleep pattern. A 30 h increase in monthly sunshine duration was associated with 1.02-unit decrease in the sleep score. Unhealthy sleep pattern, in turn, demonstrated a detrimental impact on mental health, with a unit increase in the sleep score significantly associated with 1.28-fold increase in the risk of incident depressive disorders (95% CI: 1.26, 1.31) (Table S9). Furthermore, sleep pattern was identified as a mediator, accounting for 9.0% (95% CI: 7.2%, 14.0%) of the association between sunshine duration and depressive disorders after adjusting for potential confounders (Figure 3).

3.5. Parallel Mediation Analysis

As shown in Figure 3, the combined mediating effect of selected biomarkers and sleep pattern accounted for 30.0% of the total effect of sunshine duration on the risk of incident depression.

3.6. Sensitivity Analyses

The E-value was summarized in Table S10, showing that our results were overall robust to residual confounding. The results remained robust when we excluded participants with depressions or death within one year of follow-up, conducted the mediation analysis using the method developed by Huang et al., further including chronic diseases and vitamin D intake as covariates, used educational level as a surrogate for household income, included latitude group into covariates, included the skin pigmentation as the covariate, and included season as the covariate (Tables S8 and S10).

4. Discussion

We provided evidence for the protective role of sunshine duration on depression. Employing a causal framework model, we observed a significant reduction in depression risk with increased sunshine duration across different polygenetic risk groups. Mediation analyses indicated the potential causal role of biochemical markers and sleep pattern as mediators in the pathway from sunshine duration to the risk of incident depression. The effects were stronger among the elderly, alcohol consumers, individuals who spent less time outdoors, and those who were less physically active.

The existing evidence regarding the impact of sunshine duration on depression is currently limited and inconclusive. Some studies have reported no significant associations between sunlight and depression [6,39]. In contrast, an ecological study conducted in Anhui Province, China, identified a U-shaped exposure–response association, implying that both excessive and insufficient sunlight may have detrimental effects on psychological health ⁵. Notably, a recent study based on the UK Biobank cohort observed a nonlinear (J-shaped) relationship between time spent in outdoor light and the risk of depression, indicating potential benefits of moderate exposure while highlighting possible risks at both extremes [40]. Unlike our study, which used satellite-based sunshine duration as an objective environmental exposure, this study relied on self-reported time spent in outdoor light as a proxy for sunlight exposure. This distinction in exposure assessment may account for the differences in findings. There are also some studies supporting our findings. For example, one study in Korea reported that increased sunshine duration was linked with a decreased risk of depression [7]. Similarly, another study demonstrated that prolonged exposure to sunlight led to a reduction in depression using a negative binomial regression model [41]. The inconsistencies in results may stem from variations in study design, methodologies, and study populations. Unlike prior research studies, which were mainly based on ecological or cross-sectional designs, our study was conducted by a national, longitudinal cohort and used a model with a causal framework that may provide stronger evidence for causal inference.

Stratified analyses indicated that the effects of sunshine duration on incident depressive disorders were more pronounced in individuals over 65 years of age, moderate to heavy alcohol consumers, those who spend less time outdoors, and those who engage in lower levels of physical activity. One possible explanation is that vitamin D deficiency is more prevalent among the elderly, and sunlight exposure is a primary source of vitamin D synthesis [42]. Therefore, prolonged exposure to sunlight may confer greater benefits in this population. Furthermore, individuals who spend less time outdoors and engage in irregular physical activity have decreased opportunities for sunlight exposure, which may contribute to their higher risk of depression. However, given that physical activity itself has well-documented antidepressant effects, future studies should consider disentangling the independent contributions of sunlight exposure and exercise. Additionally, previous research has suggested that limited sunlight exposure may lead to increased alcohol consumption, potentially due to its effects on mood regulation [43]. This raises the possibility that alcohol intake may partially influence the relationship between sunshine duration and depression risk, although further investigation is needed to clarify this interaction.

Biochemical markers in peripheral blood serve as essential biomarkers reflecting potential structural or functional changes within the human body influenced by exogenous factors. Sunlight is involved in vitamin D synthesis, and vitamin D facilitates the absorption of calcium in the gut [44,45]. These biomarkers have been reported to be involved in the regulation of neurotransmitter synthesis and release, including serotonin and dopamine, which play crucial roles in the onset and progression of depression [46,47]. Studies have indicated that the activated form of vitamin D, 25-hydroxyvitamin D, can up-regulate erythropoietin receptor expression and collaborate with erythropoietin to promote erythropoietic cell proliferation [48]. Furthermore, sunshine has been proposed to elicit anti-inflammatory systemic effects, and inflammation has been implicated in depression. These pieces of evidence support the mediating effects of inflammatory biomarkers observed in our study [49].

Sunshine exposure is also proposed to influence the onset of depression through its impact on sleep pattern, potentially achieved by regulating circadian rhythms and health behaviors [50,51,52]. Sunlight exposure during the early morning or late evening plays a significant role in influencing the melatonin precursor, serotonin, with melatonin being a key signaling factor in circadian rhythm regulation [9,53]. As a result, aberrant exposure to sunshine may disrupt circadian rhythms, which can lead to sleep disorders and, in turn, affect the onset of depression [9,54]. Moreover, daily physical activity has also been linked to weather conditions. For instance, one study reported that more sunshine hours led to an increase in physical activity and the time spent outside the home among older adults [55]. These changes in human behaviors induced by sunshine duration may consequently impact sleep pattern and mental health. In addition, emerging evidence suggests that sunshine exposure, through its role in promoting vitamin D synthesis, may indirectly affect mental health via the gut–brain axis [56]. Specifically, vitamin D deficiency has been shown to alter the gut microbiome, reducing the production of B vitamins such as pantothenic acid, which are essential for immune regulation and sleep function [57,58]. This suggests a possible biological pathway linking inadequate sunshine to sleep disturbances and immune dysfunction, thereby contributing to depression onset.

Our study has several strengths. First, we investigated the effects of long-term exposure to sunshine duration on the incidence of depression in the UK Biobank. The UK Biobank, being a national cohort study with a large sample size, significantly enhanced statistical power. Second, we used a time-dependent Cox proportional hazard model combined with propensity scoring to provide evidence for causal inference. Lastly, our mediation analysis provides evidence of potential mechanisms for depression including vitamin D metabolism and inflammation.

Although sunshine duration is an uncontrollable environmental factor, our study identified several effect modifiers and potential mediators that may help inform targeted public health strategies. We found that the associations between sunshine duration and depression were modified by age, time spent outdoors, physical activity, and alcohol consumption. These findings suggest that individuals living in high-latitude regions should be particularly aware of the potential mental health risks associated with limited sunlight exposure. In particular, older adults and those experiencing shorter daylight periods during colder seasons may benefit from engaging in more outdoor activities and reducing alcohol intake.

Moreover, our mediation analysis indicated that the relationship between sunshine duration and depression may be partially mediated by changes in vitamin D metabolism, sleep patterns and systemic inflammation. This underscores the importance of promoting vitamin D supplementation, anti-inflammatory behaviors, and maintaining a regular sleep–wake cycle during periods of limited sunlight. For instance, increasing the intake of vitamin D–rich and anti-inflammatory foods, as well as adopting a consistent early-to-bed, early-to-rise routine, may help mitigate the potential adverse effects of insufficient sunshine exposure.

There are also some limitations to the present study. First, the data on biochemical markers and sleep pattern were mainly assessed at baseline, constraining our ability to establish the chronological relationship between exposure and the mediators. Second, although meteorological variables, air pollution, and outcomes were time-varying, information on other covariates throughout the follow-up period was not available. Third, since exposure was assessed based on residential location rather than direct individual-level data, the absence of information on actual sunlight exposure, such as duration of outdoor activity and the potential influence of building shading, may have led to exposure misclassification. Finally, this study focused mainly on environmental factors, while depression is influenced by a variety of factors, including social and psychological elements, which were not considered.

Future studies should aim to more precisely assess individual-level sunlight exposure by collecting accurate outdoor activity durations or utilizing personal wearable devices to monitor light exposure. Additionally, longitudinal cohort studies with repeated measurements of exposures, outcomes, and covariates over time would allow for evaluation of how changes in these variables influence health outcomes. Furthermore, it would be valuable to investigate the dynamic effects of sunshine exposure on the progression of depression among individuals with pre-existing depressive disorders.

5. Conclusions

In conclusion, sunshine duration may have protective effects against depressive disorders across different genetic risk groups, primarily mediated through vitamin D metabolism, sleep patterns, and inflammation. Longer sunshine exposure appears to confer greater benefits, especially among the elderly, alcohol consumers, individuals who spend less time outdoors, and those with lower physical activity levels. These findings suggest that targeting potential mediators and modifiers may help prevent the adverse effects of insufficient sunlight exposure on depression.