The joint and interaction associations of physical activity and sleep with depression: a cohort study Authors: Zhemin Fu, Yuwei Wang, Boxiao Gu, Xinyuan Zhao, Yifanlin Yang, Shengmei Cui, Zhixing Fan Published: 2026 ABSTRACT AbstractBackgroundDespite substantial evidence regarding independent associations of physical activity and sleep quality with depression, their joint and interactive effects remain inadequately characterized. This study aimed to investigate the independent, joint, and interactive associations of physical activity and sleep quality with incident depression in a large prospective cohort.MethodsThis prospective cohort study included 323,910 depression-free UK Biobank participants recruited between 2006 and 2010, with a median follow-up of 13.44 years. Physical activity was assessed using the International Physical Activity Questionnaire-Short Form, measuring total (TPA), moderate (MPA), and vigorous physical activity (VPA). Sleep quality was evaluated using six traits to construct a comprehensive score (0–6). Incident depression was ascertained through ICD-10 codes from linked health records. Cox proportional hazards regression models explored the independent, joint, and interactive associations of physical activity and sleep quality with depression risk.ResultsDuring follow-up, 16,983 (5.24%) participants developed depression. Both physical activity and sleep quality demonstrated significant associations with depression. Participants with the lowest TPA tertile and poor sleep quality (HR: 3.802, 95% CI: 3.405–4.247) showed substantially higher depression risk than those with recommended TPA and high sleep quality. Compared with participants with recommended VPA (600–1200 MET-minutes/week) and high sleep quality, those with VPA < 600 MET-minutes/week (HR: 3.626, 95% CI: 3.137–4.190) or VPA > 1200 MET-minutes/week (HR: 3.338, 95% CI: 2.856–3.901) combined with low sleep quality had markedly elevated depression risk. Physical activity’s protective associations were substantially attenuated in poor sleep contexts, whereas poor sleep’s adverse effects persisted across all physical activity levels. A modest but statistically significant additive interaction was observed specifically between VPA and sleep quality (P < 0.05), while no significant interactions were detected for TPA or MPA. Multiplicative interactions were not statistically significant for any physical activity measure.ConclusionThere were independent and joint associations of physical activity and sleep quality with incident depression, with a modest additive interaction observed specifically for vigorous physical activity and sleep quality. Physical activity’s protective associations are attenuated in the context of poor sleep, while sleep quality’s adverse associations persists across all activity levels.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28119-0.Keywords: Physical activity, Sleep quality, Depression, Joint and interactive effects, UK Biobank Introduction Depression represents a major global public health challenge that significantly contributes to the worldwide burden of disease. According to the Global Burden of Disease 2021 study, the global prevalence of depressive disorders reached 332.4 million cases in 2021, accounting for 9.1% of all diseases and a substantial proportion of mental health disorders [1]. Between 1990 and 2021, the age-standardized disability-adjusted life years rate for depressive disorders increased by 13.4%, with population growth being the primary driver of this escalation [1, 2]. The burden varies considerably across different populations, with females experiencing substantially higher rates than males, and marked disparities exist between regions of different sociodemographic index (SDI) levels [2, 3]. Projections indicate that by 2040, global cases of depressive disorders will exceed 466 million, with middle- and low-SDI regions contributing the majority of this burden [4]. Among older adults specifically, over one-third of the global population experiences depression, highlighting the urgent need for effective preventive strategies and timely interventions [5]. These alarming epidemiological trends underscore the critical importance of identifying modifiable risk factors that could inform targeted prevention and intervention programs to reduce the growing burden of depression worldwide. The protective role of physical activity against depression is now well-established, supported by numerous large-scale prospective cohort studies and quantified by comprehensive meta-analyses. For instance, comprehensive meta-analysis of 49 prospective studies involving 266,939 participants revealed that individuals with high levels of physical activity had 17% lower odds of developing depression compared to those with low activity levels, with protective effects observed across all age groups and geographical regions [6]. Similarly, another systematic review and meta-analysis of 15 studies encompassing 191,130 participants found an inverse curvilinear dose-response relationship, indicating that adults who accumulated half the recommended volume of physical activity (4.4 metabolic equivalent task hours per week) had 18% lower depression risk, while those meeting full recommendations demonstrated 25% risk reduction [7]. Recent large-scale studies have further elucidated the optimal dose and pattern of physical activity for depression prevention. An analysis of 74,715 UK Biobank participants showed that higher doses of moderate-to-vigorous physical activity were curvilinearly associated with lower depression risk, with the greatest incremental benefits occurring during the first 4–5 h per week and optimal benefits around 15 h weekly [8]. Notably, even patterns such as the “weekend warrior” approach, where physical activity is concentrated into one or two sessions per week, have demonstrated comparable antidepressant benefits to regular activity patterns [9, 10]. Parallel to the evidence on physical activity, a substantial body of research has identified sleep disturbances as a critical and modifiable risk factor for depression, with studies consistently demonstrating a robust bidirectional relationship between the two [11, 12]. Large-scale prospective studies have consistently demonstrated that various dimensions of sleep, including duration, quality, and regularity, independently predict depression risk across different age groups. A recent analysis of 79,666 UK Biobank participants found that individuals with regular sleep patterns had 38% lower depression risk compared to irregular sleepers, with this protective effect persisting even among those meeting recommended sleep duration guidelines [13]. Similarly, sleep disorders in childhood and adolescence have been shown to substantially increase subsequent depression risk, with German youth diagnosed with sleep disorders exhibiting more than twice the hazard of developing depression within five years [14]. Beyond duration and regularity, specific sleep characteristics such as chronotype and sleep architecture also play important roles, with evening chronotype being associated with higher depression risk through mechanisms involving sensation seeking and poor sleep quality [15]. Notably, both short and long sleep durations have been implicated in depression risk, with long sleep duration (≥ 9 h/night) showing particularly strong associations with depressive symptoms in individuals already experiencing depression [16]. Despite the robust evidence linking physical activity and sleep independently with depression risk, critical gaps remain in understanding their interaction. Prior joint analyses have largely relied on mediation frameworks or descriptive co-exposure categorizations, without formally quantifying additive interaction or characterizing the directional asymmetry of the two factors’ interplay. Whether physical activity’s protective effects are differentially attenuated by poor sleep—or vice versa—remains unresolved. Therefore, the present study aimed to investigate the independent, joint, and interactive associations of physical activity and sleep quality with incident depression in a large prospective cohort from the UK Biobank. Method Participants The UK Biobank is a large-scale prospective population-based cohort study that recruited approximately 502,175 participants aged 40–69 years from 22 assessment centers across England, Scotland, and Wales between 2006 and 2010. At baseline, participants underwent comprehensive assessments including questionnaires, physical measurements, and biological sample collection. Detailed information on the study design, recruitment procedures, and methodology has been published elsewhere [17]. At baseline, trained research nurses and assessment center staff conducted all assessments at dedicated UK Biobank assessment centers. Data collection included standardized touchscreen questionnaires completed by participants, physical measurements conducted by staff using calibrated equipment, and biological sample collection. Physical activity and sleep information were self-reported by participants via touchscreen questionnaires at the assessment center visit. The UK Biobank received ethical approval from the North West Multi-Centre Research Ethics Committee, and all participants provided written informed consent for data collection and future analyses. For the present analysis, participants were sequentially excluded based on the following criteria: (1) missing sleep trait data including snoring, chronotype, daytime sleepiness, sleep duration, insomnia symptoms, and ease of morning awakening (n = 91,961); (2) missing physical activity information on total physical activity (TPA), moderate physical activity (MPA), or vigorous physical activity (VPA) measured in metabolic equivalent of task (MET) levels (n = 83,525), resulting in 326,689 participants. We further excluded those with a history of depression at baseline (n = 2,779). The final analytical cohort comprised 323,910 depression-free participants with complete physical activity and sleep data (Figure S1). Physical activity Physical activity was assessed using the International Physical Activity Questionnaire-Short Form (IPAQ-SF) during baseline recruitment, a validated instrument for capturing frequency, duration, and intensity of walking, moderate-intensity, and vigorous-intensity activities over the preceding seven days [18]. Physical activity volume was quantified as metabolic equivalent of task (MET)-minutes/week by weighting time spent in each activity domain by its energy requirement. Total physical activity (TPA) was calculated as the sum of MET-minutes across all domains. For primary analyses, TPA was categorized into tertiles (Q1, Q2, Q3) with the middle tertile as reference, allowing simultaneous assessment of associations at both lower and higher activity levels relative to a moderate baseline—an approach more informative than a lowest-tertile reference when a non-linear relationship is hypothesized a priori based on prior literature [19]. Moderate physical activity (MPA) and vigorous physical activity (VPA) were classified according to World Health Organization guidelines [20] for adults (150–300 min of moderate-intensity or 75–150 min of vigorous-intensity activity weekly, equivalent to 600–1200 MET-minutes/week) into three categories: <600, 600–1200, and > 1200 MET-minutes/week, with 600–1200 designated as the reference group representing minimum recommended activity levels. Sleep quality Sleep quality was evaluated using six self-reported sleep characteristics previously validated in UK Biobank research [21]: snoring, chronotype, daytime sleepiness, sleep duration, insomnia symptoms, and ease of morning awakening. Each sleep trait was dichotomized into low-risk and high-risk categories based on established associations with health outcomes (Table S1). Low-risk sleep patterns were defined as: absence of snoring (“No”), early chronotype (preferring “morning” or “morning rather than evening”), infrequent daytime sleepiness (“never/rarely” or “sometimes”), optimal sleep duration (7–8 h per day), infrequent insomnia symptoms (“never/rarely” or “sometimes”), and ease of morning awakening (“fairly easy” or “very easy”). A comprehensive sleep quality score was constructed by assigning one point for each low-risk sleep characteristic and zero points for each high-risk characteristic, yielding a continuous score ranging from 0 (poorest sleep quality) to 6 (optimal sleep quality). For analytical purposes, participants were further stratified into three sleep quality categories: high (score 5–6), medium (score 3–4), and low (score 0–2), with higher scores indicating better overall sleep quality. Depression The outcome was incident depression ascertained through linkage to multiple health records in the UK Biobank. Depression diagnoses were identified using International Classification of Diseases, 10th Revision (ICD-10) codes F32 (depressive episode) and F33 (recurrent depressive disorder), captured from hospital inpatient admissions, primary care electronic health records, and national death registries. Follow-up time was calculated from the date of baseline assessment to whichever occurred first: date of first depression diagnosis, date of death, date lost to follow-up, or administrative censoring date. The administrative censoring dates varied by country: September 30, 2021 for England, July 31, 2021 for Scotland, and February 28, 2018 for Wales, reflecting the most recent available data linkage at the time of analysis. Covariate Covariates were selected a priori based on established associations with both physical activity, sleep, and depression risk, and their potential role as confounders in the exposure-outcome pathway [22–24]. Sociodemographic characteristics included age (continuous variable in years), sex (female or male), ethnicity (white or other), educational attainment (university degree or no university degree), and household income (annual income <£18,000 or ≥£18,000), given that socioeconomic disadvantage and demographic factors are consistently associated with both lifestyle behaviors and mental health outcomes [22]. Lifestyle and behavioral factors comprised smoking status (never, previous, or current smoker), alcohol consumption status (never, previous, or current drinker), adherence to the Dietary Approaches to Stop Hypertension (DASH) diet (yes or no), and body mass index (BMI, calculated as weight in kilograms divided by height in meters squared), as these factors independently predict depression risk and are associated with both physical activity and sleep [23]. The DASH diet emphasizes high intake of fruits, vegetables, whole grains, low-fat dairy, nuts, and legumes, and low intake of red meat, sweets, sugary drinks, and sodium. As detailed in Table S2, participants scored 1 point for meeting each component target and 0 otherwise, yielding a total score from 0 to 10. Adherence was defined as a total score ≥ 5 (“Yes”) versus < 5 (“No”). Pre-existing medical conditions were assessed through self-reported physician diagnoses and included history of diabetes mellitus, hypertension, cardiovascular disease (CVD), and cancer, each coded as binary variables (yes or no), as these comorbidities are known to be associated with both reduced physical activity capacity and increased depression risk [24]. Statistical analysis Missing values for covariates were addressed using multiple imputation with chained equations. Continuous variables were presented as means with standard deviations and compared using Student’s t-tests, while categorical variables were expressed as frequencies with percentages and compared using chi-square tests. Independent associations between physical activity, sleep quality, and incident depression were conducted by Cox proportional hazards regression models to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Three sequential models were constructed: Model 1 adjusted for age, sex, ethnicity, education, and income; Model 2 additionally adjusted for smoking status, alcohol consumption, and DASH diet; Model 3 further adjusted for BMI and history of diabetes mellitus, hypertension, cardiovascular disease, and cancer. In the mutual adjustment models, sleep quality and TPA were included simultaneously. For MPA and VPA analyses, additional adjustments were made for other physical activity components and walking. The proportional hazards assumption was verified using Schoenfeld residuals. Dose-response relationships between continuous physical activity levels (TPA, MPA, VPA) and sleep quality scores with depression risk were examined using restricted cubic spline (RCS) regression with three knots positioned at the 10th, 50th, and 90th percentiles. Nonlinearity was tested using the likelihood ratio test comparing models with and without the cubic spline terms. Joint associations of physical activity and sleep quality with depression risk were evaluated by creating combined exposure categories. Participants with recommended physical activity levels (TPA Q2, MPA/VPA 600–1200 MET-minutes/week) and high sleep quality served as the reference group. Both multiplicative and additive interactions were assessed. HRs for all combined exposure categories were estimated using fully adjusted Cox models. Both multiplicative and additive interactions were assessed. Multiplicative interaction was tested by including product terms between physical activity categories and sleep quality categories in Cox regression models, with statistical significance determined by the Wald test. Additive interaction was evaluated using three measures: relative excess risk due to interaction (RERI), attributable proportion due to interaction (AP), and synergy index (SI), calculated from the HRs of the joint exposure categories. Presence of additive interaction was indicated by RERI > 0, AP > 0, or SI > 1. Stratified analyses were conducted to examine whether the association between physical activity and depression varied across sleep quality subgroups (high, medium, low), and conversely, whether the sleep-depression association differed by physical activity levels. Interaction P-values were calculated from likelihood ratio tests. Additive interactions between physical activity and each of the six individual sleep traits were also examined. Three sensitivity analyses were performed to assess robustness. First, we repeated all analyses after excluding participants with missing covariate data to evaluate the impact of multiple imputation. Second, we excluded participants who developed depression within the first two years of follow-up to minimize potential reverse causation whereby prodromal depressive symptoms might influence physical activity or sleep behaviors. Third, we excluded participants with a history of cardiovascular disease at baseline, as this comorbidity might confound the associations of interest. All statistical analyses were conducted using R software version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria). Two-sided P-values < 0.05 were considered statistically significant. This study was an exploratory analysis and multiple corrections were not performed. Result Baseline characteristic The final analytical cohort comprised 323,910 participants (mean age 56.18 ± 8.12 years; 52.68% female) free from depression at baseline, among whom 16,983 (5.24%) developed incident depression during a median follow-up of 13.44 years (Table 1). Participants who developed depression were more likely to be female, have lower socioeconomic status, and exhibit unhealthier lifestyle profiles including current smoking and obesity (all P < 0.001). All examined comorbidities including diabetes mellitus, hypertension, cardiovascular disease, and cancer were significantly more prevalent among those who developed depression (all P < 0.001). Table 1Baseline characteristics of participants stratified by depressionCharacteristicsLevelOverall(n = 323910)Control(n = 306927)Depression(n = 16983) P Age (year)56.18 (8.12)56.18 (8.12)56.33 (8.16)0.014< 65 year266,171 (82.17)252,437 (82.25)13,734 (80.87)< 0.001≥ 65 year57,739 (17.83)54,490 (17.75)3249 (19.13)Sex (%)Female170,628 (52.68)160,093 (52.16)10,535 (62.03)< 0.001Male153,282 (47.32)146,834 (47.84)6448 (37.97)Race (%)Other15,425 ( 4.76)14,745 ( 4.80)680 ( 4.00)< 0.001White308,485 (95.24)292,182 (95.20)16,303 (96.00)Education (%)No university degree207,399 (64.03)195,083 (63.56)12,316 (72.52)< 0.001University degree116,511 (35.97)111,844 (36.44)4667 (27.48)Income (%)<£18,00062,921 (19.43)57,368 (18.69)5553 (32.70)< 0.001>£18,000260,989 (80.57)249,559 (81.31)11,430 (67.30)Smoke (%)Never177,450 (54.78)169,742 (55.30)7708 (45.39)< 0.001Previous114,499 (35.35)108,046 (35.20)6453 (38.00)Current31,961 ( 9.87)29,139 ( 9.49)2822 (16.62)Alcohol (%)Never12,048 ( 3.72)11,246 ( 3.66)802 ( 4.72)< 0.001Previous10,377 ( 3.20)9243 ( 3.01)1134 ( 6.68)Current301,485 (93.08)286,438 (93.32)15,047 (88.60)DASH (%)No130,438 (40.27)123,331 (40.18)7107 (41.85)< 0.001Yes193,472 (59.73)183,596 (59.82)9876 (58.15)Sleep score4.01 (1.13)4.03 (1.12)3.55 (1.21)< 0.001High115,128 (35.54)111,200 (36.23)3928 (23.13)< 0.001Medium202,777 (62.60)190,579 (62.09)12,198 (71.82)Low6005 ( 1.85)5148 ( 1.68)857 ( 5.05)TPA (MET min/week)2654.14 (2643.94)2658.57 (2635.85)2574.08 (2784.87)< 0.001Q2162,073 (50.04)154,418 (50.31)7655 (45.07)< 0.001Q380,738 (24.93)76,569 (24.95)4169 (24.55)Q181,099 (25.04)75,940 (24.74)5159 (30.38)MPA (MET-minutes/week)938.96 (1213.83)938.61 (1210.93)945.33 (1265.21)0.483600–120055,074 (17.00)52,443 (17.09)2631 (15.49)< 0.001> 120088,081 (27.19)83,432 (27.18)4649 (27.37)< 600180,755 (55.80)171,052 (55.73)9703 (57.13)VPA (MET-minutes/week)680.36 (1164.47)684.10 (1160.62)612.70 (1230.01)< 0.001600–120047,562 (14.68)45,531 (14.83)2031 (11.96)< 0.001> 120062,597 (19.33)59,801 (19.48)2796 (16.46)< 600213,751 (65.99)201,595 (65.68)12,156 (71.58)BMI (kg/m2)27.26 (4.66)27.19 (4.60)28.56 (5.55)< 0.001< 25 kg/m2109,982 (33.95)105,411 (34.34)4571 (26.92)< 0.00125 ~ 29 kg/m2139,236 (42.99)132,413 (43.14)6823 (40.18)≥ 30 kg/m274,692 (23.06)69,103 (22.51)5589 (32.91)History of diabetes mellitus (%)No307,679 (94.99)292,136 (95.18)15,543 (91.52)< 0.001Yes16,231 ( 5.01)14,791 ( 4.82)1440 ( 8.48)History of hypertension (%)No227,803 (70.33)217,488 (70.86)10,315 (60.74)< 0.001Yes96,107 (29.67)89,439 (29.14)6668 (39.26)History of CVD (%)No299,661 (92.51)284,691 (92.76)14,970 (88.15)< 0.001Yes24,249 ( 7.49)22,236 ( 7.24)2013 (11.85)History of cancer (%)No294,405 (90.89)279,248 (90.98)15,157 (89.25)< 0.001Yes29,505 ( 9.11)27,679 ( 9.02)1826 (10.75)DASH dietary approaches to stop hypertension, BMI body mass index, TPA Total physical activity, MPA Moderate physical activity, VPA Vigorous physical activity Baseline characteristics of participants stratified by depression DASH dietary approaches to stop hypertension, BMI body mass index, TPA Total physical activity, MPA Moderate physical activity, VPA Vigorous physical activity Figure 1 illustrates the incidence of depression across physical activity and sleep quality categories. Depression incidence decreased progressively with higher physical activity levels across all three domains (TPA, MPA, VPA), ranging from 6.36% in the lowest to 4.72% in the middle tertile for total physical activity (Fig. 1A-C). Sleep quality demonstrated a strong dose-response relationship, with depression incidence declining from 17.53% at a sleep score of 0 to 2.51% at a score of 6 (Fig. 1D). Individual sleep traits showed substantial disparities, with high-risk categories exhibiting markedly elevated depression incidence compared to low-risk categories (Fig. 1E). Incidence of depression in the physical activity and sleep groups. A-C: Incidence of depression in the TPA, MPA, and VPA groups; D: Incidence of depression in the sleep score groups; E: Incidence of depression in the 6 sleep traits group. TPA: Total physical activity; MPA: Moderate physical activity; VPA: Vigorous physical activity Independent associations of physical activity and sleep with depression In fully adjusted models (Model 3), all physical activity measures showed significant associations with incident depression (Table 2). For TPA, the lowest tertile exhibited 17.8% higher depression risk compared to the middle tertile (HR = 1.178, 95% CI: 1.137–1.221), while the highest tertile showed modest 7.4% increased risk (HR = 1.074, 95% CI: 1.034–1.116). VPA below 600 MET-minutes/week conferred 14.1% elevated risk (HR = 1.141, 95% CI: 1.089–1.197), while exceeding 1200 MET-minutes/week showed 6.4% increased risk (HR = 1.064, 95% CI: 1.004–1.126). Both MPA below 600 MET-minutes/week (HR = 1.066, 95% CI: 1.020–1.113) and above 1200 MET-minutes/week (HR = 1.066, 95% CI: 1.016–1.118) demonstrated similarly elevated risk compared to the 600–1200 reference category. Restricted cubic spline analyses revealed significant non-linear relationships (Fig. 2): TPA displayed a J-shaped curve, while MPA and VPA exhibited U-shaped associations (all P-nonlinear < 0.001). Table 2Independent associations of physical activity and sleep with depression riskExposureCrude modelModel 1Model 2Model 2HR(95%CI) P HR(95%CI) P HR(95%CI) P HR(95%CI) P TPA (MET-minutes/week) Q2refrefrefref Q31.098(1.057,1.140)< 0.00011.040(1.001,1.080)0.0421.051(1.012,1.091)0.0111.074(1.034,1.116)< 0.001 Q11.356(1.309,1.405)< 0.00011.349(1.303,1.398)< 0.00011.245(1.201,1.290)< 0.00011.178(1.137,1.221)< 0.0001MPA (MET-minutes/week) 600–1200refrefrefref >12001.109(1.058,1.164)< 0.00011.060(1.011,1.112)0.0171.055(1.005,1.106)0.0301.066(1.016,1.118)0.009 <6001.125(1.077,1.174)< 0.00011.158(1.109,1.209)< 0.00011.102(1.055,1.150)< 0.00011.066(1.020,1.113)0.004VPA (MET-minutes/week) 600–1200refrefrefref >12001.048(0.990,1.110)0.1081.042(0.984,1.103)0.1581.047(0.988,1.108)0.1181.064(1.004,1.126)0.035 <6001.340(1.279,1.405)< 0.00011.270(1.212,1.332)< 0.00011.192(1.137,1.249)< 0.00011.141(1.089,1.197)< 0.0001Sleep quality Highrefrefrefref Medium1.789(1.725,1.854)< 0.00011.763(1.701,1.828)< 0.00011.688(1.628,1.750)< 0.00011.606(1.549,1.665)< 0.0001Low4.484(4.164,4.828)< 0.00014.194(3.894,4.517)< 0.00013.678(3.412,3.964)< 0.00013.198(2.965,3.448)< 0.0001Model 1 was adjusted for age, sex, race, education, and income; Model 2 was adjusted for Model 1 + smoking status, alcohol drinker status, and DASH diet; Model 3 was adjusted for Model 1 + BMI, and the history of diabetes mellitus, hypertension, CVD, and cancer. Sleep quality and TPA were mutually adjusted in respective model. MPA was further adjusted for sleep quality, VPA, and walking. VPA was further adjusted for sleep quality, MPA, and walkingDASH Dietary approaches to stop hypertension, BMI Body mass index, CVD Cardiovascular disease, TPA Total physical activity, MPA Moderate physical activity, VPA Vigorous physical activity Independent associations of physical activity and sleep with depression risk Model 1 was adjusted for age, sex, race, education, and income; Model 2 was adjusted for Model 1 + smoking status, alcohol drinker status, and DASH diet; Model 3 was adjusted for Model 1 + BMI, and the history of diabetes mellitus, hypertension, CVD, and cancer. Sleep quality and TPA were mutually adjusted in respective model. MPA was further adjusted for sleep quality, VPA, and walking. VPA was further adjusted for sleep quality, MPA, and walking DASH Dietary approaches to stop hypertension, BMI Body mass index, CVD Cardiovascular disease, TPA Total physical activity, MPA Moderate physical activity, VPA Vigorous physical activity Dose-response associations of sleep score (A) and physical activity (B-D) with depression risk. Models were adjusted for age, sex, race, education, income, smoking status, alcohol drinker status, DASH diet, BMI, and the history of diabetes mellitus, hypertension, CVD, and cancer. Sleep score and TPA were mutually adjusted in respective model. MPA was further adjusted for sleep score, VPA, and walking. VPA was further adjusted for sleep score, MPA, and walking. DASH: Dietary approaches to stop hypertension; BMI: Body mass index; CVD: Cardiovascular disease; TPA: Total physical activity; MPA: Moderate physical activity; VPA: Vigorous physical activity Sleep quality demonstrated strong dose-response associations independent of physical activity. Compared to high sleep quality, medium quality conferred 60.6% higher risk (HR = 1.606, 95% CI: 1.549–1.665), and low quality nearly tripled the risk (HR = 3.198, 95% CI: 2.965–3.448) (Table 2). The restricted cubic spline curve showed an inverse curvilinear relationship (P-nonlinear = 0.004). All six sleep traits independently predicted depression after full adjustment (Table S3), with difficulty getting up in morning showing the strongest association (HR = 1.997), followed by frequent daytime sleepiness (HR = 1.667), frequent insomnia (HR = 1.570), abnormal sleep duration (HR = 1.434), late chronotype (HR = 1.270), and snoring (HR = 1.100) (all P < 0.001). Joint associations of physical activity and sleep with depression Physical activity and sleep quality demonstrated substantial cumulative effects on incident depression risk (Fig. 3, Table S4). Using recommended physical activity levels combined with high sleep quality as reference, participants with the least favorable profiles showed the highest risks: low TPA (Q1) with low sleep quality conferred nearly 4-fold elevated risk (HR = 3.802, 95% CI: 3.405–4.247), while insufficient MPA or VPA (< 600 MET-minutes/week) with low sleep quality yielded similar risks (HR = 3.661 and 3.626, respectively). Depression risk increased progressively with worsening sleep quality within each physical activity category and with lower physical activity within each sleep quality stratum. The joint associations of physical activity and sleep with depression risk. A: The joint associations of TPA and sleep with depression risk; B: The joint associations of MPA and sleep with depression risk; C: The joint associations of VPA and sleep with depression risk. Models were adjusted for age, sex, race, education, income, smoking status, alcohol drinker status, DASH diet, BMI, and the history of diabetes mellitus, hypertension, CVD, and cancer. MPA was further adjusted for VPA and walking. VPA was further adjusted for MPA and walking. DASH: Dietary approaches to stop hypertension; BMI: Body mass index; CVD: Cardiovascular disease; TPA: Total physical activity; MPA: Moderate physical activity; VPA: Vigorous physical activity Interaction associations of physical activity and sleep with depression Stratified analyses revealed differential associations between physical activity and depression risk across sleep quality subgroups (Fig. 4, Figure S2). Among participants with high sleep quality, higher physical activity consistently conferred lower depression risk (TPA Q1 vs. Q2: HR = 1.217; VPA < 600 vs. 600–1200: HR = 1.179), but these protective associations were substantially attenuated in medium or low sleep quality groups. Restricted cubic spline analyses demonstrated varying dose-response patterns across strata (Fig. 5, Figure S3): among participants with high sleep quality, physical activity showed steeper risk gradients, while flatter curves in low sleep quality groups. Associations of physical activity with depression risk in sleep quality subgroups. Models were adjusted for age, sex, race, education, income, smoking status, alcohol drinker status, DASH diet, BMI, and the history of diabetes mellitus, hypertension, CVD, and cancer. MPA was further adjusted for VPA and walking. VPA was further adjusted for MPA and walking. DASH: Dietary approaches to stop hypertension; BMI: Body mass index; CVD: Cardiovascular disease; TPA: Total physical activity; MPA: Moderate physical activity; VPA: Vigorous physical activity RCS of physical activity with depression risk in sleep quality subgroups. A: RCS of TPA with depression risk in sleep quality subgroups; B: RCS of MPA with depression risk in sleep quality subgroups; C: RCS of VPA with depression risk in sleep quality subgroups. Models were adjusted for age, sex, race, education, income, smoking status, alcohol drinker status, DASH diet, BMI, and the history of diabetes mellitus, hypertension, CVD, and cancer. MPA was further adjusted for VPA and walking. VPA was further adjusted for MPA and walking. DASH: Dietary approaches to stop hypertension; BMI: Body mass index; CVD: Cardiovascular disease; TPA: Total physical activity; MPA: Moderate physical activity; VPA: Vigorous physical activity Formal interaction testing revealed no significant multiplicative interactions between TPA or MPA and sleep quality (P-interaction = 0.112 and 0.097, respectively), but VPA showed a borderline non-significant multiplicative interaction (P-interaction = 0.432) (Fig. 4). Additive interaction analyses, however, demonstrated significant synergistic effects specifically for VPA and sleep quality (Table 3). The relative excess risk due to interaction (RERI = 0.17, 95% CI: 0.05–0.28), attributable proportion (AP = 0.09, 95% CI: 0.02–0.16), and synergy index (SI = 1.26, 95% CI: 1.02–1.57) all indicated positive additive interaction. Neither TPA nor MPA demonstrated significant additive interactions with sleep quality. Table 3Interaction associations of physical activity and sleep with depression riskPhysical activitySleep qualityEffect of not Recommended VPA within the lower Sleep qualityEffect of lower Sleep quality within not Recommended VPAMultiplicativeInteractionAdditive interactionHigherLowerRERIAPSITPA1.12(1.09,1.16)1.63(1.55,1.71)0.96(0.89,1.03)0.04(-0.05,0.13)0.02(-0.03,0.07)1.05(0.94,1.16)Recommended1(Reference)1.7(1.61,1.79)Not Recommended1.17(1.1,1.25)1.91(1.81,2.01)MPA1.06(1.01,1.11)1.65(1.58,1.71)0.94(0.86,1.04)-0.02(-0.14,0.1)-0.01(-0.08,0.05)0.98(0.85,1.12)Recommended1(Reference)1.74(1.59,1.91)Not Recommended1.12(1.03,1.22)1.84(1.7,1.99)VPA1.15(1.09,1.22)1.68(1.61,1.74)1.08(0.97,1.19)0.17(0.05,0.28)0.09(0.02,0.16)1.26(1.02,1.57)Recommended1(Reference)1.56(1.41,1.72)Not Recommended1.07(0.98,1.17)1.8(1.65,1.96)Models were adjusted for age, sex, race, education, income, smoking status, alcohol drinker status, DASH diet, BMI, and the history of diabetes mellitus, hypertension, CVD, and cancer. MPA was further adjusted for VPA and walking. VPA was further adjusted for MPA and walkingDASH Dietary approaches to stop hypertension, BMI Body mass index, CVD Cardiovascular disease, TPA Total physical activity, MPA Moderate physical activity VPA Vigorous physical activity, RERI Relative excess risk of interaction, AP Attributable proportion of interaction, SI Synergy indexBold text indicates statistical significance Interaction associations of physical activity and sleep with depression risk Models were adjusted for age, sex, race, education, income, smoking status, alcohol drinker status, DASH diet, BMI, and the history of diabetes mellitus, hypertension, CVD, and cancer. MPA was further adjusted for VPA and walking. VPA was further adjusted for MPA and walking DASH Dietary approaches to stop hypertension, BMI Body mass index, CVD Cardiovascular disease, TPA Total physical activity, MPA Moderate physical activity VPA Vigorous physical activity, RERI Relative excess risk of interaction, AP Attributable proportion of interaction, SI Synergy index Bold text indicates statistical significance We also conducted additive interactions between physical activity and six sleep traits (Table S5-S10). Notably, VPA demonstrated positive additive interactions with daytime doze (RERI = 0.39, 95% CI: 0.00-0.77; AP = 0.20, 95% CI: 0.01–0.39) and difficulty getting up in the morning (RERI = 0.33, 95% CI: 0.16–0.51; AP = 0.15, 95% CI: 0.07–0.23; SI = 1.38, 95% CI: 1.13–1.69). Additionally, TPA exhibited a significant additive interaction with difficulty getting up (RERI = 0.12, 95% CI: 0.00-0.24; AP = 0.05, 95% CI: 0.00-0.11; SI = 1.11, 95% CI: 1.00-1.23). Sensitivity analyses Three sensitivity analyses were conducted to assess the robustness of the main findings. First, after excluding participants with missing covariate data (n = 23,421), the joint associations, interaction patterns, and dose-response relationships remained largely consistent with the primary analyses (Figure S4, Figure S5, Table S11). The additive interaction between VPA and sleep quality persisted (RERI = 0.14, 95% CI: 0.01–0.27; AP = 0.08, 95% CI: 0-0.15; SI = 1.21, 95% CI: 0.97–1.51). Second, exclusion of participants who developed depression within the first two years of follow-up (n = 4,892) to minimize potential reverse causation yielded nearly identical results (Figure S6, Figure S7, Table S12), with the VPA-sleep interaction remaining significant (RERI = 0.17, 95% CI: 0.05–0.29; AP = 0.09, 95% CI: 0.02–0.17; SI = 1.28, 95% CI: 1.01–1.62). Third, after excluding participants with baseline cardiovascular disease (n = 24,249), all main findings were preserved (Figure S8, Figure S9, Table S13), including the additive interaction for VPA and sleep quality (RERI = 0.13, 95% CI: 0-0.25; AP = 0.07, 95% CI: 0-0.14; SI = 1.18, 95% CI: 0.97–1.44). Discussion In this prospective cohort study of 323,910 UK Biobank participants, we examined the independent, joint, and interactive associations of physical activity and sleep quality with incident depression. Both factors demonstrated significant independent associations with non-linear dose-response relationships, indicating optimal protection at intermediate physical activity levels. Consistent with these patterns, participants who subsequently developed depression exhibited markedly poorer sleep quality and lower physical activity levels at baseline compared with those who remained free of depression. Participants with both low physical activity and poor sleep quality experienced nearly 4-fold elevated depression risk, demonstrating substantial cumulative effects. Significant additive synergistic effects were identified specifically between vigorous physical activity and sleep quality, while this interaction was not observed for total or moderate physical activity. Notably, physical activity’s protective associations were substantially attenuated in poor sleep contexts, whereas poor sleep’s adverse effects persisted across all physical activity levels. These findings remained robust across sensitivity analyses, highlighting the importance of addressing both factors in depression prevention strategies. Our findings align with and extend previous research on the interrelationships between physical activity, sleep, and depression. The observed non-linear dose-response relationships are consistent with recent evidence demonstrating threshold effects and optimal activity levels [8, 25]. Qiu et al. [25] identified a threshold at 2.48 MET-min/1000-week in weekend catch-up sleepers, paralleling our U-shaped and J-shaped curves indicating both insufficient and excessive activity may increase depression risk. The substantial cumulative effects we observed—with individuals exhibiting both low physical activity and poor sleep experiencing nearly 4-fold elevated risk—corroborate findings from the Canadian Longitudinal Study showing low activity patterns were associated with poorest mental health outcomes [26]. Our identification of significant additive synergistic effects between vigorous physical activity and sleep quality extends mechanistic understanding beyond previous mediation analyses. While Kaseva et al. [27] demonstrated sleep problems partially mediated the physical activity-depression association (proportion mediated = 0.36), and Wei et al. [28] showed sleep and physical activity traits mediate depression’s causal associations with age-related diseases, our study uniquely quantifies additive interactions exceeding simple mediation. Recent accelerometry-based research revealed that joint components of sleep, physical activity, and circadian rhythms explained substantial variation in depression [29]. Our stratified analyses complement this by demonstrating sleep quality’s dominant role persists across physical activity levels, suggesting hierarchical rather than purely additive relationships in depression etiology. Notably, the gradient of depression risk across physical activity levels was most pronounced among participants with high sleep quality, where low physical activity was associated with only modest risk elevation (e.g., VPA < 600 vs. 600–1200 MET-minutes/week: HR = 1.179; TPA Q1 vs. Q2: HR = 1.217); by contrast, among participants with low sleep quality, depression risk remained markedly elevated regardless of physical activity level (HR = 3.338–3.802), confirming that the inverse association between physical activity and depression was substantially attenuated in the context of poor sleep, while the adverse association of poor sleep persisted across all physical activity levels. Compared with prior work, this study advances existing knowledge in two specific respects: formally quantifying additive (not merely multiplicative) interaction, and demonstrating a directional asymmetry whereby sleep quality’s detrimental effects persist across all activity levels while physical activity’s benefits are attenuated by poor sleep. The observed associations likely operate through multiple interconnected pathways. Physical activity exerts antidepressant effects via diverse biological mechanisms, including enhanced brain-derived neurotrophic factor expression promoting hippocampal neuroplasticity, reduced inflammatory cytokines, decreased oxidative stress, and normalized HPA axis functioning [30]. Importantly, psychosocial pathways also contribute substantially, with physical activity improving depression outcomes through enhanced self-efficacy, increased social support, and elevated self-esteem [31]. These multilevel mechanisms may explain why meeting WHO physical activity guidelines confers significant mental health benefits. Conversely, sleep disturbances disrupt emotional regulation through altered sleep macrostructure, reduced slow-wave sleep, and dysfunctional autonomic activity [32]. The HPA axis represents a critical convergence point where both factors intersect—physical inactivity and sleep disturbance independently dysregulate cortisol reactivity [33], creating synergistic stress vulnerability. Circadian rhythm dysregulation emerges as particularly salient, with misalignment between circadian biology and sleep-wake behavior directly predicting depression severity [34, 35]. Our finding that sleep quality’s detrimental effects persisted across all physical activity levels suggests circadian disruption may supersede physical activity’s protective mechanisms. Our findings carry substantial clinical and public health implications for depression prevention. The hierarchical relationship between sleep quality and physical activity necessitates integrated rather than isolated interventions. Depression prevention strategies may consider emphasizing both sleep optimization and physical activity as complementary non-pharmacological strategies [36], with evidence supporting their synergistic benefits [37]. For individuals with comorbid sleep disturbances, multimodal approaches simultaneously addressing sleep and physical activity represent a plausible hypothesis for future intervention trials, though causal conclusions cannot be drawn from the present data [38]. From a public health perspective, time reallocation strategies—replacing sedentary behaviors with physical activity or extending sleep duration—offer accessible preventive interventions [39]. Population-level initiatives may consider individuals with poor sleep quality as a potentially high-risk subgroup warranting closer monitoring, given their persistently elevated depression risk regardless of physical activity level. Our dose-response findings are consistent with existing public health guidelines emphasizing moderate, sustainable activity targets, though the paradoxical risk at extreme volumes requires experimental validation given the observational design. Several methodological strengths enhance our findings’ robustness. First, the large prospective cohort with extended follow-up provides substantial statistical power while minimizing reverse causation. Second, our comprehensive assessment of multiplicative and additive interactions represents a methodological advancement, with the identification of additive synergy specifically for vigorous physical activity and sleep quality offering novel mechanistic insights. Third, the multi-dimensional sleep quality score incorporating six validated characteristics provides more nuanced evaluation than single-item measures. Fourth, systematic sensitivity analyses demonstrated remarkable consistency across different scenarios, confirming result reliability. Several limitations warrant consideration. First, self-reported physical activity and sleep measures are subject to recall bias and social desirability bias, potentially introducing non-differential misclassification that may attenuate true effect estimates. Second, the predominantly white European UK Biobank sample limits generalizability to diverse ethnic and socioeconomic populations. Third, despite extensive adjustments, residual confounding from unmeasured factors cannot be excluded; notably, anxiety disorders, perceived stress, and prior subclinical mental health symptoms—each strongly correlated with both physical inactivity and sleep disturbance, which may partially account for the observed associations. Fourth, the observational design precludes definitive causal inference—randomized controlled trials are necessary to establish causality. Fifth, we lacked temporal data on behavioral changes across follow-up, precluding examination of how activity and sleep modifications influence depression trajectories. Sixth, the construction of our composite sleep score involved assigning equal weight to six dichotomized sleep characteristics, which assumes equivalent contributions of each trait to depression risk and may result in information loss due to dichotomization. Seventh, health record-based depression ascertainment may underestimate true incidence given underdiagnosis, potentially introducing selection bias toward more severe cases. Eight, the analyses encompassed multiple exposure definitions, interaction models, and sleep trait combinations without formal adjustment for multiple comparisons due to the exploratory analyses, should therefore be interpreted cautiously and treated as hypothesis-generating pending independent replication. Finally, the significant additive interaction observed between VPA and sleep quality should be interpreted with caution, as it was not replicated across other physical activity measures. In conclusion, this prospective cohort study demonstrates that physical activity and sleep quality are independently, cumulatively, and synergistically associated with depression risk, with sleep quality’s adverse associations persisting across all activity levels. The identification of additive interaction specifically between vigorous physical activity and sleep quality advances mechanistic understanding. These findings underscore that effective depression prevention requires integrated interventions simultaneously addressing both factors rather than isolated approaches. Supplementary Information Supplementary Material 1. Acknowledgements We thank the faculties from UK Biobank Access Management Team for helping us in the data preparation (Application Number 170605). Authors' contributions Zhemin Fu and Yuwei Wang contributed to Conceptualization, Project administration, Supervision, Investigation, and Writing–original draft. Zhixing Fan contributed to Data curation, Formal analysis, and Writing–review & editing. Boxiao Gu, Xinyuan Zhao, Yifanlin Yang and Shengmei Cui contributed to Conceptualization, Investigation, and Writing–original draft. All authors had the final responsibility for the decision to submit for publication. Funding No founding. Data availability Data availability The data are available from the UK Biobank https://www.ukbiobank.ac.uk/enable-your-research/. Declarations Ethics approval and consent to participate All participants provided written informed consent, and the study received ethical approval from the North West Multi-Centre Research Ethics Committee (REC ID: 16/NW/0274). The study was carried out in accordance with the guidelines of the Declaration of Helsinki. The UKB study was approved by the North West Multicenter Research Ethical Committee. All participants provided informed written consent. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Zhemin Fu, Email: fuzhemin@syty.edu.cn. Zhixing Fan, Email: fanzhixing@ctgu.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement Data availability The data are available from the UK Biobank https://www.ukbiobank.ac.uk/enable-your-research/.