Background

The Global Burden of Disease Study in 2019 has revealed that mental disorders have consistently ranked among the top 10 leading causes of burden worldwide since 1990, with depressive disorders accounting for 37.3% of mental disorder disability-adjusted life years in 2019, affecting approximately 4.7% of the global population.1 Recurrent depression in adults with pre-existing depression presents a significant challenge throughout their lifespan, with estimates ranging as high as 75–90%.2 Recurrent depression has a profound impact on individuals’ quality of life, relationships and daily functioning. Therefore, developing prognostic interventions for adults with pre-existing depression is necessary to prevent further episodes.

Psychotherapies and pharmacotherapies are often the first-line treatments for depression. However, the existing study has demonstrated a ceiling effect on the effectiveness of psychotherapies and pharmacotherapies.3 The use of antidepressant medication often leads to iatrogenic comorbidity due to accompanying side effects such as gastrointestinal symptoms, genitourinary symptoms, sexual dysfunction and central nervous system disturbances.4 Meanwhile, the treatment coverage for depression remains low across various income levels and regions worldwide, primarily due to financial and resourcing-related barriers.5 Consequently, directing the focus towards modifiable risk factors in adults with pre-existing depression has become a priority for both public health and clinical practice.

There have been numerous studies investigating the efficacy of lifestyle-based approaches for managing depression, which typically involve applying environmental, behavioural and motivational principles to self-care and self-management of lifestyle-related health issues. Several reviews have demonstrated that exercise can serve as an accessible and cost-effective alternative or adjunctive intervention for reducing depressive symptoms.6 7 Additionally, other studies have suggested that maintaining a balanced diet, obtaining sufficient sleep and using stress management techniques such as mindfulness mediation may also be advantageous for reducing the likelihood of depression recurrence.8 9 However, the relatively small sample size and short follow-up precluded extensive investigations of how depression prognosis would be modified by lifestyle factors. Moreover, inconsistencies in defining and measuring healthy lifestyle factors across studies hinder the ability to draw definitive conclusions. Meanwhile, the focus of the most studies has predominantly been on individual lifestyle factors, neglecting to examine their aggregated effects. This may overlook the potential correlation and interaction between various lifestyle factors. It is therefore necessary to consider aggregated effect and elucidate overall patterns of healthy lifestyle.

The aim of this study was to examine the associations between aggregated and individual healthy lifestyle factors with the risk of first-time hospitalisation for the recurrence of depression and mortality among adults with pre-existing depression in the UK Biobank (UKB). Furthermore, we explored the potential effect modification by antidepressant medication use and other potential risk factors (online supplemental file 1).

MethodsStudy population and participants

This prospective cohort study used data from the UKB, a large-scale population-based study in the UK that has comprehensively collected health and lifestyle information on over 500 000 individuals aged between 37 and 73 years. Participants attended one of the 22 assessment centres located throughout England, Scotland and Wales to complete touchscreen questionnaires and physical examinations, and provide biological samples for collection between 2006 and 2010.10

Participants with pre-existing depression were identified through the integration of multiple data sources, including self-reported depression, prescribed antidepressant medication usage (codes of antidepressants used in the UKB can be seen in online supplemental table 1) and codes F32–F33 in the electronic health records (England and Wales: Health Episode Statistics; Scotland: Scottish Morbidity Records) based on the 10th Revision of the International Classification of Diseases (ICD-10). The prevalence and characteristics of probable depression within UKB have been previously described.11 Briefly, 47 043 participants with pre-existing depression were included; we then excluded the individuals with missing lifestyle factors and covariates, remaining 26 164 participants in the final analysis (online supplemental figure 1). All participants provided informed consent for linkage to national electronic health-related datasets. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline for cohort studies.

Lifestyle assessment and scoring

Healthy lifestyle factors adopted in this study conformed with the clinical guidelines developed by the World Federation of Societies for Biological Psychiatry, and included eight domains of behaviours for depression healthcare8: (1) pack years of smoking, (2) alcohol consumption, (3) diet score, (4) sleep score, (5) physical activity, (6) social health score, (7) time spent on work and (8) greenspace interaction. These lifestyle factors were collected through a touchscreen questionnaire at baseline (2006–2010), subsequent to the diagnosis of depression in participants.

The aggregated healthy lifestyle score (HLS) was calculated through the following formula, whereby a higher score indicates a healthier lifestyle: (β1×pack years of smoking+β2×alcohol consumption+β3×diet score+β4×sleep score+β5×physical activity+β6×social health score+β7×time spent on work+β8×greenspace interaction)×(8/sum of the β coefficients). Cox regression models were used to derive β coefficient for each lifestyle factor, with the exact values of β1–8 detailed in online supplemental table 2. In subsequent multivariable analyses, the HLS was divided into tertiles, with the highest tertile representing the healthiest lifestyle, and the lowest tertile representing the least healthy lifestyle. The individual lifestyle factors were used as continuous variables in the analyses. Detailed methods for assessing individual lifestyles and calculating are provided in the online supplemental eMethods.

Clinical outcome ascertainment

The primary analytical outcomes of interest were the hazards of first-time hospitalisation due to recurrence of depression and all-cause mortality. Hospital-treated depression recurrence was assessed via linkage data to the UK National Health Service (NHS) Hospital Episode Statistics (HES) database for hospital admissions according to the ICD-10 codes (F32–F33), and mortality was determined through the NHS Central Registry, from March 1995 until May 2021. Follow-up of this study was censored at the date of incident recurrence recorded in HES, death or the end of follow-up (19 December 2022), whichever occurred first.

Covariates

Covariates were selected based on a prior-defined directed acyclic graph (online supplemental figure 2). Our study finally included age, gender (female and male), ethnicity (white and non-white), education attainment (college or university degree, professional qualifications and other), Townsend Deprivation Index (TDI), body mass index (BMI, kg/m2, weight in kilograms divided by height in meters squared), C reactive protein (CRP, mg/L), overall health rating (healthy/unhealthy), self-reported long-standing illness, disability or infirmity (yes and no), self-reported other serious medical conditions or disabilities diagnosed by doctors (yes/no) and use of antidepressant medication (yes/no). TDI is a composite index of deprivation that takes into account unemployment, lack of car and cottage ownership and household overcrowding. Higher values indicate lower socioeconomic status.12 Covariates with a missing rate of less than 1% (eg, TDI and ethnicity) were excluded, covariates with missing rate greater than 1% (eg, education attainment) were either coded as an additional category for categorical variables or replaced by mean values for continuous variables (missingness of the covariates can be seen in online supplemental table 3). All covariates were evaluated during the baseline phase of the UKB study, spanning 2006–2010, subsequent to the diagnosis of depression in participants.

Statistical analysis

In descriptive analyses, continuous variables were expressed as means and SDs, categorical variables were represented by frequency and percentage. Cox proportional hazards models were performed to evaluate the associations of HLS with first-time hospitalisation due to recurrence of depression and mortality, with the calculation of HRs and 95% CIs. Linear trends were examined by entering the median value of each tertile of HLS as a continuous variable into the models. Three multivariable-adjusted models were constructed to account for potential confounding. Model 1 was adjusted for age (timescale) and sex; model 2 was adjusted as in model 1 and for TDI, ethnicity, education attainment and BMI; model 3 was adjusted as in model 2 and for health rating, long-standing illness, disability or infirmity, other serious medical conditions diagnosed by doctor, CRP and use of antidepressants. The proportional hazard assumption was assessed for all Cox models using Schoenfeld residuals, and no evidence of violation was detected. Dose-response associations between HLS, individual lifestyle factors and the risk of first-time hospitalisation for the recurrence of depression and mortality were evaluated using restricted cubic splines (RCS) fitted in the Cox models.

Stratified analyses were conducted according to use of antidepressant medication (yes/no), age (<60 and ≥60 years), sex (female and male), ethnicity (white and non-white), education attainment (college or university degree, professional qualifications and others), TDI (<median and ≥median), BMI (<25 and ≥25 kg/m2), CRP (≤10 and >10 mg/L), self-reported long-standing illness, disability or infirmity (yes/no), self-reported other serious medical conditions or disabilities diagnosed by doctors (yes/no). Interactions were tested by a likelihood ratio test comparing models with and without product terms between HLS and stratified factors.

Sensitivity analyses were also conducted to assess the robustness of our results. First, we excluded participants with poor self-rated health status (categorised as healthy and unhealthy according to questionnaire), considering that they were less likely to adopt healthy lifestyles. Second, we excluded participants with limited follow-up (≤2 years) to avoid the potential risk of reverse causation. Third, complete case analyses were used to assess the associations between HLS with first-time hospitalisation due to recurrence of depression and mortality.

All UKB analyses were conducted using Stata V.16.0 (StataCorp, College Station, Texas) and R software (V.4.1.3). The statistical significance was set as p<0.05 (two-sided test).

ResultsPopulation characteristics

Table 1 shows the baseline characteristics of the study participants according to tertiles of HLS. Among 26 164 participants with a mean age of 56.0 years (SD, 7.9 years), 17 259 (65.96%) were female. The majority self-identified as white accounting for 24 088 participants (92.07%), and 7584 participants had completed at least some college education (28.99%). Compared with those within the lowest tertile of HLS, participants who were categorised into the highest tertile were generally younger, had lower TDI values, higher levels of education attainment, lower BMI and CRP levels, self-reported healthier status, less likely to have a prevalence of long-standing illness, disability or infirmity, serious medical conditions or use of antidepressant medications.

Table 1

Baseline characteristics of adults with pre-existing depression according to aggregated healthy lifestyle scores (HLS)

Aggregated and individual healthy lifestyle and first-time hospitalisation due to recurrence of depression

Over a mean follow-up period of 13.3 years, 9740 cases of first-time hospitalisation due to recurrence of depression were documented. The risk of depression recurrence decreased with increasing HLS (HR=0.91, 95% CI 0.90 to 0.92). Compared with the lowest tertile of HLS, the adjusted HRs for depression recurrence were 0.86 (95% CI 0.82 to 0.91) and 0.73 (95% CI 0.69 to 0.77) across the medium and highest tertiles of HLS, respectively (P for trend <0.001, table 2). Regardless of the level of HLS, there was a reverse linear association between HLS and depression recurrence (P for non-linearity=0.329, figure 1).

Table 2

Multivariable associations of aggregated healthy lifestyle scores (HLS) with first-time hospitalisation for the recurrence of depression and mortality among adults with pre-existing depression

Figure 1

Dose-response associations between HLS and individual lifestyle factors with the risk of first-time hospitalization for the recurrence of depression among adults with pre-existing depresssion. A: Healthy lifestyle score. B: Pack years of smoking. C: Alcohol consumption. D: Diet score. E: Sleep score. F: Physical activity. G: Social health score. H: Time spent on work. I: Greenspace interaction. Models were adjusted for age, sex, ethnicity, education attainment, Townsend Deprivation Index, body mass index, self-assessed overall health rating, self-reported long-standing illness, self-reported serious medication conditions or disabilities, C reactive protein levels and use of antidepressant medication.

The associations of individual lifestyle factors with depression recurrence were also evaluated separately (figure 2). After adjusting for potential confounders, the HR for depression recurrence in participants with the highest smoking intensity compared with the lowest category was 1.14 (95% CI 1.09 to 1.19). Notably, the highest level of physical activity adherence was associated with a 5% higher risk of depression recurrence in this study. In contrast, we observed a decreased risk of depression recurrence among those who consumed higher amounts of alcohol (HR=0.84, 95% CI 0.79 to 0.88), adhered to healthier diet (HR=0.88, 95% CI 0.82 to 0.94) and sleep pattern (HR=0.86, 95% CI 0.82 to 0.91), spent more time employed in current job (HR=0.81, 95% CI 0.76 to 0.85) or engaged more frequently with greenspace (HR=0.87, 95% CI 0.83 to 0.92) when compared with those who exhibited the lowest levels of these individual lifestyle factors. Results from the multivariable-adjusted RCS regression showed that the risk of depression recurrence decreased linearly with increasing diet and sleep score (P for non-linearity=0.179 and 0.508, respectively), whereas the associations between pack years of smoking, alcohol consumption, time spent on work, greenspace interaction and depression recurrence were more non-linear (all P for non-linearity <0.05). In addition, increased physical activity was linearly associated with increased risk of depression recurrence (P for non-linearity=0.877) (figure 1).

Figure 2

Multivariable-adjusted associations of individual lifestyle factor with first-time hospitalization for the recurrence of depression and mortality among adults with pre-existing depression. Models were adjusted for age, sex, ethnicity, education attainment, Townsend Deprivation Index, body mass index, self-assessed overall health rating, self-reported long-standing illness, self-reported serious medication conditions or disabilities, C reactive protein levels and use of antidepressant medication.

Aggregated and individual healthy lifestyle and mortality

We documented 1527 deaths over a mean follow-up period of 13.3 years. In the fully adjusted model, higher HLS were significantly associated with a lower risk of mortality; multivariable HR of the highest tertile compared with lowest tertile for mortality was 0.92 (95% CI 0.87 to 0.95) (table 2). The RCS model indicates a linear association between HLS and mortality (P for non-linearity=0.442) (figure 3).

Figure 3

Dose-response association between HLS and individual lifestyle factors with mortality among adults with pre-existing depression. A: healthy lifestyle score. B: pack years of smoking. C: alcohol consumption. D: diet score. E: sleep score. F: physical activity. G: social health score. H: time spent on work. I: green space interaction. Models were adjusted for age, sex, ethnicity, education attainment, Townsend Deprivation Index, body mass index, self-assessed overall health rating, self-reported long-standing illness, self-reported serious medication conditions or disabilities, C reactive protein levels and use of antidepressant medication.

The associations between individual lifestyle factors and mortality were shown in figure 2. Similarly, participants with the highest smoking intensity experienced a 55% increase in mortality compared with those with the lowest smoking intensity (HR=1.55, 95% CI 1.38 to 1.74). Conversely, a decreased risk of mortality was observed among those who adhered to a healthy diet (HR=0.78, 95% CI 0.66 to 0.93), engaged in higher levels of physical activity (HR=0.81, 95% CI 0.72 to 0.93), achieved higher social health scores (HR=0.79, 95% CI 0.63 to 0.98) and spent more time in their current job (HR=0.74, 95% CI 0.64 to 0.87). Results from the multivariable-adjusted RCS regression indicated a linear decrease in mortality risk with increasing diet score (P for non-linearity=0.262). Conversely, the relationship between physical activity and time spent on work exhibited greater non-linearity (all P for non-linearity <0.05). Furthermore, an increase in pack years of smoking was associated with a non-linear increase in mortality risk (figure 3).

Effect modification and sensitivity analyses

When considering the heterogeneity of whether adults with pre-existing depression take medication, we found a significant interaction between the use of antidepressant medication and HLS (P for interaction <0.001, as shown in table 3). The inverse association with first-time hospitalisation due to recurrence of depression was more pronounced among participants who reported using antidepressant medication than those who did not take medication. Conversely, the association with mortality was more pronounced among participants who reported no use of antidepressant medication than those who took.

Table 3

Associations of aggregated healthy lifestyle scores (HLS) with first-time hospitalisation for the recurrence of depression and mortality among patients with pre-existing depression stratified by antidepressant medication use

When stratified by other potential risk factors, no statistically significant interactions were observed between HLS and all stratified factors except for BMI (online supplemental tables 4 and 5). The association between HLS and first-time hospitalisation due to recurrence of depression was stronger among participants with lower BMI than higher BMI (P for interaction=0.001). Similar results were noted regarding the association between HLS and mortality (P for interaction=0.004).

The findings remained largely consistent with the main analyses when further excluding participants with poor self-rated health status (online supplemental table 6), those with limited follow-up of ≤2 years (online supplemental table 7) or conducting complete analyses by dropping all missing values of covariates (online supplemental table 8).

Discussion

In this large prospective cohort study of 26 164 adults with pre-existing depression from UKB, we found that adhering to a healthy lifestyle, particularly for less smoking, healthy diet or more time employed in current job, was associated with a lower risk of first-time hospitalisation for the recurrence of depression and mortality (online supplemental file 1). Notably, the associations remained regardless of the use of antidepressant medication. These findings highlight the benefits of adopting a healthy lifestyle in reducing the risks for depression recurrence and mortality among adults with depression.

Despite a growing body of evidence showing associations between combined and individual lifestyle factors and the risk of new-onset depression, to our knowledge, this is the first study to assess whether aggregated modifiable risk factors are associated with prognosis outcomes in adults with pre-existing depression. Our findings suggest that higher scores on measures of healthier lifestyle were associated with reduced first-time hospitalisation due to recurrence of depression and improved survival rates among adults diagnosed with depression. Moreover, we observed that the inverse association between HLS and first-time hospitalisation due to recurrence of depression was more pronounced among participants who reported using antidepressant medication than those who did not take medication. This could be attributed to the effectiveness of antidepressants in relieving depressive symptoms and the implementation of lifestyle-based interventions.13 However, the mortality risk was higher among participants who did not report using antidepressant medication compared with those who did. Several observational studies have substantiated the potential detrimental effects of these medications on various outcomes such as falls, fracture, epilepsy, hyponatraemia, attempted suicide, self-harm, stroke and transient ischaemic attack.14 15 Consequently, in this study, the association between HLS and mortality risk was found to be stronger among the non-medicated population.

Moreover, the aggregated HLS in this study encompass eight domains, providing a more comprehensive insight into the impacts of lifestyle and environmental factors on individuals with pre-existing depression. The role of lifestyle-based approaches in depression managing may involve multiple biological and psychological mechanisms. Randomised controlled trials have demonstrated that diet interventions can augment the concentration of antioxidant substances in plasma and mitigate oxidative stress. Furthermore, it can significantly elevate serum levels of serotonin, a crucial neurotransmitter strongly associated with alleviating symptoms of depression.16 Extensive reviews have explained on the potential impact of smoking, sleep, physical activity and psychosocial stress on inflammation and oxidative and nitrosative stress that underlie depression progression.17 18 Interventions targeting lifestyle and environmental domains can also extend into the management of substance use disorders through an increase in positive affect, acute reduction in cravings and urges and improvement in comorbid physical disease. These mechanisms are pertinent as substance use can be a means of ‘self-medicating’ psychological distress, past and ongoing trauma and mental illness.19

When examining the role of individual lifestyles, participants who smoked less, adhered to a healthier diet and sleep pattern, spent more time employed in their current jobs and had greater exposure to greenspace were associated with significantly lower risk of depression recurrence. Similarly, adopting healthy lifestyles such as less smoking, healthy diet, physically active and more time employed in current job was associated with a lower risk of mortality among adults with depression. Interestingly, moderate alcohol consumption was significantly associated with the lowest risk of depression recurrence (HR=0.77, 95% CI 0.73 to 0.81). This finding is not entirely unexpected given the conflicting evidence and competing hypotheses surrounding the relationship between alcohol consumption and depression, which may not follow a linear pattern.20 21 Meanwhile, we have found that adherence to the highest level of physical activity was marginally linked to an increased risk of depression recurrence (HR=1.05, 95% CI 1.00 to 1.11). The findings of this study may diverge from those of other studies that have reported more substantial enhancements in physical activity for depressive symptoms.22 Some potential biological mechanisms could explain these findings. High-intensity physical activity might prompt an increased inflammatory response, such as higher levels of interleukin-6 and tumour necrosis factor-alpha, which are closely associated with depression. In addition, high volume of physical activity may cause an increase in the release of stress hormones such as cortisol, which could lead to reduced sleep quality, increased anxiety and the onset of depressive symptoms.23–25 Furthermore, the efficacy of physical activity as a treatment for depression may be contingent on factors such as sample size, research design and methodology.

Evidence from previous reviews has indicated that prolonged exposure to long working hours may lead to depression.26 27 This could be attributed to the limited decision-making autonomy, job-related stress and harassment experienced by both male and female employees, leading to a gradual increase in depressive symptoms over time. However, our findings suggest that longer job engagement in the current position can serve as a protective factor against depression recurrence and mortality. It may serve as compelling evidence for the implementation of a crucial preventive strategy in managing depression within clinical settings. We have also discovered that adults diagnosed with depression could benefit from the long-term exposure to residential greenspace, which is consistent with the results of the previous published cross-sectional study.16 The findings of this study serve as a crucial reference for promoting mental health and addressing depression, particularly in the face of rapid urbanisation and dwindling greenspace resources. Therefore, strengthening the protection and development of urban greenspaces is highly significant for enhancing public mental health. We did not observe a significant association between social health and the remission of the depression, despite the social health in our study was defined according to frequency of friend and family visits, engagement in leisure and social activities and ability to confide.

We also found BMI-dependent effects of HLS on the risks of first-time hospitalisation for the recurrence of depression requiring hospitalisation and mortality, with the HRs being relatively lower at milder BMI (<30 kg/m2) and the effect sizes decreasing with increasing BMI. The relationship between depression and obesity is confirmed to be bidirectional, the presence of one increases the risk for developing the other. The association between depression and obesity can be elucidated by the shared biological pathways, which encompass genetics, alterations in systems involved in homeostatic adjustments (such as the hypothalamic-pituitary-adrenal axis, immunoinflammatory activation, neuroendocrine regulators of energy metabolism including leptin and insulin, and microbiome), as well as brain circuitries integrating homeostatic and mood regulatory responses.28 29 Further investigations are warranted to validate these associations.

Limitations

The main strengths of this study lie in the development of an aggregated HLS and the utilisation of a large-scale population with pre-existing depression, which have enabled us to investigate the relationship between healthy lifestyles and both first-time hospitalisation due to recurrence of depression and mortality. However, there are several limitations still needed to be considered. First, the information on healthy lifestyle factors was primarily obtained from the self-reported questionnaires, which may be subject to recall bias. Additionally, we cannot rule out the potential influence of lifestyle changes over time, as exposure data were only collected at baseline and no time-varying variables were assessed during follow-up. Second, the primary outcome is first-time hospitalisation due to recurrence of depression and mortality. Therefore, we are unable to account for any comorbidities that may have arisen after depression diagnosis or treatment, such as anxiety disorders, sleep disturbances and physical illnesses. Third, it is difficult to make causality conclusions in an observational cohort study. Despite controlling for numerous relevant confounding factors, unmeasured variables could introduce bias into our findings. Nevertheless, through a series of sensitivity analyses, we have substantiated the robustness of our findings.