Participants were enrolled from May 2023 to March 2024 at the Struttura Complessa Psichiatria Universitaria, Dipartimento di Neuroscienze e Salute Mentale, Azienda Ospedaliero-Universitaria “Città della Salute e della Scienza di Torino”, Turin, Italy. Clinical assessment and venous blood sampling of all subjects were performed at this location.

The patients recruited were outpatients aged between 18 and 65 years, living in an urban area, and having a diagnosis of MDD or BD (according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR) criteria) with an ongoing MDE necessitating the initiation or modification of antidepressant treatment. MDD, BD, and MDE were diagnosed according to the DSM-5-TR criteria and assessed with the Structured Clinical Interview for DSM-5-TR Research Version (SCID-RV). Exclusion criteria were neurodevelopmental disorders, alcohol and/or substance use disorders, cognitive impairment, intellectual disability, and psychiatric comorbidities (e.g., obsessive-compulsive disorders and anxiety disorders), difficulties with walking or mobility that impair access to green areas, and pregnancy.

Due to the lack of clear reference values for the biological markers analyzed, we recruited a sample of healthy volunteers (HV). HV were defined as people not affected by any mental disorder and with a negative family history for MDD and/or BD up to second-degree relatives. To minimize potential confounding factors in the comparison between patients and HV, we selected a sample of HV comparable to the clinical sample in terms of age, sex, education, and community of origin, specifically urban areas. Diagnoses of mental disorders were excluded using the same standardized clinical tools (SCID-5-RV).

The study had a prospective design: patients were assessed by a psychiatrist, and a venous blood sample was taken at baseline (t0), i.e., during the MDE at the time of antidepressant therapy modification, and after six weeks at the follow-up visit (t1).

At baseline, patients were instructed to spend time in natural green environments. We defined green environments as all those areas with low anthropization that allow patients to be in contact with nature while distancing themselves as much as possible from the urban environment. Therefore, we considered green environments to be forests, woodlands, and extensive urban gardens or parks, allowing patients an immersive greenery experience through sight, hearing, and smell. This operational definition aligns with the Japanese concept of Shinrin-Yoku (62, 63, 65). We suggested patients be exposed to green environments for 45–60 minutes, three non-consecutive days per week, until the follow-up visit. Exposure was explained as simply spending time in nature without engaging in physical activity. Since no established scientific protocols recommended a specific duration of green exposure, we based our approach on the WHO Guidelines on Physical Activity and Sedentary Behavior, which advise at least 45 minutes of physical activity three times a week on non-consecutive days (86). Following this model, we similarly suggested a comparable frequency and duration for green exposure. HV’s data collection and blood sampling were only performed at one time point, since the only information regarding green exposure that was collected for this group concerned their usual habits, namely whether or not they had frequented green environments during the six weeks prior to recruitment.

Comprehensive information was provided on the procedures and goals of the study, and signed consent was obtained from all subjects. The study was designed in accordance with the Declaration of Helsinki and was approved by the Local Research Ethics Committee (Protocol number: 0010765).

At the baseline, all participants were evaluated using a semi-structured interview to assess socio-demographic characteristics and to investigate the relationship with green in terms of time spent in green environments, including parks, gardens, and woods, in the six weeks before the interview. HV were interviewed to assess whether they met the exclusion criteria and were considered in the exposed to greenery group if they stayed in green environments for at least 45 minutes on two non-consecutive days per week for the six weeks before enrollment and blood sampling, otherwise they were considered in the non-exposed to greenery group.

The clinical characteristics of MDD or BD and the severity of depressive symptoms were assessed by two expert psychiatrists (C.B.; A.C.) The severity of symptoms was evaluated with the Hamilton Rating Scale for Depression (HAM-D) (87), where a higher score indicates greater severity.

After six weeks from baseline (t0), patients performed the follow-up visit (t1). They were reassessed with the HAM-D scale and asked whether, during the six weeks of the study, they were exposed to green environments for at least 45 minutes on two non-consecutive days per week. According to their response at the six-week follow-up visit (t1), patients were classified as exposed to or nonexposed to green environments during the study, i.e., between t0 and t1. This classification was applied to divide patients into groups both at follow-up (t1) and retrospectively at baseline (t0).

The nursing staff conducted venous sampling of all subjects at baseline and follow-up. In the serum, we measured the concentration of cortisol, CRP, IL-6, C3 and C4 complement fractions, leptin, and adiponectin at the Struttura Complessa Biochimica Clinica, Azienda Ospedaliero-Universitaria “Città della Salute e della Scienza”, Turin, Italy, and BDNF at “Neuroscience Institute Cavalieri Ottolenghi” laboratory in Orbassano.

The procedure for measuring biomarker concentrations in participants’ blood samples is described in the Supplementary Material.

The normal distribution of the continuous variables was verified with the Shapiro-Wilk test. Since the sample size was relatively small, to provide a more robust and appropriate summary of the data, variables were expressed as median and interquartile range.

The Mann-Whitney test was employed for independent sample comparisons, and the Wilcoxon test for paired samples. The effect size of the comparisons was calculated using the rank biserial correlation method. The results were reported as absolute values. The false discovery rate (FDR) was controlled with the Benjamini-Hochberg (BH) correction procedure and set at.05 (88).

The significant results of the comparisons between exposed and nonexposed patients at follow-up (t1) and within exposed or nonexposed patients between baseline (t0) and follow-up (t1) were used to perform linear backward regression models. Follow-up (t1) variables showing a significant variation from the baseline in the between- and/or within-group comparisons were chosen as outcomes for the regression models. Exposure to green environments between baseline and follow-up was entered as the regressor into all regression models. The models were controlled for baseline (t0) diagnosis (MDD or BD), depressive symptoms, drug naivety, age, sex, education, body mass index (BMI), self-reported physical activity, and relationship with green, and for the baseline (t0) variable corresponding to the dependent follow-up (t1) variable. Regarding this latter control, for example, if IL-6 at follow-up (t1) was the outcome of the regression model, we controlled for IL-6 at baseline (t1) as an additional control variable. The regressor and control variables entered the regression model as independent variables.

We set the alpha error at .05, the beta error at .10, and the effect size for non-parametric comparisons at .5. All comparisons were one-tailed, assuming a reduction in depressive symptoms and inflammatory biomarkers concentrations in participants who reported engaging in green exposure. Independent comparison had a between-groups ratio of 1.5. For multiple linear regressions, we set the alpha error at .05, the beta error at .10, three significant control predictors, and an effect size f² of .15 related to the reported participation in the intervention of green exposure. This led to a minimal sample size of 38 patients for paired comparisons, 58 and 88 for independent ones, and 73 for multiple linear regressions.

Analyses were conducted with the IBM Statistical Package for Social Science (SPSS), version 29.0; Jasp, version 0.18.3; and Gpower, version 3.1.

Ninety-three patients during a MDE requiring a modification of the antidepressant were screened. Twenty-three (24.7%) were excluded for age or mental and/or physical comorbidities, constituting an exclusion criterion. Of the seventy eligible patients, seventeen (24.3%) refused to participate in the study because they felt it was burdensome or did not want to take additional blood samples. Fifty-three patients were recruited at baseline; twenty-two participants dropped out at follow-up, and 31 (58.5%) completed the study, leading to a drop-out rate of 41.5%. Nineteen completers (61%) engaged with the intervention, reporting regular exposure to greenery during the six weeks of the study observation. Consequently, the attrition rate for the suggested intervention was 39%. These results are represented in Figure 1.

All inferential analyses were underpowered. In particular, the post-hoc statistical power of the pairwise comparison was 65% for exposed patients and 47% for nonexposed. For the between-group comparison between exposed and nonexposed patients, it was 36%, and for the linear regression, it was 55%. Due to the low statistical power, the results of the following inferential analyses will be considered as trends since the low sample size of this pilot study may not bring out significant results or, conversely, despite FDR correction, make differences evident due to small sample size “instability”.

Eighty-four subjects (53 patients, 31 HV) participated in the study. No trends suggest differences at baseline between completers and drop-outs (Table 1). Most completer patients were female (71%), had a median age of 47, and had an education of 13 years. The median duration of illness was six years, and the median severity of the MDE was moderate (89).

Sex-dependent distributions of completers and HVs’ variables are reported in Supplementary Material (Supplementary Figures 1, 2, Tables 1, 2). No trends were found after FDR correction.

The number of patients with MDD or BD diagnosis remained stable throughout the study, with no patients experiencing a switch to hypomania or mania.

At baseline, compared to nonexposed patients, the 19 participants who were exposed to greenery during the 6-week study follow-up showed the following trends: lower CRP, IL-6, C3, C4, and leptin levels and higher adiponectin concentrations (Figure 2A, Table 2).

At follow-up, we found the same trends except for C4 (Figure 2B, Table 2).

In the paired sample comparison of patients exposed to greenery, we found the following trends at the six-week follow-up visit: a reduction of depressive symptoms and IL-6 levels and an increase in adiponectin serum concentration (Figure 2C, Table 2).

No trends in biomarkers and depressive symptoms were found in nonexposed patients (Table 2).

The 10 HV exposed to green in the six weeks before blood sampling (32%) had lower leptin levels than the 21 nonexposed (68%) (Figure 3A, Table 3). The comparison between patients and HV exposed to green did not show trends either at baseline or at follow-up (Table 3).

At baseline, nonexposed patients compared to nonexposed HV showed the following trends: higher levels of IL-6, C3, and leptin and lower levels of adiponectin (Figure 3B, Table 3).

At follow-up, we found a trend of lower adiponectin serum concentrations in nonexposed patients compared to nonexposed HV (Figure 3C, Table 3).

The regression models (Table 4) showed the following trends of associations: patients’ adherence to the intervention of regular exposure to green environments during the six weeks of the study and a reduction of depressive symptoms, CRP and IL-6 plasmatic levels, and an increase in adiponectin plasma concentrations. No trends related to sex differences were observed (Supplementary Material, Tables 1, 2, Figures 1, 2).

This study has some limitations that must be acknowledged. The sample size is highly limited, leading to low statistical power and consequent non-generalizability of the results and the impossibility of establishing sound associations, but only possible trends to be verified by future studies. The observational design of this study lacks the control capacity of a randomized trial, which may introduce biases, particularly the selection bias related to attrition. Five other sources of potential selection bias are discussed above and represent a further limitation for the generalizability of the feasibility of the intervention of active exposure to greenery proposed in this study. We focused exclusively on exposure to large green spaces, but due to the small sample size, we could not account for differences between environments, such as forests or large urban parks. Patients self-selected into green exposure, potentially leading to the exclusion of less motivated individuals. We did not use a checklist to ensure that the intervention of exposure to green environments was adequately explained to patients. Green exposure was not objectively verified, but self-reported because, for ethical and privacy reasons, using GPS tracking was impossible. This introduced uncertainty. We omitted a dose-response analysis since, at the six-week follow-up, we only had a dichotomous variable indicating whether patients frequented green spaces at least twice a week during the six weeks of follow-up. Patients were treated with different antidepressants or mood stabilizers, and we could not account for all possible variations in medication type and dosage; rather, we could only account for whether patients had a medication adjustment or a new prescription. Again, regarding possible confounding variables, the study lacks an assessment of economic status factors related to lifestyle, e.g., daily hours of sleep. Concerning regression models, we could not control for changes in psychopharmacological therapy nor socioeconomic status. However, it is worth noting that, since antidepressant therapy was set or modified at baseline, psychiatrists generally avoided major changes during the following six weeks to properly assess treatment efficacy, considering the delayed onset of antidepressant effects. This waiting period coincided with the follow-up visit of this study, likely reducing the confounding effect of therapy adjustments. Regarding the lack of assessment of socioeconomic status, we further acknowledge that socioeconomic disadvantage may have negatively impacted patients’ clinical outcomes, acting as a putative confounding variable in the analysis. Finally, concerning physical activity, we controlled for self-reported physical activity in the regression models; however, we could not exclude the possibility of inaccuracies in patients’ reporting. Therefore, part of the observed trends might be due to increased exercise rather than exposure to nature alone. The short follow-up period hinders the ability to assess long-term effects on symptom remission, functioning, and quality of life and account for the influence of seasonal variations on both mood and biomarker responses. Lastly, biomarkers in HV were measured at a single time point and, unlike patients for whom we assessed exposure to greenery during the study, this variable was assessed retrospectively, introducing possible biases.

Despite these limitations, this study has notable strengths. Unlike previous studies, which typically involve non-clinical populations, do not account for the specific phase of the illness, or focus on living in proximity to green spaces rather than exposure into them (56), this study is pioneering as the first to examine green space exposure in a clinical population (MDD and BD patients), during a specific phase of these disorders (MDE), in a real-world setting, showing that the majority of patients adhere to spending time in green spaces despite acute depression. We also show preliminary trends of improvements in depressive symptoms and inflammatory balance linked to self-reported adherence to an intervention of green space exposure. Additionally, the study’s holistic approach considers not only depressive symptoms but also factors related to inflammation, stress, metabolism, and neurotrophic factors and focuses on the human-nature interaction in terms of active exposure into green environments.

Following the preliminary results of this study, future research should aim first at performing a randomized controlled trial on this topic, to allow stronger causal inferences and reduce selection biases. Additionally, increasing the sample size to reach higher statistical power and extending the follow-up period to eight months would help assess the long-term impacts of green exposure on patients’ functioning and quality of life, strengthening the sustainability of the benefits resulting from green exposure for a prolonged period. Furthermore, with a longer follow-up, it will also be possible to better highlight the long-term impact of green exposure, to stratify patients based on sex, on the type of green space they were exposed to, distinguishing, for example, between a dense forest or an urban park, and it will also be possible to collect more detailed data concerning frequency and duration of green exposure. Moreover, to provide more objective measures of green exposure and its immediate effects on patients' well-being, thereby reducing potential biases, it would be valuable to organize rehabilitation groups, led by a mental health expert, for example, a psychiatric rehabilitation technician, or by an expert in Shinrin-Yoku practice. By doing so, we could also consider using a checklist to assess the intervention’s explanation better and perform a dose-response analysis. Moreover, evaluating neurocognitive functions and their relationships with MDE, inflammation, and exposure to green spaces could offer insights into the underlying mechanisms. Studying epigenetic processes, such as miRNA transcription and histone modifications, could help clarify the role of green space exposure in regulating gene expression associated with depression and inflammation, further bridging the gap between environmental exposure and biological mechanisms. Finally, this protocol could be extended to anxiety and chronic stress disorders to explore the potential benefits of green exposure in these conditions as well.

In conclusion, the intervention of exposure to green environments was feasible for patients during a MDE, as more than half of the participants who attended the follow-up incorporated green exposure into their routine. This suggests that a larger study with a more rigorous evaluation might be possible in the future. Furthermore, even if underpowered due to the low number of participants completing the study and engaging in green exposure, the trends resulting from this pilot study suggest that self-reported exposure to green environments was associated with reductions in depressive symptoms, IL-6 and CRP levels, and an increase in adiponectin after six weeks. Future studies with larger sample sizes should verify the potential therapeutic benefits of green space exposure on MDEs. Information from this work is relevant to psychiatrists treating depression and epidemiologists or public mental health experts who would like to test the hypotheses of this pilot study. If the results are confirmed in larger studies, they can inform policymakers and urban green planners and consequently encourage the design and distribution of large parks throughout the cities to enhance urban mental health.