This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Cardio, is properly cited. The complete bibliographic information, a link to the original publication on https://cardio.jmir.org, as well as this copyright and license information must be included.
The health benefits of urban green space have been widely reported in the literature; however, the biological mechanisms remain unexplored, and a causal relationship cannot be established between green space exposure and cardiorespiratory health.
Our aim was to conduct a panel study using personal tracking devices to continuously collect individual exposure data from healthy Chinese adults aged 50 to 64 years living in Hong Kong.
A panel of cardiorespiratory biomarkers was tested each week for a period of 5 consecutive weeks. Data on weekly exposure to green space, air pollution, and the physical activities of individual participants were collected by personal tracking devices. The effects of green space exposure measured by the normalized difference vegetation index (NDVI) at buffer zones of 100, 250, and 500 meters on a panel of cardiorespiratory biomarkers were estimated by a generalized linear mixed-effects model, with adjustment for confounding variables of sociodemographic characteristics, exposure to air pollutants and noise, exercise, and nutrient intake.
A total of 39 participants (mean age 56.4 years, range 50-63 years) were recruited and followed up for 5 consecutive weeks. After adjustment for sex, income, occupation, physical activities, dietary intake, noise, and air pollution, significant negative associations with the NDVI for the 250-meter buffer zone were found in total cholesterol (–21.6% per IQR increase in NDVI, 95% CI –32.7% to –10.6%), low-density lipoprotein (–14.9%, 95% CI –23.4% to –6.4%), glucose (–11.2%, 95% CI –21.9% to –0.5%), and high-sensitivity C-reactive protein (–41.3%, 95% CI –81.7% to –0.9%). Similar effect estimates were found for the 100-meter and 250-meter buffer zones. After adjustment for multiple testing, the effect estimates of glucose and high-sensitivity C-reactive protein were no longer significant.
The health benefits of green space can be found in some metabolic and inflammatory biomarkers. Further studies are warranted to establish the causal relationship between green space and cardiorespiratory health.
Previous studies have demonstrated the health benefits of the natural environment and urban green space on mental health [
With the aim to explore the effects of green space on respiratory and cardiovascular biomarkers to provide evidence for the underlying biological pathways, we conducted a panel study using personal tracking devices to continuously collect the individual exposure data in healthy Chinese adults aged 50 to 64 years living in Hong Kong. Using these data, we estimated the independent effects of green space exposure on different metabolic, inflammatory, and oxidative biomarkers for cardiovascular and respiratory health.
The target population was Chinese adults aged 50 to 64 years who had been living in Hong Kong for the past 2 years. This age group was chosen because of their high risk of preclinical chronic conditions [
Using the formula for mixed models with repeated measurements [
The participants were recruited and followed up for 5 consecutive weeks during October to December 2017. On the same weekday of each week during the study period, each individual participant was invited to visit the Integrative Health Clinic (IHC) of the Hong Kong Polytechnic University for blood sample collection and physical examination. Each participant had a total of 6 visits during the whole study period (1 at enrollment and 5 at follow-up). These visits were scheduled on early mornings (8 AM-10 AM) of the same weekday to reduce the bias caused by the diurnal change of biomarkers. If participants took any medication, experienced allergies in the preceding 7 days, or worked a night shift on the day before their scheduled visit, these visits would be postponed for 1 week.
A summary of the data collection procedure at the clinic visits is shown in Table S1 in
At the first visit, a GPS device (BT-Q1000XT, Qstarz International Co Ltd) was distributed to each participant. Research assistants provided the participants with both verbal and paper instructions on device use. The GPS device logged the participant’s location coordinates, date, and time at 1-minute epochs, and the real-time data were automatically collected by a server on the university campus. Raw GPS data were first screened in ArcGIS software (Esri) for missing or suspicious data by comparing them with the daily activity log. Days with less than 600 minutes of GPS data recorded were regarded as missing and excluded from subsequent analysis. The research assistants regularly checked the GPS data collected on the server to ensure the completeness of these data. The participants showed good compliance and yielded no missing data.
We matched cleaned GPS data to a map of the normalized difference vegetation index (NDVI) in the entire territory of Hong Kong to calculate the urban greenness within a 100-, 250-, and 500-meter–radius buffer zone of individual GPS coordinates, based on a Satellite Pour l’Observation de la Terre (SPOT) 6 image obtained in 2016. The NDVI is a normalized ratio of infrared and red bands ranging between –1 and 1, with higher numbers indicating more green vegetation [
We also estimated individual exposure to two ambient air pollutants, nitrogen dioxide (NO2) and particulate matter with a diameter less than 2.5 μm (PM2.5), by matching the raw GPS coordinates with the hourly maps of PM2.5 and NO2 at 10-meter spatial resolution. These hourly maps were remodeled from temporal information reported by the Kwai Chung Monitoring Station and adjusted by the spatial patterns of air pollutants reported in previous local studies [
The annual average household exposure to road traffic noise was estimated by matching the residential addresses with a 3D-built environment database in Hong Kong that was validated using real-time field data in a local study [
At the first visit, an accelerometer data logger (ActiGraph GT3X, ActiGraph LLC) was also distributed to each participant, and the research assistants reminded them to wear the accelerometer on the wrist of their nondominant hand all the time during the study period to continuously record their physical activities. Raw accelerometer data were collected at 1-minute epochs. Individual daily physical activities were calculated from the vector magnitudes of the cleaned data. The cutoff points of light, moderate, and vigorous physical activity were determined according to a formula validated in a Chinese population [
We asked each participant to record their daily dietary intakes in weeks 2 and 5 using a standard dietary journal widely used in nutritional studies [
The above collected data were converted into weekly averages by taking the arithmetic mean—with the exception of the weekly average access to green space, which was calculated using the geometric mean of the NDVI—in the corresponding period between two visits of each participant.
At each visit, participants first measured their body weight and height, systolic blood pressure, diastolic blood pressure, and heart rate (HR). Lung function tests were conducted using a spirometer (microQuark, COSMED), following the guidelines of the American Thoracic Society [
We chose a panel of biomarkers used in previous environmental studies on the cardiovascular health effects of green space and air pollution [
Spearman coefficients were calculated to assess the correlations between variables. A generalized linear mixed-effects model (GLME), which includes two components, namely, a fixed effect (green space effects and confounding) and a random effect (within-subject variations), was used to estimate the association between green space exposure and biomarker levels or lung functions. The potential confounding factors, including sex, total household income, occupation, physical activities, and household traffic noise exposure, were included as fixed effects in the model. Because there were categorical confounding factors, the generalized version of the variance inflation factor (VIF) was used to detect multicollinearity [
We fit 4 different models to the weekly measurements of each biomarker to estimate the independent effects of green space. In model 1, the NDVI was added as the only explanatory factor; in model 2, variables of sex, physical activity (moderate to vigorous physical activity levels), occupation, household income, and traffic noise exposure were added to model 1 as confounding factors; in models 3 and 4, PM2.5 and NO2 were respectively added to model 2 to assess the independent effects of green space exposure with adjustment for ambient air pollutants; and in model 5, daily consumption amounts of protein and carbohydrates were added to model 2 to adjust for the confounding effects of dietary intake. The effects of green space were quantified by the percentage changes of biomarker concentrations associated with a per interquartile range increase of the NDVI. The goodness of fit was evaluated by the Akaike information criterion (AIC), with the minimum indicating the best fit. The likelihood ratio tests of full and partial models were used to show the statistical significance of the variables. All data analysis was conducted using R software, version 3.6.2 (R Foundation for Statistical Computing). The statistical significance was set to
This project was approved by the Human Subjects Ethics Subcommittee of the Hong Kong Polytechnic University. All participants signed a consent form, and no personal information except residential address was collected in this study.
We recruited 40 participants during October to December 2017. One participant withdrew from the study after the first week due to unforeseen family issues. The remaining 39 participants finished all the scheduled visits. The 39 participants had a mean age of 56.4 years (range 50-63 years); 27 (69%) were women, 31 (80%) were married, 35 (90%) were living with family members (family size range 1-6 people), and 14 (36%) received postsecondary education (
Baseline characteristics of the participants (N=39).
Characteristics |
Value | ||
Age (years), mean (SD) | 56.4 (3.5) | ||
|
|||
|
Male | 12 (31) | |
|
Female | 27 (69) | |
|
|||
|
Self-owned | 25 (64) | |
|
Rent | 14 (36) | |
|
|||
|
Alone | 4 (10) | |
|
With family member | 35 (90) | |
Household size, mean (SD) | 3.1 (1.2) | ||
|
|||
|
Secondary | 25 (64) | |
|
Postsecondary | 14 (36) | |
|
|||
|
Working | 21 (54) | |
|
Homemaker and retired | 18 (41) | |
|
|||
|
Low (10,500 or below) | 21 (54) | |
|
Medium (10,501-23,000) | 13 (33) | |
|
High (23,001 or above) | 5 (13) | |
|
|||
|
Married | 31 (80) | |
|
Single | 4 (10) | |
|
Widowed/divorced | 4 (10) | |
BMI, mean (SD) | 23.0 (3.4) |
aHK $1=US $0.13.
The weekly average NDVIs at the 100-, 250-, and 500-meter buffer zones were –0.54 (IQR 0.34), –0.61 (IQR 0.25), and –0.61 (IQR 0.25), respectively (
Weekly average of green space exposure and air pollution for the 3 buffer zones, as well as BMI and physical activity at the 250-meter buffer zone, in the 5-week follow-up period.
Variable | Value, mean (SD) | |||||||||||
|
Overall | Week 1 | Week 2 | Week 3 | Week 4 | Week 5 |
||||||
|
||||||||||||
|
100-m buffer zone | –0.54 (0.33) | –0.67 (0.35) | –0.53 (0.34) | –0.53 (0.29) | –0.46 (0.35) | –0.51 (0.33) | |||||
|
250-m buffer zone | –0.61 (0.22) | –0.70 (0.29) | –0.61 (0.19) | –0.59 (0.17) | –0.57 (0.21) | –0.61 (0.21) | |||||
|
500-m buffer zone | –0.61 (0.22) | –0.70 (0.29) | –0.60 (0.19) | –0.59 (0.17) | –0.57 (0.21) | –0.60 (0.21) | |||||
|
||||||||||||
|
100-m buffer zone | 25.57 (1.28) | 25.60 (1.62) | 25.68 (1.20) | 25.46 (1.22) | 25.67 (1.42) | 25.46 (1.07) | |||||
|
250-m buffer zone | 29.26 (1.12) | 29.07 (1.47) | 29.25 (1.00) | 29.17 (1.05) | 29.44 (1.07) | 29.35 (1.02) | |||||
|
500-m buffer zone | 29.25 (1.01) | 29.04 (1.44) | 29.25 (0.82) | 29.18 (0.95) | 29.41 (0.89) | 29.33 (0.93) | |||||
|
||||||||||||
|
100-m buffer zone | 23.46 (12.89) | 23.58 (14.84) | 23.48 (13.26) | 23.63 (12.38) | 22.78 (11.82) | 23.84 (13.36) | |||||
|
250-m buffer zone | 39.73 (5.30) | 40.43 (8.19) | 39.78 (4.00) | 39.44 (4.64) | 39.08 (4.19) | 40.00 (5.20) | |||||
|
500-m buffer zone | 39.78 (5.21) | 40.47 (7.81) | 39.83 (4.09) | 39.52 (4.56) | 39.12 (4.34) | 40.03 (5.09) | |||||
BMI (kg/m2) | 22.83 (3.26) | 22.85 (3.30) | 22.78 (3.35) | 22.89 (3.30) | 22.91 (3.29) | 22.73 (3.25) | ||||||
MVPAd (minutes) | 177.61 (76.64 | 180.77 (79.75) | 184.02 (79.01) | 176.64 (77.24) | 172.96 (77.39) | 173.91 (73.37) |
aNDVI: normalized difference vegetation index.
bPM2.5: particulate matter with a diameter less than 2.5 μm.
cNO2: nitrogen dioxide.
dMVPA: moderate to vigorous physical activity.
Summary statistics of the outcome measurements at the weekly visits.
Variable | Outcome, mean (SD) | ||||||
|
|
Week 1 | Week 2 | Week 3 | Week 4 | Week 5 | |
|
|||||||
|
FVCa (%) | 100.69 (9.65) | 99.85 (8.39) | 99.67 (10.26) | 100.66 (10.06) | 98.84 (9.54) | |
|
FEV1b (%) | 103.76 (10.43) | 102.11 (9.48) | 102.01 (11.19) | 103.01 (10.62) | 101.17 (10.30) | |
|
FEV1/FVC ratio | 103.11 (5.02) | 102.33 (5.63) | 102.37 (5.29) | 102.44 (5.90) | 102.44 (5.46) | |
|
Heart rate (beats per minute) | 71.67 (10.13) | 69.46 (9.17) | 69.44 (9.25) | 71.05 (9.68) | 70.28 (9.12) | |
|
|||||||
|
Systolic | 127.64 (15.78) | 123.59 (16.28) | 124.08 (14.24) | 122.77 (16.98) | 122.92 (13.02) | |
|
Diastolic | 79.90 (12.30) | 77.46 (9.60) | 77.46 (10.50) | 76.18 (10.18) | 75.82 (9.07) | |
|
|||||||
|
Total cholesterol (mmol/L) | 5.86 (0.91) | 5.72 (1.02) | 5.77 (1.07) | 5.78 (1.09) | 5.52 (1.05) | |
|
Triglycerides (mmol/L) | 1.48 (0.82) | 1.43 (0.80) | 1.44 (0.72) | 1.53 (0.86) | 1.45 (0.74) | |
|
LDL-Cc (mmol/L) | 3.83 (0.84) | 3.73 (0.92) | 3.76 (0.97) | 3.73 (0.95) | 3.57 (0.92) | |
|
HDL-Cd (mmol/L) | 1.47 (0.34) | 1.44 (0.34) | 1.45 (0.33) | 1.45 (0.36) | 1.38 (0.32) | |
|
Total cholesterol/HDL-C ratio | 4.19 (1.12) | 4.17 (1.15) | 4.16 (1.10) | 4.18 (1.10) | 4.17 (1.11) | |
|
Triglycerides/HDL-C ratio | 1.15 (0.88) | 1.14 (0.91) | 1.12 (0.76) | 1.20 (0.86) | 1.18 (0.85) | |
|
LDL-C/HDL-C ratio | 2.77 (0.90) | 2.75 (0.95) | 2.74 (0.92) | 2.72 (0.89) | 2.72 (0.89) | |
|
Glucose (mmol/L) | 5.75 (0.82) | 5.60 (0.78) | 5.59 (0.72) | 5.63 (0.68) | 5.45 (0.74) | |
|
|||||||
|
hs-CRPe (mg/L) | 1.98 (3.11) | 1.56 (3.20) | 1.01 (0.76) | 1.07 (1.20) | 1.05 (0.82) | |
|
IL-6f (pg/mL) | 1.72 (2.40) | 1.78 (2.58) | 1.31 (1.66) | 1.74 (2.26) | 1.25 (1.75) | |
|
TNF-αg (pg/mL) | 1.6 (1.22) | 1.65 (1.25) | 1.49 (1.22) | 1.73 (1.45) | 1.50 (1.27) | |
|
sP-selectinh (ng/ml) | 3.51 (2.62) | 3.20 (2.14) | 2.90 (1.98) | 3.27 (1.89) | 3.24 (1.92) | |
|
|||||||
|
MDAi (μM) | 40.85 (22.26) | 31.83 (21.24) | 30.11 (18.88) | 36.35 (23.69) | 31.75 (19.76) | |
|
8-OHdGj (ng/mL) | 1.32 (0.54) | 1.41 (0.54) | 1.22 (0.39) | 1.37 (0.49) | 1.47 (0.55) | |
|
SODk (U/μL) | 0.77 (0.38) | 0.74 (0.35) | 0.77 (0.32) | 0.79 (0.36) | 0.80 (0.50) | |
|
GPx-1l (ng/mL) | 3.25 (7.06) | 3.17 (6.91) | 2.79 (6.31) | 2.54 (5.84) | 4.25 (10.8) |
aFVC: forced vital capacity.
bFEV1: forced expiratory volume in the first second.
cLDL-C: low-density lipoprotein cholesterol.
dHDL-C: high-density lipoprotein cholesterol.
ehs-CRP: high-sensitivity C-reactive protein.
fIL-6: interleukin-6.
gTNF-α: tumor necrosis factor-α.
hsP-selectin: soluble platelet selectin.
iMDA: malondialdehyde.
j8-OHdG: 8-hydroxy-2'-deoxyguanosine.
kSOD: superoxide dismutase.
lGPx-1: glutathione peroxidase 1.
We compared the AICs of the different GLME models and found that model 3 with the 100-meter NDVI, model 1 with the 250-meter NDVI, and model 1 with the 500-meter NDVI generally returned smaller AICs for most biomarkers than the rest of the models (Table S2,
Percentage change in metabolic biomarker concentrations associated with per IQR increase in normalized difference vegetation index (NDVI) at (A) the 100-meter buffer zone, (B) the 250-meter buffer zone, and (C) the 500-meter buffer zone. Vertical bars are 95% CI. Note: model 1: outcome ~ NDVI; model 2: outcome ~ NDVI + covariates (sex, income, occupation, moderate to vigorous physical activity, noise); model 3: outcome ~ NDVI + covariates + PM2.5; model 4: outcome ~ NDVI + covariates + NO2; model 5: outcome ~ NDVI + covariates + protein + carbohydrates. hs-CRP: high-sensitivity C-reactive protein, IL-6: interleukin-6, TNF-α: tumor necrosis factor α, 8-OHdG: 8-hydroxy-2'-deoxyguanosine, SOD: superoxide dismutase, GPx-1: glutathione peroxidase.
Percentage change in inflammatory and oxidative biomarker concentrations associated with per-IQR increase in normalized difference vegetation index (NDVI) at (A) the 100-meter buffer zone, (B) the 250-meter buffer zone, and (C) the 500-meter buffer zone. Vertical bars are 95% CIs. Note: model 1: outcome ~ NDVI; model 2: outcome ~ NDVI + covariates (sex, income, occupation, moderate to vigorous physical activity, noise); model 3: outcome ~ NDVI + covariates + PM2.5; model 4: outcome ~ NDVI + covariates + NO2; model 5: outcome ~ NDVI + covariates + protein + carbohydrates. LDL-C: low-density lipoprotein, HDL-C: high-density lipoprotein, TC: total cholesterol, TG: triglyceride.
In this panel study, we found a negative association of NDVI exposure with different metabolic and inflammatory biomarkers. The findings were generally consistent with the beneficial effects of neighborhood green space on biomarkers of respiratory and cardiovascular health in the literature [
After adjustment for air pollution, physical activities, and dietary intake, we found that lower levels of TC, LDL-C, and fasting glucose were associated with higher green space exposure. Our findings are consistent with 2 cross-sectional studies and 1 cohort study in Chinese populations, which reported that a larger amount of green space in working places or residential areas was associated with lower levels of TC, TG, LDL-C, and fasting glucose [
Compared to the metabolic biomarkers, the associations of green space and proinflammatory biomarkers were less significant and inconsistent in our study. We found negative associations between green space exposure and hs-CRP, although they were not statistically significant. This echoes the findings of a cohort study of school-aged children in Portugal [
Elevated MDA and lower SOD levels have been linked to increased risks of coronary artery disease, heart failure, and other chronic diseases [
We did not find any significant effects of green space on lung functions, except FEV1/FVC. This could be due to the fact that the participants were relatively healthy, without any pre-existing chronic respiratory conditions. Sinharay et al [
Similar to previous green space studies [
Our study had several limitations. First, the small sample size may render low statistical power, which could explain the few significant effect estimates and wide confidence intervals. Therefore, additional studies with a larger sample size are needed to further elucidate the inconsistent findings. Second, green space exposure was measured by daily average NDVI, which could not reflect the activities performed by participants around the green spaces. Nevertheless, we simultaneously collected the physical activities by personal trackers, which can reduce the confounding effect of these activities. Third, due to a limited budget, we only tested a selected panel of biomarkers, although we attempted to cover a wide range of biomarkers for metabolism, respiratory functions, oxidative stress, and proinflammation. Future studies could adopt more biomarkers to gain a comprehensive understanding of the pathways involved in health effects of green space. Last but not least, sampling bias may exist due to the convenience sampling approach we used in this study. The volunteers tend to be healthier and more educated than the general population, as shown in
By combining data collected via personal tracking devices with green space GIS data, we were able to demonstrate that higher exposure to green space was associated with a better lipid profile and lower inflammatory biomarkers; however, no significant associations were found with respiratory and oxidative biomarkers. The findings of this study provide more clues to the potential biological pathways for the health benefits of green space. From the public health perspective, the health effects of green space identified from this study will also aid the design of future intervention programs to improve the quality of life of the general public.
Supplementary materials.
8-hydroxy-2'-deoxyguanosine
Akaike information criterion
chronic obstructive pulmonary disease
copper–zinc superoxide dismutase
enzyme-linked immunosorbent assay
forced expiratory volume in the first second
forced vital capacity
generalized linear mixed-effects model
glutathione peroxidase
high-density lipoprotein
heart rate
high-sensitivity C-reactive protein
Integrative Health Clinic
interleukin-6
low-density lipoprotein
malondialdehyde
normalized difference vegetation index
nitrogen dioxide
particulate matter with a diameter less than 2.5 μm
soluble platelet selectin
Satellite Pour l’Observation de la Terre
total cholesterol
triglyceride
tumor necrosis factor α
variance inflation factor
We thank Fiona Y Wong and Shi Zhao for their assistance in data analysis.
This work was supported by the Faculty of Health and Social Science Dean’s Reserve Fund from the Hong Kong Polytechnic University.
None declared.