Abstract
Greenness exposure is nominated as a potential beneficial factor for health, but evidence is limited on its diabetes effects. We conducted a cross-sectional study between May and September 2016 in rural areas of northwestern China, including 4670 Uyghur adults, to explore the associations between residential greenness and fasting glucose levels and diabetes prevalence. Fasting glucose levels were determined, and information on covariates was collected by questionnaire. Normalized difference vegetation index (NDVI) and soil-adjusted vegetation index (SAVI) were calculated to assess greenness levels. Generalized linear mixed models were applied to evaluate the associations of greenness with fasting glucose levels and diabetes prevalence. The prevalence of diabetes was 11.6%. We found that living in rural areas characterized by increased amounts of greenness was associated with reduced diabetes prevalence (e.g., NDVI1000m: OR, 0.92; 95% CI, 0.86, 0.99). Stratified analyses showed that the protective effects of greenness on diabetes prevalence were found only in women (NDVI1000m: OR, 0.90; 95% CI, 0.82, 0.99). However, none of the interaction was statistically significant. Our study suggests that greater residential greenness levels were associated with a lower odds ratio of diabetes prevalence in Xinjiang Uyghur adults. Further well-designed longitudinal studies are needed to confirm our findings.
Keywords: greenness, diabetes, fasting glucose, Uyghur
1. Introduction
Diabetes mellitus is one of the major chronic diseases worldwide, and its prevalence has increased in recent years [1]. In China, diabetes mellitus has also become a public health problem, and the most recent national representative survey showed that the overall prevalence of diabetes in Chinese adults has increased from 9.7% in 2007 to 11.6% in 2010, representing an estimated 114 million Chinese adults with diabetes mellitus [2]. Therefore, there is an urgent need to take effective intervention strategies to control the uprising trend.
The etiology of diabetes mellitus is complex and is thought to be influenced by both genetic and environmental factors [3]. For environmental factors, high-sugar, high-fat, and high-energy diet habits, lack of physical activity, being overweight/obese, and exposure to air pollution have been reported to increased risk of diabetes mellitus [4,5,6]. Recently, diabetic effects of greenness have attracted more and more researchers’ attention worldwide. Mechanistically, greenness can bring health benefits by reducing levels of ambient air pollutants, extreme heat waves, and noise exposure, by encouraging physical activity, by reducing adiposity, and by enriching microbial diversity [7]. Several studies have explored the associations of greenness exposure with glucose levels, insulin resistance, and diabetes mellitus risk, but showed inconsistent results [8,9,10,11,12,13,14,15,16,17,18,19,20]. For example, cross-sectional studies conducted in Netherlands, Australia, the United States, the United Kingdom, and Canada have found that green space had a protective effect on diabetes mellitus [13,14,15,17,18,20], while a cross-sectional study among the Norwegian population found no significant association [8]. Two prospective cohort studies from England and Canada observed a significant association between increased greenness exposure and lower relative hazard of developing diabetes mellitus [9,16], but the Framingham study did not find such a longitudinal association [11]. While most prior studies were conducted among adults, a few have also explored potential diabetic effects of greenness among children and adolescents [12,19]. For example, a study of 837 German adolescents identified significant associations between higher greenness and lower insulin resistance [12]. Similarly, in a study of 3844 Iranian children, Dadvand et al. observed inverse associations between time spent in green spaces and fasting blood glucose levels and the risk of impaired fasting glucose [19]. Despite this, the prior studies were mostly conducted in developed countries, and only one study was conducted in China [10].
Additionally, although most prior studies on greenness and health were conducted in urban areas, few were performed among people living in rural areas (i.e., unurbanized areas like countryside or village). Xinjiang Uyghur Autonomous Region is located in northwest of China and is the largest province in China. The Uyghur is the largest inhabitant minority group in the province, whose lifestyle, culture, religion, genetic, and diet habits are different from Han ethnicities. It is also usually reported that the prevalence of diabetes mellitus is high in this region [21]. Further, Xinjiang belongs to the typical temperate continental arid climate, where it is dry, receiving little rainfall (with the mean annual precipitation of 100 mm in Southern Xinjiang), with little green space [22]. Additionally, rural areas in Xinjiang are economically backward with low health service, and people may have low health awareness, thus, rural population may be more vulnerable to environmental exposure [23]. Therefore, the aim of the current study was to evaluate the effects of residential greenness on diabetes prevalence and fasting glucose levels among rural Uyghur adults in Xinjiang Uyghur Autonomous Region.
2. Materials and Methods
2.1. Study Participants and Inclusion Criteria
The current study was conducted between May and September 2016 in Kashgar city. The city is an important economic, political, and cultural center in northwestern China. Due to Xinjiang having vast land area, low population density, and inconvenient transportation, we adopted a multistage city–county–towns–village stratified cluster random sampling method to select participants for saving cost. We chose Shufu county of Kashagar city as the study area. Shufu county consists of 10 towns with 123 villages (16 villages in Bulacksou town, 11 in LanGan town, 9 in Mushi town, 16 in Sayibarg town, 16 in Tashmeric town, 4 in Tierimu town, 8 in Toquzak town, 16 in WuPal town, 8 in Wucusack town, and 19 in Zhanmin town). Population ranged between 2000 and 30,000 among the villages. Over 290,000 people reside in Shufu county, of whom 98% are Uyghurs. First, we selected one village from each of the 10 towns using the random number method. Second, from each study village, all adults aged ≥18 years old and who had lived in the village for at least two years were selected for study enrollment.
A total of 5087 potential individuals were invited to take part in the current study, of whom 4772 individuals returned the questionnaire, yielding a response rate of 93.81%. We excluded 102 individuals due to having malignant diseases (e.g., chronic liver disease, renal failure, or cancer), being pregnant, and absence of a blood sample for glucose testing. Finally, a total of 4670 participants were included in the current analysis. The study procedures and protocols were reviewed and approved by the Ethics Committee of Guangzhou Center for Disease Control and Prevention prior to the present study (Identification code: GZCDC-ER[A]2016007). Written informed consent was obtained from all participants before all survey data and specimens were collected.
2.2. Data Collection and Diabetes Mellitus Definition
We used a self-administered questionnaire delivered by a face-to-face interview to collect information on demographic characteristics (e.g., age, sex, education levels, and marital status), tobacco use and alcohol drinking status, and physical activity levels. As detailed in our previous paper [24], tobacco use and alcohol drinking status were categorized as current, former, and non-smokers or drinkers, while physical activity levels were categorized as low, moderate, and high levels.
All participants were asked for permission to collect a blood sample for biochemical analysis after an overnight fasting (≥8 h). Fasting glucose concentration was determined by an enzymatic colorimetric method, using a Roche Autoanalyzer (Cobas c702 type; Roche Ltd.; Mannheim, Germany) in Guangzhou Center for Disease Control and Prevention.
Diabetes mellitus is defined as the fasting glucose level ≥7.0 mmol/L and/or a physician diagnosis and/or currently receiving pharmacological treatment for diabetes, according to the recommendations of the American Diabetes Association [25].
2.3. Confounders and Mediators
We chose confounders according to the recommendations by Jager et al. [26], in which a confounder has to meet the following three criteria: (1) it should be a risk factor for diabetes or elevated blood glucose; (2) it must be a cause of the greenness exposure and unequally distributed among participants with different greenness levels; and (3) it must not be an “effect” of greenness exposure, nor be an intermediate factor in the causal pathway of diabetes. Then, we developed a directed acyclic graph (DAG, Figure S1) to select a minimally sufficient set of covariates to adjust for confounding [27]. According to the DAG, the following confounders were retained: age, sex, marital status, and educational levels. Also, physical activity was chosen as a potential mediator (Figure S1).
2.4. Residential Greenness
We defined residential greenness using two satellite-based vegetation indexes, normalized difference vegetation index (NDVI) [28] and soil-adjusted vegetation index (SAVI) [29]. Derivations of both NDVI and SAVI indexes were calculated according to the land surface reflectance of the visible red and near-infrared parts of the light wavelengths. SAVI is similar to NDVI while additionally includes a correction factor to minimize soil influences. Both NDVI and SAVI values range from −1 (water) to +1 (completely vegetated areas). We used two cloud-free Landsat 5 Thematic Mapper satellite images at a resolution of 30 m (http://earthexplorer.usgs.gov) obtained in September 2017 to calculate the greenness levels. Both NDVI and SAVI were defined as mean values in circular buffers of 100 m, 300 m, 500 m, and 1000 m around each participant’ residential address centroid. These calculations were conducted by the ArcGIS 10.4 (ESRI, Redlands, CA, USA).
2.5. Statistical Analysis
Means and standard deviations were used to describe normally distributed continuous variables (e.g., age, body mass index (BMI), and fasting glucose levels). Frequency and percentage were used to describe the categorical variables (e.g., sex, education level, marital status, smoking and drinking status, and physical activity). Student’s t-tests were used to test differences in mean levels for continuous normally distributed variables between non-diabetes mellitus and diabetes mellitus groups, and chi-square tests were used to test differences in percentages for categorical variables. Spearman correlation analysis was performed to evaluate the intercorrelations between NDVI and SAVI.
Generalized linear mixed models (GLMMs) were used to explore the association between greenness with fasting glucose levels and diabetes mellitus prevalence, as described previously [10]. In the models, study village was incorporated as random effect and greenness and covariates were included as fixed effects. The effect estimates are presented as regression coefficients (β) and odds ratios (OR) and their corresponding 95% confidence intervals (CI). Basic regression models were unadjusted, and we further evaluated models with adjustment for confounders selected using DAG. We also built models with additional adjustment for smoking and alcohol status to explore their potential impact. Then, we categorized NDVI1000m and SAVI1000m into quartiles to test for nonlinear relationships. Additionally, we assessed the associations between greenness and fasting glucose levels with diabetes mellitus by sex (female vs. male) and age (<48 years vs. ≥48 years). We also explored potential mediation effects of physical activity by comparing the effect estimates before and after additional adjustment for physical activity [30]. For each model, effect estimates for fasting glucose levels and diabetes mellitus were calculated per interquartile range (IQR) increase in NDVI and SAVI. All statistic analyses were performed on SAS software (version 9.4; SAS institute, Inc. Cary, NC, USA). Levels of statistical significance were set at an α level of 0.05.
3. Results
3.1. Study Population Characteristics
Mean age of the participants was 47.2 years old, and 64.7% were women (Table 1). About half of the participants (49.3%) had low educational level (i.e., primary school), 88.6% were self-reported nonsmokers, 97.0% were self-reported nondrinkers, and 93.2% had low physical activity level. The overall prevalence of diabetes mellitus was 11.6%, of whom 110 reported a physician-diagnosed diabetes or were receiving antidiabetic medicine. Compared with individuals without diabetes mellitus, those with diabetes mellitus manifested significantly higher age, BMI, and fasting glucose levels, and were more likely to be male and have lower education levels (no school and primary school) (all p < 0.05).
Table 1.
Characteristics of the study participants.
| Variables | No Diabetes Mellitus (n = 4127) | Diabetes Mellitus (n = 543) | p |
|---|---|---|---|
| Age, years | 46.4 ± 14.2 | 53.6 ± 13.1 | <0.001 |
| Sex, n (%) | |||
| Men | 1418 (34.4) | 229 (42.2) | <0.001 |
| Women | 2709 (65.6) | 314 (57.8) | |
| Education level, n (%) | |||
| No school | 370 (9.0) | 78 (14.4) | <0.001 |
| Primary school | 2002 (48.5) | 299 (55.1) | |
| Middle school | 1497 (36.3) | 136 (25.1) | |
| >Junior college | 258 (6.3) | 30 (5.5) | |
| Marital status, n (%) | |||
| Not married | 164 (4.0) | 9 (1.7) | 0.025 |
| Get married | 3577 (86.7) | 479 (88.2) | |
| Divorce or widowed | 386 (9.4) | 55 (10.1) | |
| Smoking status, n (%) | |||
| Non-smokers | 3662 (88.7) | 474 (87.3) | 0.443 |
| Former-smokers | 380 (9.2) | 59 (10.9) | |
| Current-smokers | 85 (2.1) | 10 (1.8) | |
| Drinking status, n (%) | |||
| Non-drinkers | 3998 (96.9) | 533 (98.2) | 0.238 |
| Former-drinkers | 81 (2.0) | 7 (1.3) | |
| Current-drinkers | 48 (1.2) | 3 (0.6) | |
| Physical activity, n (%) | |||
| Low strength | 3848 (93.2) | 506 (93.2) | 0.963 |
| Moderate or high strength | 279 (6.8) | 37 (6.8) | |
| Body mass index, kg/m2 | 24.9 ± 4.7 | 26.3 ± 4.8 | <0.001 |
| Fasting glucose, mmol/L | 5.3 ± 0.7 | 9.6 ± 4.5 | <0.001 |
Table 2 shows distributions and correlations of NDVI and SAVI. Greenness levels of the study participants differed markedly among the study participants. For example, NDVI1000m ranged from 0.28 to 0.57 with a median value of 0.47. Additionally, we observed that NDVI was significantly correlated with SAVI, and Spearman’s correlation coefficients ranged from 0.40 to 0.99 (all p < 0.01).
Table 2.
Distributions and intercorrelations (Spearman correlation coefficients) for NDVI and SAVI.
| Median (IQR) | Min | Max | NDVI100m | NDVI300m | NDVI500m | NDVI1000m | SAVI100m | SAVI300m | SAVI500m | SAVI1000m | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| NDVI100m | 0.45 (0.20) | 0.20 | 0.67 | 1 | |||||||
| NDVI300m | 0.48 (0.12) | 0.21 | 0.63 | 0.74 ** | 1 | ||||||
| NDVI500m | 0.47 (0.09) | 0.23 | 0.57 | 0.57 ** | 0.88 ** | 1 | |||||
| NDVI1000m | 0.47 (0.06) | 0.28 | 0.57 | 0.42 ** | 0.63 ** | 0.82 ** | 1 | ||||
| SAVI100m | 0.29 (0.14) | 0.13 | 0.44 | 0.99 ** | 0.74 ** | 0.55 ** | 0.42 ** | 1 | |||
| SAVI300m | 0.32 (0.08) | 0.14 | 0.42 | 0.71 ** | 0.99 ** | 0.90 ** | 0.63 ** | 0.71 ** | 1 | ||
| SAVI500m | 0.32 (0.06) | 0.15 | 0.39 | 0.53 ** | 0.86 ** | 0.99 ** | 0.79 ** | 0.52 ** | 0.89 ** | 1 | |
| SAVI1000m | 0.32 (0.04) | 0.19 | 0.39 | 0.41 ** | 0.64 ** | 0.84 ** | 0.99 ** | 0.40 ** | 0.65 ** | 0.82 ** | 1 |
Abbreviations: IQR, interquartile range (computed by subtracting the first quartile from the third quartile); Max, maximum; Min, minimum; NDVI, normalized difference vegetation index; SAVI, soil-adjusted vegetation index. ** Correlation is significant at the 0.01 level (two-tailed).
3.2. Relationship of Greenness with Fasting Glucose and Diabetes Mellitus
Table S1 and Table 3 summarize the results of linear regression analysis for greenness with fasting glucose levels in crude and adjusted models, respectively. We observed that there was no significant association for any buffers of NDVI and SAVI with fasting glucose levels.
Table 3.
Associations of NDVI and SAVI (per IQR increase) around the residential address with fasting glucose levels and diabetes mellitus prevalence (adjusted models).
| Fasting Glucose | Diabetes Mellitus | |||
|---|---|---|---|---|
| β (95% CI) * | p | OR (95% CI) * | p | |
| NDVI100m | 0.04 (−0.07, 0.14) | 0.487 | 0.97 (0.83, 1.13) | 0.683 |
| NDVI300m | −0.01 (−0.08, 0.07) | 0.864 | 0.92 (0.82, 1.03) | 0.128 |
| NDVI500m | −0.01 (−0.08, 0.05) | 0.657 | 0.92 (0.84, 1.01) | 0.089 |
| NDVI1000m | −0.02 (−0.06, 0.03) | 0.531 | 0.92 (0.86, 0.99) | 0.024 |
| SAVI100m | 0.07 (−0.04, 0.18) | 0.238 | 1.00 (0.85, 1.17) | 0.966 |
| SAVI300m | −0.01 (−0.09, 0.07) | 0.836 | 0.92 (0.82, 1.03) | 0.132 |
| SAVI500m | −0.02 (−0.08, 0.04) | 0.540 | 0.92 (0.84, 1.01) | 0.075 |
| SAVI1000m | −0.02 (−0.07, 0.03) | 0.458 | 0.92 (0.86, 0.99) | 0.020 |
Abbreviations: IQR, interquartile range; NDVI, normalized difference vegetation index; SAVI, soil-adjusted vegetation index; OR, odds ratio. * Adjusted for age, sex, marital status, and educational levels.
In crude logistic regression analysis, an IQR increase in NDVI1000m and SAVI1000m was significantly associated with 7% (NDVI1000m: OR = 0.93, 95% CI = 0.87, 1.00; SAVI1000m: OR = 0.93, 95% CI = 0.87, 0.99) reduction in odds for diabetes risk (Table S1). The effect estimates were not influenced by further adjustments for age, sex, marital status, and education levels (Table 3). For example, an IQR increase in NDVI1000m and SAVI1000m was associated with 8% (both OR = 0.92, 95% CI = 0.86, 0.99) reduction in odds for diabetes risk. However, no significant associations were found for NDVI and SAVI in smaller buffers.
We additionally adjusted the main models for smoking and alcohol drinking status and found that the effect estimates were not materially changed (Table S2). Also, we performed mediation analysis but did not find any mediating role of physical activity (Table S3). Further, we categorized NDVI1000m and SAVI1000m into quartiles and found that effects estimates were lower in higher quartiles than in lower quartiles, and p values for trend were significant for all (Table S4).
3.3. Relationship of Greenness with Fasting Glucose and Diabetes Mellitus Stratified by Sex and Age
Table 4 shows the sex- and age-stratified associations for residential greenness and fasting glucose levels. We found that the relationship of residential greenness with fasting glucose levels failed to reach significance in both sex and age subgroups. And the p values for the interactions were not significant.
Table 4.
Sex- and age-stratified associations between per IQR increase in NDVI and SAVI around the residential address with fasting glucose levels (adjusted models).
| Males | Females | <48 | ≥48 | |
|---|---|---|---|---|
| β (95% CI) * | β (95% CI) * | β (95% CI) * | β (95% CI) * | |
| NDVI100m | 0.07 (−0.10, 0.24) | 0.00 (−0.13, 0.13) | 0.08 (−0.04, 0.19) | 0.01 (−0.15, 0.16) |
| NDVI300m | −0.04 (−0.17, 0.09) | 0.01 (−0.09, 0.10) | 0.05 (−0.03, 0.14) | −0.04 (−0.15, 0.08) |
| NDVI500m | −0.06 (−0.16, 0.05) | 0.01 (−0.07, 0.09) | 0.02 (−0.06, 0.09) | −0.02 (−0.11, 0.07) |
| NDVI1000m | −0.04 (−0.12, 0.05) | −0.01 (−0.07, 0.05) | −0.03 (−0.08, 0.03) | 0.00 (−0.07, 0.07) |
| SAVI100m | 0.11 (−0.08, 0.29) | 0.03 (−0.01, 0.17) | 0.11 (−0.01, 0.23) | 0.04 (−0.13, 0.20) |
| SAVI300m | −0.03 (−0.16, 0.10) | 0.00 (−0.09, 0.10) | 0.05 (−0.03, 0.14) | −0.04 (−0.15, 0.08) |
| SAVI500m | −0.06 (−0.16, 0.05) | −0.01 (−0.08, 0.08) | 0.01 (−0.06, 0.09) | −0.03 (−0.12, 0.07) |
| SAVI1000m | −0.04 (−0.12, 0.04) | −0.01 (−0.07, 0.05) | −0.03 (−0.08, 0.03) | −0.01 (−0.08, 0.07) |
Abbreviations: IQR, interquartile range; NDVI, normalized difference vegetation index; SAVI, soil-adjusted vegetation index. * Adjusted for age, marital status, and educational levels.
Table 5 shows the results of analyses for residential greenness in relation to diabetes mellitus stratified by sex and age. The protective association of greenness and diabetes prevalence was only found in women (e.g., IQR increases in NDVI1000m and SAVI1000m were both associated with a 10% (NDVI1000m: OR = 0.90, 95% CI = 0.82, 0.99; SAVI1000m: OR = 0.90, 95% CI = 0.82, 0.98) reduction in odds for diabetes mellitus prevalence). However, none of the interactions were statistically significant. When the analysis was stratified by age, the relationship of residential greenness with diabetes prevalence failed to reach significance in both <48 and ≥48 age groups.
Table 5.
Sex- and age-stratified associations between per IQR increase in NDVI and SAVI around the residential address with diabetes mellitus prevalence (adjusted models).
| Males | Females | <48 | ≥48 | |
|---|---|---|---|---|
| OR (95% CI) * | OR (95% CI) * | OR (95% CI) * | OR (95% CI) * | |
| NDVI100m | 1.01 (0.79, 1.28) | 0.94 (0.77, 1.15) | 0.85 (0.61, 1.17) | 1.01 (0.85, 1.20) |
| NDVI300m | 0.96 (0.80, 1.14) | 0.87 (0.76, 1.03) | 0.93 (0.73, 1.19) | 0.92 (0.81, 1.05) |
| NDVI500m | 0.96 (0.83, 1.02) | 0.90 (0.80, 1.01) | 0.95 (0.77, 1.16) | 0.93 (0.84, 1.03) |
| NDVI1000m | 0.95 (0.85, 1.07) | 0.90 (0.82, 0.99) | 0.89 (0.76, 1.03) | 0.94 (0.87, 1.02) |
| SAVI100m | 1.05 (0.81, 1.35) | 0.96 (0.77, 1.20) | 0.88 (0.62, 1.24) | 1.04 (0.86, 1.26) |
| SAVI300m | 0.96 (0.80, 1.15) | 0.89 (0.76, 1.03) | 0.95 (0.74, 1.21) | 0.92 (0.81, 1.05) |
| SAVI500m | 0.95 (0.82, 1.11) | 0.89 (0.79, 1.01) | 0.95 (0.78, 1.17) | 0.92 (0.83, 1.03) |
| SAVI1000m | 0.95 (0.85, 1.07) | 0.90 (0.82, 0.98) | 0.90 (0.77, 1.05) | 0.93 (0.86, 1.01) |
Abbreviations: IQR, interquartile range; NDVI, normalized difference vegetation index; SAVI, soil-adjusted vegetation index, OR, odds ratio. * Adjusted for age, marital status, and educational levels.
4. Discussion
The findings from the population-based study of 4670 Uyghur adults showed that living in greener areas was associated with reduced diabetes mellitus prevalence. In addition, neither age nor sex significantly modified the associations. This is the first study to examine the relationship between residential greenness and diabetic metrics among Uyghur rural population.
The relationships of greenness and diabetes mellitus have received a great deal of attention during the past decade. For instance, Maas et al. conducted a large cross-sectional study of 345,143 individuals in Dutch and found that the annual prevalence rate of diabetes mellitus was lower in people living in areas with higher green space levels [17]. Another study of 267,072 adult Australians (≥45 years old) found that the rate of self-report diabetes mellitus was decreased with the increased green space of residential neighborhoods [13]. A study of 249,405 USA elderly (≥65 years) also found greenness significantly reduced diabetes mellitus risk; an increase in NDVI from 1 SD less to 1 SD more than mean was associated with a reduced risk of 14% for diabetes mellitus [14]. Likewise, another three cross-sectional studies conducted in USA [15], the United Kingdom [18], and Canada [20] observed increased neighborhood or street green space was significantly associated with lower individual diabetes mellitus prevalence. Additionally, several cohort studies have also documented a significant and inverse relationship between greenness levels and diabetes mellitus incidence [9,16]. However, there are also a few studies that found no association between them [8,11]. Our research group has also investigated the association of greenness with diabetes mellitus in Chinese Han adults and found that higher residential greenness was significantly associated with a lower prevalence of diabetes mellitus [10]. Our current findings were consistent with most of the prior studies that residential greenness was associated with diabetes mellitus prevalence.
However, we did not find an association of residential greenness with fasting glucose levels. To the best of our knowledge, there were only three studies that have explored association between them. Specifically, in a cross-sectional study of 3844 Iranian schoolchildren aged 7–18 years, Dadvand et al. found that more time spent in green spaces, especially in natural green spaces, was significantly associated with decreased fasting glucose levels [19]. In the Framingham Heart Study, Lee and colleagues showed that higher green space was associated with lower fasting glucose levels in cross-sectional analysis, although the association disappeared in longitudinal analysis [11]. Additionally, the cross-sectional study conducted in the Chinese adults found that a 0.1-unit increase in NDVI500m was significantly associated with a 1.14% reduction in fasting glucose levels [10]. Inconsistent with these cross-sectional findings, our current study did not observe an inverse association of residential greenness with fasting glucose levels among Uyghur adults. Additionally, we did not detect a significant modification effect of sex and age on the associations between greenness and diabetes. The current evidence in this topic is mixed. For example, consistent with our findings, Dadvand et al. [19] did not find a difference in the strength of association of greenness and fasting glucose between Iranian boys and girls. However, in the study by Yang et al., which we have already described, stronger associations of greenness with fasting glucose levels were observed in men than in women [10]. Also, unlike our study, Yang et al. reported that the association of greenness with diabetes mellitus prevalence was stronger in younger participants, while no effect modification by age was observed for fasting glucose levels [10]. Further studies with large sample size and longitudinal design therefore are needed to validate these findings.
The precise mechanisms by which greenness benefits health are not fully understood. Several investigators have proposed that green spaces might have beneficial effects on glucose homeostasis markers and the development of diabetes mellitus via actively filtering out ambient air pollutions levels [12]. Ambient air pollutions, such as NO, NO2, PM1, PM2.5, and PM10, were significantly associated with increased risk of diabetes mellitus, which was observed in many previous studies [5,31,32]. Green space would reduce ambient air pollutions levels. Thus, the association of residential greenness with fasting glucose and diabetes mellitus risk may be plausible. Furthermore, previous studies reported a significant negative relationship of green space with adiposity [33,34], which is a major risk factor for diabetes mellitus. Therefore, it is also proposed that greenness may benefit glucose metabolism by reducing adiposity. Reducing physical activity is also proposed as a potential mechanism by which greenness affects health. We explored mediation analysis to explore this but failed to detect a mediation role of physical activity. Additionally, greenness was also associated with psychosocial stress, reduced noise exposure, and greater and more diverse microbial exposure, all of them might influence glucose homeostasis and the development of diabetes mellitus [9,35,36,37].
Our study has several limitations, which need to be acknowledged. First, the cross-sectional design of the study precludes us to establish a temporal association between greenness exposure with fasting glucose levels and diabetes mellitus risk. However, the possibility of the reverse causality, that is, people with diabetes moved to areas with lower greenness levels, is very low. Second, although the sample size of our study was relatively large, the statistical power remained not ideal and thus most associations were not statistically significant, especially in linear regression analyses. Therefore, future well-designed studies with large sample size are expected to validate our findings. Third, although we calculated greenness exposure levels using high-resolution satellite images, the NDVI and SAVI just represent general vegetation levels but cannot distinguish the type, utility, structure, and quality of green space, which also may lead to exposure misclassification. Fourth, data on covariates were collected by questionnaire, so recall bias cannot be ignored because the responses might be subject to the participant’s memory influence. Fifth, although we considered and adjusted for some confounders, information on potential mediators or confounders, such as air pollutions, mental stress, dietary habits, noise exposure, and time spent of green space, were unavailable in the study. Therefore, we cannot further explore potential mechanisms underlying the association between greenness and diabetes, and residual confounding caused by unmeasured confounders (e.g., village-level income) remains possible. Finally, we defined diabetes based on a single point measurement of fasting glucose, and detailed data on fasting period were not available, which might have caused outcome misclassification and precludes us from performing sensitivity analyses. Nonetheless, our study still has several apparent strengths. First, the participants are homogenous, which improves the effectiveness of statistical analysis, because they were of Uyghur nationality and from the same region of Kashgar. Second, to the best of our knowledge, our findings provide the first evidence on the association of residential greenness and diabetes mellitus risk in Uyghur population in Kashgar region.
5. Conclusions
In summary, exposure to higher residential greenness (large buffer in particular) was significantly associated with a lower odds ratio of diabetes mellitus in Xinjiang Uyghur nationality adults. However, further well-designed longitudinal studies with larger sample size are needed to confirm our findings when considering the limitations of our study.
Acknowledgments
We gratefully acknowledge the assistance and cooperation of the faculty and staff of Shufu Center for Disease Control and Prevention and thank all of the participants in our study.
Supplementary Materials
The following are available online at https://www.mdpi.com/1660-4601/16/24/5131/s1, Figure S1: Directed acyclic graph for the association between greenness and diabetes. Pink lines indicate potential confounders, and green lines indicate potential mediators. Table S1: Associations of NDVI and SAVI (per IQR increase) around the residential address with fasting glucose levels and diabetes mellitus prevalence (crude models). Table S2: Associations of NDVI and SAVI (per IQR increase) around the residential address with fasting glucose levels and diabetes mellitus prevalence after additionally adjusting main models for smoking and alcohol drinking status. Table S3: Associations of NDVI and SAVI (per IQR increase) around the residential address with fasting glucose levels and diabetes mellitus prevalence after additionally adjusting main models for physical activity. Table S4: Associations of categorical NDVI and SAVI levels with fasting glucose levels and diabetes mellitus prevalence.
Author Contributions
Conceptualization, S.F., G.D., and Z.Z. (Zhoubin Zhang); Data curation, Z.X.; Formal analysis, S.F. and G.D.; Funding acquisition, S.F. and Z.Z. (Zhoubin Zhang); Investigation, Z.X., J.Y., Z.Z. (Ziyan Zhou), and Y.W.; Methodology, B.Y. and G.D.; Project administration, Z.Y. and Z.Z. (Zhoubin Zhang); Resources, Z.Y., B.Y., and G.D.; Validation, B.Y.; Writing—original draft, S.F.; Writing—review & editing, Z.Z. (Zhoubin Zhang).
Funding
This research was funded by the National Natural Science Foundation of China (81803196) and the Science and Technology Project of Guangdong Province (2014A020209006).
Conflicts of Interest
The authors declare no conflict of interest.
References
- 1.World Health Organization Global Report on Diabetes, Geneva, Switzerland. [(accessed on 24 November 2019)];2016 Available online: http://apps.who.int/iris/bitstream/10665/204871/1/9789241565257_eng.pdf.
- 2.Xu Y., Wang L., He J., Bi Y., Li M., Wang T., Wang L., Jiang Y., Dai M., Lu J., et al. 2010 China Noncommunicable Disease Surveillance Group. Prevalence and control of diabetes in Chinese adults. JAMA. 2013;310:948–959. doi: 10.1001/jama.2013.168118. [DOI] [PubMed] [Google Scholar]
- 3.Franks P.W., Pearson E., Florez J.C. Gene-environment and gene-treatment interactions in type 2 diabetes: Progress, pitfalls, and prospects. Diabetes Care. 2013;36:1413–1421. doi: 10.2337/dc12-2211. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Merino J., CHARGE Consortium Nutrition Working Group Quality of dietary fat and genetic risk of type 2 diabetes: Individual participant data meta-analysis. BMJ. 2019;366:I4292. doi: 10.1136/bmj.l4292. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Eze I.C., Hemkens L.G., Bucher H.C., Hoffmann B., Schindler C., Künzli N., Schikowski T., Probst-Hensch N.M. Association between ambient air pollution and diabetes mellitus in Europe and North America: Systematic review and meta-analysis. Environ. Health Perspect. 2015;123:381–389. doi: 10.1289/ehp.1307823. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Zhang Y., Pan X.F., Chen J., Xia L., Cao A., Zhang Y., Wang J., Li H., Yang K., Guo K., et al. Combined lifestyle factors and risk of incident type 2 diabetes and prognosis among individuals with type 2 diabetes: A systematic review and meta-analysis of prospective cohort studies. Diabetologia. 2019;63:21–33. doi: 10.1007/s00125-019-04985-9. [DOI] [PubMed] [Google Scholar]
- 7.Markevych I., Schoierer J., Hartig T., Chudnovsky A., Hystad P., Dzhambov A.M., de Vries S., Triguero-Mas M., Brauer M., Nieuwenhuijsen M.J., et al. Exploring pathways linking greenspace to health: Theoretical and methodological guidance. Environ. Res. 2017;158:301–317. doi: 10.1016/j.envres.2017.06.028. [DOI] [PubMed] [Google Scholar]
- 8.Ihlebæk C., Aamodt G., Aradi R., Claussen B., Thorén K.H. Association between urban green space and self-reported lifestyle-related disorders in Oslo, Norway. Scand. J. Public Health. 2018;46:589–596. doi: 10.1177/1403494817730998. [DOI] [PubMed] [Google Scholar]
- 9.Clark C., Sbihi H., Tamburic L., Brauer M., Frank L.D., Davies H.W. Association of Long-Term Exposure to Transportation Noise and Traffic-Related Air Pollution with the Incidence of Diabetes: A Prospective Cohort Study. Environ. Health Perspect. 2017;125:087025. doi: 10.1289/EHP1279. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Yang B.Y., Markevych I., Heinrich J., Bowatte G., Bloom M.S., Guo Y., Dharmage S.C., Jalaludin B., Knibbs L.D., Morawska L., et al. Associations of greenness with diabetes mellitus and glucose-homeostasis markers: The 33 Communities Chinese Health Study. Int. J. Hyg. Environ. Health. 2019;222:283–290. doi: 10.1016/j.ijheh.2018.12.001. [DOI] [PubMed] [Google Scholar]
- 11.Lee J.J., Hwang S.J., Mutalik K., Corey D., Joyce R., Block J.P., Fox C.S., Powell-Wiley T.M. Association of built environment characteristics with adiposity and glycaemic measures. Obes. Sci. Pract. 2017;3:333–341. doi: 10.1002/osp4.115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Thiering E., Markevych I., Brüske I., Fuertes E., Kratzsch J., Sugiri D., Hoffmann B., von Berg A., Bauer C.P., Koletzko S., et al. Associations of Residential Long-Term Air Pollution Exposures and Satellite-Derived Greenness with Insulin Resistance in German Adolescents. Environ. Health Perspect. 2016;124:1291–1298. doi: 10.1289/ehp.1509967. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Astell-Burt T., Feng X., Kolt G.S. Is neighborhood green space associated with a lower risk of type 2 diabetes? Evidence from 267,072 Australians. Diabetes Care. 2014;37:197–201. doi: 10.2337/dc13-1325. [DOI] [PubMed] [Google Scholar]
- 14.Brown S.C., Lombard J., Wang K., Byrne M.M., Toro M., Plater-Zyberk E., Feaster D.J., Kardys J., Nardi M.I., Perez-Gomez G., et al. Neighborhood Greenness and Chronic Health Conditions in Medicare Beneficiaries. Am. J. Prev. Med. 2016;51:78–89. doi: 10.1016/j.amepre.2016.02.008. [DOI] [PubMed] [Google Scholar]
- 15.Nguyen Q.C., Sajjadi M., McCullough M., Pham M., Nguyen T.T., Yu W., Meng H.W., Wen M., Li F., Smith K.R., et al. Neighbourhood looking glass: 360° automated characterisation of the built environment for neighbourhood effects research. J. Epidemiol. Community Health. 2018;72:260–266. doi: 10.1136/jech-2017-209456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Dalton A.M., Jones A.P., Sharp S.J., Cooper A.J., Griffin S., Wareham N.J. Residential neighbourhood greenspace is associated with reduced risk of incident diabetes in older people: A prospective cohort study. BMC Public Health. 2016;16:1171. doi: 10.1186/s12889-016-3833-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Maas J., Verheij R.A., de Vries S., Spreeuwenberg P., Schellevis F.G., Groenewegen P.P. Morbidity is related to a green living environment. J. Epidemiol. Community Health. 2009;63:967–973. doi: 10.1136/jech.2008.079038. [DOI] [PubMed] [Google Scholar]
- 18.Bodicoat D.H., O’Donovan G., Dalton A.M., Gray L.J., Yates T., Edwardson C., Hill S., Webb D.R., Khunti K., Davies M.J., et al. The association between neighbourhood greenspace and type 2 diabetes in a large cross-sectional study. BMJ Open. 2014;4:e006076. doi: 10.1136/bmjopen-2014-006076. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Dadvand P., Poursafa P., Heshmat R., Motlagh M.E., Qorbani M., Basagaña X., Kelishadi R. Use of green spaces and blood glucose in children; a population-based CASPIAN-V study. Environ. Pollut. 2018;243:1134–1140. doi: 10.1016/j.envpol.2018.09.094. [DOI] [PubMed] [Google Scholar]
- 20.Ngom R., Gosselin P., Blais C., Rochette L. Type and Proximity of Green Spaces Are Important for Preventing Cardiovascular Morbidity and Diabetes—A Cross-Sectional Study for Quebec, Canada. Int. J. Environ. Res. Public Health. 2016;13:423. doi: 10.3390/ijerph13040423. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Wang L., Gao P., Zhang M., Huang Z., Zhang D., Deng Q., Li Y., Zhao Z., Qin X., Jin D., et al. Prevalence and ethnic pattern of diabetes and prediabetes in China in 2013. JAMA. 2017;317:2515–2523. doi: 10.1001/jama.2017.7596. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Liu Y.L., Luo K.L., Lin X.X., Gao X., Ni R.X., Wang S.B., Tian X.L. Regional distribution of longevity population and chemical characteristics of natural water in Xinjiang, China. Sci. Total Environ. 2014;473:54–62. doi: 10.1016/j.scitotenv.2013.11.134. [DOI] [PubMed] [Google Scholar]
- 23.Hartley D. Rural health disparities, population health, and rural culture. Am. J. Public. Health. 2004;94:1675–1678. doi: 10.2105/AJPH.94.10.1675. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Zhang Z., Fan S., Xue Z., Yuan J., Zhou Z., Wang T., Liu J., Bawudun A., Nurmamat N., Wang Y., et al. Evaluation of the appropriate predictive contributor and diagnostic threshold for the cardio-metabolic syndrome in Chinese Uyghur adults. BMC Public Health. 2019;19:613. doi: 10.1186/s12889-019-6803-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.American Diabetes Association 2. Classification and diagnosis of diabetes: Standards of Medical Care in Diabetes-2018. Diabetes Care. 2018;41(Suppl. 1):S13–S27. doi: 10.2337/dc18-S002. [DOI] [PubMed] [Google Scholar]
- 26.Jager K.J., Zoccali C., MacLeod A., Dekker F.W. Confounding: What it is and how to deal with it. Kidney Int. 2008;73:256–260. doi: 10.1038/sj.ki.5002650. [DOI] [PubMed] [Google Scholar]
- 27.Greenland S., Pearl J., Robins J.M. Causal diagrams for epidemiologic research. Epidemiology. 1999;10:37–48. doi: 10.1097/00001648-199901000-00008. [DOI] [PubMed] [Google Scholar]
- 28.Tucker C.J. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 1979;8:127–150. doi: 10.1016/0034-4257(79)90013-0. [DOI] [Google Scholar]
- 29.Huete A. A soil-adjusted vegetation index (SAVI) Remote Sens. Environ. 1988;25:295–309. doi: 10.1016/0034-4257(88)90106-X. [DOI] [Google Scholar]
- 30.Yang B.Y., Markevych I., Bloom M.S., Heinrich J., Guo Y., Morawska L., Dharmage S.C., Knibbs L.D., Jalaludin B., Jalava P., et al. Community greenness, blood pressure, and hypertension in urban dwellers: The 33 Communities Chinese Health Study. Environ. Int. 2019;126:727–734. doi: 10.1016/j.envint.2019.02.068. [DOI] [PubMed] [Google Scholar]
- 31.Yang B.Y., Qian Z.M., Li S., Chen G., Bloom M.S., Elliott M., Syberg K.W., Heinrich J., Markevych I., Wang S.Q., et al. Ambient air pollution in relation to diabetes and glucose-homoeostasis markers in China: A cross-sectional study with findings from the 33 Communities Chinese Health Study. Lancet Planet Health. 2018;2:e64–e73. doi: 10.1016/S2542-5196(18)30001-9. [DOI] [PubMed] [Google Scholar]
- 32.Liu F., Chen G., Huo W., Wang C., Liu S., Li N., Mao S., Hou Y., Lu Y., Xiang H. Associations between long-term exposure to ambient air pollution and risk of type 2 diabetes mellitus: A systematic review and meta-analysis. Environ. Pollut. 2019;252:1235–1245. doi: 10.1016/j.envpol.2019.06.033. [DOI] [PubMed] [Google Scholar]
- 33.Sarkar C. Residential greenness and adiposity: Findings from the UK Biobank. Environ. Int. 2017;106:1–10. doi: 10.1016/j.envint.2017.05.016. [DOI] [PubMed] [Google Scholar]
- 34.Persson Å., Pyko A., Lind T., Bellander T., Östenson C.G., Pershagen G., Eriksson C., Lõhmus M. Urban residential greenness and adiposity: A cohort study in Stockholm County. Environ. Int. 2018;121:832–841. doi: 10.1016/j.envint.2018.10.009. [DOI] [PubMed] [Google Scholar]
- 35.Crump C., Sundquist J., Winkleby M.A., Sundquist K. Stress resilience and subsequent risk of type 2 diabetes in 1.5 million young men. Diabetologia. 2016;59:728–733. doi: 10.1007/s00125-015-3846-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Van Aart C.J.C., Michels N., Sioen I., De Decker A., Bijnens E.M., Janssen B.G., De Henauw S., Nawrot T.S. Residential landscape as a predictor of psychosocial stress in the life course from childhood to adolescence. Environ. Int. 2018;120:456–463. doi: 10.1016/j.envint.2018.08.028. [DOI] [PubMed] [Google Scholar]
- 37.Rook G.A. Regulation of the immune system by biodiversity from the natural environment: An ecosystem service essential to health. Proc. Natl. Acad. Sci. USA. 2013;110:18360–18367. doi: 10.1073/pnas.1313731110. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.