Association of ambient air pollution with cardiovascular disease risks in people with type 2 diabetes: a Bayesian spatial survival analysis

Abstract

Background: Evidence is limited on excess risks of cardiovascular diseases (CVDs) associated with ambient air pollution in diabetic populations. Survival analyses without considering the spatial structure and possible spatial correlations in health and environmental data may affect the precision of estimation of adverse environmental pollution effects. We assessed the association between air pollution and CVDs in type 2 diabetes through a Bayesian spatial survival approach.

Methods: Taiwan's national-level health claims and air pollution databases were utilized. Fine individual-level latitude and longitude were used to determine pollution exposure. The exponential spatial correlation between air pollution and CVDs was analyzed in our Bayesian model compared to traditional Weibull and Cox models.

Results: There were 2072 diabetic patients included in analyses. PM_2.5 and SO₂ were significant CVD risk factors in our Bayesian model, but such associations were attenuated or underestimated in traditional models; adjusted hazard ratio (HR) and 95% credible interval (CrI) or confidence interval (CI) of CVDs for a 1 μg/m³ increase in the monthly PM_2.5 concentration for our model, the Weibull and Cox models was 1.040 (1.004-1.073), 0.994 (0.984-1.004), and 0.994 (0.984-1.004), respectively. With a 1 ppb increase in the monthly SO₂ concentration, adjusted HR (95% CrI or CI) was 1.886 (1.642-2.113), 1.092 (1.022-1.168), and 1.091 (1.021-1.166) for these models, respectively.

Conclusions: Against traditional non-spatial analyses, our Bayesian spatial survival model enhances the assessment precision for environmental research with spatial survival data to reveal significant adverse cardiovascular effects of air pollution among vulnerable diabetic patients.

Keywords: Bayesian approach; Cardiovascular disease; Spatial correlation; Survival; Time-to-event; Type 2 diabetes.

Fig. 1

Determination of air pollution at individual level and spatial correlations and event rate of composite cardiovascular disease events and 76 air quality monitoring station (dots) across 20 geographic regions in Taiwan. a shows the two main problems inherent to existing studies on the environmental pollution effects on health. First, personal location is typically defined at the geographic area (community) level (e.g., a person is known to live in Changhua but their exact location in Changhua is unclear). As a result, the conventional approach for the assessment of air pollution is to assign all residents within an area the same value of air pollution, regardless of the individual’s relative location with respect to the pollution source. For example, consider the 13 people living in Changhua County shown in Fig. 1a. The average concentration for the three air monitoring stations within the area (i.e., [46.49+ 49.67 + 53.88]/3 = 50.01 μg/m³ for the monthly concentration of PM_2.5 in Dec. 2009) is generally used for people living in a given area. Second, due to the coarse resolution used for individual locations, the distance between two individuals or their proxies (i.e., their assigned air quality monitoring stations) is not ascertained and thus the spatial correlations in the pattern of air pollution and adverse pollution effects on health are not modeled in the analyses. b shows the spatial variations in the event rate of cardiovascular diseases across Taiwan

Fig. 2

Collinearity between ambient air pollutants tested using Pearson correlation matrix. This figure shows that several ambient air pollutants are highly correlated to each other (e.g., there are correlations between PM₁₀ and PM_2.5 [i.e., 0.93] and between NO_x and NO₂ [i.e., 0.94])

Fig. 3

Estimated spatial correlations and predicted risk-exceedance probabilities (P [exp(Y) > 1]) of cardiovascular diseases in Taiwan. a Estimated spatial correlations of exposure to air pollution and consequent pollution effects on cardiovascular diseases. Based on the estimated spatial parameters σ = 1.28 and ϕ = 0.44, when the distance between the proxies for two individuals (i.e., two air quality monitoring stations assigned to them) is more than 100 km, the corresponding relative spatial correlations with the hazard of CVDs between these two individuals reduce to almost zero; this means the spatial variability that cannot be explained by the covariates in the model is almost zero when two stations are located more than 100 km apart from each other. b Predicted risk-exceedance probabilities (P [exp(Y) > 1]) of cardiovascular diseases for individual geographic areas in Taiwan. This plot shows the posterior probability that the covariate-adjusted relative risk of CVDs is larger than 1 (P [exp(Y) > 1]), which maps the likelihood of excessive CVD risks as the “CVD risk-exceedance probability” to demonstrate the risk level of CVDs for areas in Taiwan. Specifically, for the areas with a color spot close to red that indicates the likelihood of having CVDs (i.e., covariate-adjusted relative risk > 1) is close to 100%, they could be classified as the area with a higher CVD risk. In contrast, for those with a color spot close to yellow that implies the likelihood of having CVDs > 1 is close to 0%, they could be considered as the area with a lower CVD risk

Fig. 4

Association between patients’ characteristics and the risk of cardiovascular diseases in the multivariable Bayesian survival spatial, Cox, and Weibull models. This figure shows the association between patients’ characteristics (i.e., demographics, drug treatments, diabetes status/severity, comorbidities, and air pollution exposure) and the risk of composite cardiovascular disease events, estimated by the multivariable Bayesian survival spatial, Cox, and Weibull models. And, in the Bayesian survival spatial model, the estimated prior parameters α, λ, σ, and φ are 0.932, 0.0001, 1.2829, and 0.440, respectively. For the Weibull model, α and λ are 0.8971 and 0.0007, respectively. Abbreviations: aDCSI, adapted Diabetes Complication Severity Index; CVD, cardiovascular disease; GLA, glucose-lowering agent; SU, sulfonylurea; DPP-4i, dipeptidyl-peptidase 4 inhibitor; TZD, thiazolidinedione; CrI, credible interval; CI, confidence interval