2.1 Study design and participants
Guangdong province, located in southern China, has a developed economy and a large population. The population in Guangdong province is about 126 million according to the seventh population census of China in 2020 [13]. The area has a subtropical monsoon climate with plenty of sunshine and precipitation. We collected data from surveys on the risk factors of noncommunicable diseases from Guangdong province in 2004, 2007, 2010, 2013 and 2015. These surveys conducted by Guangdong Provincial Center for Disease Control and Prevention, and in order to understand the prevalence of major chronic disease and their risk factors in Guangdong province, China.
Details of those surveys were described elsewhere [14–16]. Similar study designs and methods were used to all surveys implemented in 2004, 2007, 2010, 2013 and 2015. Stratified multistage cluster sampling with probability proportional to size was used to obtain research objects. First, 21 survey site including cities or counties in Guangdong province were randomly selected. Second, four districts or townships from each survey site were chosen; Third, three streets or villages constituted at least 50 households from each district or township were chosen; Fourth, one household from each street or village was randomly sampled; Finally, one participant aged ≥ 18 years was selected using the Kish grid method from each selected household. When the selected respondent did not agree to participate or there was no people aged ≥ 18 years in the selected household, the household would be replaced by another randomly selected household nearby.
All participants received a face-to-face questionnaire interview by investigators with professionally trained experience and on-site health examination.
Questionnaires were consisted of demographics, behavioral factors and disease history. General demographics include gender (male and female), age (< 60 and ≧ 60), marital status (never married, married, and others), education level. Behavior factors include smoking, drinking and physical activity. The history of related diseases includes hypertension, diabetes etc.
Health examination main included blood pressure measurement, height and weight measurement, blood sample test. Blood pressure measured in our surveys by physicians through standardized mercury sphygmomanometers or Omron sphygmomanometers. Before measurement, participants were asked to rest at least 5 minutes, then BP was taken on the right arm 2 or 3 times 1 minute apart in a seated position. The average of readings was used for analysis. According to 2018 European society Arterial Hypertension Guidelines [17, 18], we divided participants into three different blood pressure states. Participants who have been diagnosed with hypertension by physicians before the surveys were defined as known hypertensive patients. Newly detected hypertensive patients in the surveys were defined as SBP is greater than 140 mmHg and/or DBP was large than 90mmHg but not diagnosed as hypertension before the surveys. Normotensives were defined as participants whose SBP was less than 140 mmHg and DBP was less than 90mmHg without hypertension history.
Each participant was provided a detailed introduction and explanation of this study and signed the informed consent form. Ethics was approved by the Ethics Committee of Guangdong Provincial Center for Disease Control and Prevention (Ethical review code:2019025).
2.2 Meteorological data and air pollution data
Daily meteorological data during study periods were obtained from Guangdong meteorological center, including daily maximum temperature, daily minimum temperature, daily mean temperature, relative humidity, wind speed and precipitation. We connected daily meteorological data to hypertension data by climate station. If there is no climate station at survey sites, the nearest weather observation station will be selected. The air pollution data during 2013–2015 were collected from Guangdong Environmental Monitoring Center, including daily PM10, SO2, and O3. We linked air pollution data to hypertension data through each individual’s districts/counties. Similarly, if there is no air pollution station at survey site, the nearest observation station will be selected.
2.3 Statistical analysis
In this study, we employed DTR and TV0-n (n = 1–7) to assess the BP effects of temperature variability within 1 to 8 days. DTR was defined as the difference between the daily maximum temperature and minimum temperature within a day. TV0-n was calculated by the standard deviation (SD) of daily minimum temperature and daily maximum temperature within exposure days. For example, temperature variability within 2 days’ exposure (TV0–1) was calculated as follows: TV0-1 = SD (maximum temperaturelag0, minimum temperaturelag0, maximum temperaturelag1, minimum temperaturelag1). Therefore, the definition of temperature variability could explain both intra-day and inter-day temperature variability, as well as the delayed effects of temperature variation [19].
BP was described as mean ± sd, and other basic characteristics of the study sample were presented as counts or constituent ratios.
A nonlinear relationship between temperature and blood pressure was showed in previous study [20]. Therefore, the generalized additive model (GAM) was used in this study to explore the nonlinear relationship between temperature variability and BP. Since the distribution of BP in the population was approximately normal, Gaussian function was selected as the connection function in the model. Cubic spline function and three degrees of freedom (df) was chosen to fit model, and covariates included daily mean temperature, precipitation, relative humidity, wind speed, BMI, age, sex, physical activity, smoking, alcohol use, and resting time. The regression model was described as the following:
Yi = β0 + βtv0_n s (Xtv0_n, k = 3) + βtm s (Xmtemp, k = 3) +βn s(Xcovariate)+ εi
Yi refers to BP; β0 is the overall intercept, βtv0_n means coefficient of TV0-n. βtm means coefficient of daily mean temperature. βn corresponds to coefficient of covariates. s () refers to penalized cubic spline function, εi is the residual error.
To simple identification the relationships between temperature variability and BP, we further fitted linear model to obtain the linear effect of temperature on BP by replacing the cubic spline function of temperature variability in main model with a linear term.
In addition, we further analyzed the potential modifications of individual characteristics on the effects of temperature variability on BP. The following formula was used to test the significant difference between two-point estimates. Q1 and Q2 were estimated effects, SE1 and SE2 were estimated standard errors [21].
(Q1-Q2) ± 1.96\(\sqrt[2]{S{e}_{1}^{2}+S}{e}_{2}^{2}\)
In sensitivity analysis, biological information (LDL) and air pollutants (et. PM2.5, SO2, O3) were separately added to the model to test the robustness of our findings.
All the statistical analyses were performed in R software version 4.0.3 using “mgcv” packages. Two-sided statistical test was conducted, and effects of P < 0.05 were considered statistically significant.