Associations of household solid fuel for heating and cooking with hypertension in Chinese adults.

OBJECTIVE
The association between indoor air pollution resulting from household solid fuel use for heating and cooking with hypertension or blood pressure (BP) remains less clear. This study aims to rectify these knowledge gaps in a large Chinese population.


METHODS
During 2005-2009, 44 007 individuals aged 35-70 years with complete information on household solid fuel use for cooking and heating were recruited from 279 urban and rural communities of 12 centers. Solid fuel referred to charcoal, coal, wood, agriculture crop, animal dung or shrub. Annual concentration of ambient atmospheric particulate matter that have a diameter of less than 2.5 μm for all communities was collected. Generalized linear mixed models using community as the random effect were performed to estimate the association with hypertension prevalence or BP after considering ambient atmospheric particulate matter that have a diameter of less than 2.5 μm and a comprehensive set of potential confounding factors at the individual and household level.


RESULTS
A total of 47.6 and 61.2% of participants used household solid fuel for heating and cooking, respectively. Solid fuel use for heating was not associated with an increase in hypertension prevalence (adjusted odds ratio = 1.08, 95% confident interval: 0.98, 1.20) or elevated SBP (0.62 mmHg, 95% confident interval: -0.24, 1.48). No association was found between solid fuel for cooking and hypertension or BP, and no additional risk was observed among participants who had both exposures to solid fuel for heating and cooking compared with those used for heating only.


CONCLUSION
The current large Chinese study revealed a statistically insignificant increase in the association between solid fuel use for heating and hypertension prevalence or BP. As this cross-sectional study has its inherent limitation on causality, findings from this study would have to be confirmed by prospective cohort studies.


INTRODUCTION
I ndoor air pollution resulting from burning solid fuel such as coal, wood, crop residues for heating and cooking has been a major public health problem in China and other developing countries for decades [1][2][3][4]. Burning fuels for heating in winter or rainy season is especially common in many northern rural areas of China [5,6], which generates large amounts of particulate matter [e.g. atmospheric particulate matter that have a diameter of less than 2.5 mm (PM2. 5), PM10], carbon monoxide nitrogen dioxide, sulfur dioxide, polycyclic aromatic hydrocarbons and/or other volatile organic compounds [7,8]. Many of these chemical substances are known to increase oxidative stress and systematic inflammation, and could potentially increase blood pressure (BP) and promote vascular atherosclerosis [8,9].
Association between household solid fuel use for cooking and hypertension has been investigated by several epidemiological studies with mixed findings [10][11][12][13][14][15][16][17]. A study examining 137 809 individuals from 10 countries in the Prospective Urban and Rural Epidemiology (PURE) study observed no association between cooking with solid fuels and hypertension or BP [10]. However, the PURE study did not examine solid fuel use for heating since many of the included countries did not have high prevalence of heating. However, in China, prevalence of using solid fuel for heating and cooking is high particularly in north China [5,6], and also it has been known that burning solid fuel for heating may generate more PM2.5 than cooking [18]. Until now, only two cross-sectional studies conducted in China have examined the association between using household solid fuels and hypertension prevalence and reported a positive association [16,17], but both of them did not distinguish the effect of using solid fuel for cooking from that for heating. More recent evidence has shown that longterm exposure to ambient PM2.5 may harm blood vessels and increase BP via increasing inflammatory factors and oxidative stress [19,20], but the specific effect of using household solid fuel for heating and the joint effect with solid fuel for both cooking and heating remains largely not known.
The current study is the first to examine the association of household solid fuel use for both heating and cooking with hypertension and BP among a large community-based population in China, after considering the influence of ambient PM2.5 exposure and a comprehensive set of potential confounding factors at the individual and household level.

Study design and participants
The PURE study is a large-scale epidemiological study that recruited 153 996 individuals aged 35-70 years from 600 communities in 17 countries around the world [21,22]. This report included the baseline survey of the PURE study in China (PURE-China). Three-level cluster sampling (province, community and household) was used to enroll potential participants between 2005 and 2009 [23]. For the PURE-China study, a total of 47 863 participants aged 35-70 years were recruited from 279 communities in 12 centers (provinces), and the selection of communities was based on the feasibility for a long-term follow-up. A total of 3563 participants were excluded due to incomplete information of age, sex or primary cooking and/or heating source, education, smoking and alcohol status, food intake and household wealth index. We further excluded 223 individuals who used kerosene as the main heating source because its emission characteristic differs from that of the solid fuels [4]. Therefore, a total of 44 007 participants were retained in the current analyses.

Data collection
Trained research staff interviewed each participant according to standardized individual and household questionnaires to collect information on socio-demographic data, lifestyle behaviors (e.g. tobacco smoking, alcohol drinking, physical activity, and dietary habits), household fuels uses and medical history [22]. Physical activity was collected using the International Physical Activity Questionnaire and then evaluated in metabolic equivalents [24]. Daily energy intake was assessed using a validated Food Frequency Questionnaire [25]. Smoking and alcohol use were defined if an individual who had ever or currently regularly used any tobacco or alcohol products, otherwise, the participant was referred to being a never smoker or alcohol drinker. BMI was calculated by body weight (kilogram) divided by the square of the body height (meter) [26]. We categorized BMI according to the guideline for the prevention and treatment of hypertension in China and defined participants as normal (BMI < 24), overweight (28 > BMI ! 24) and obese (BMI ! 28) [27]. Household wealth index was developed based on household possessions from the PURE questionnaire and classified into three categories using the tertile value [28].

Definitions of outcomes
BP was measured twice separately on the right arm by trained personnel using a digital sphygmomanometer (Omron HEM-757; Omron Healthcare Co. Ltd., Scarborough, Ontario, Canada) with a cuff size of 14 Â 48 cm supplied to all sites, and the device was validated for accuracy [29,30]. The mean value of two separate measurements was calculated for further analysis. Before BP was measured, each participant had at least 5 min for resting and stayed 30 min before the last smoking, exercising, eating or climbing stairs. Each measurement was recorded for both SBP and DBP simultaneously. Hypertension was defined either using a self-reported physician diagnosed hypertension, receiving BP control drug within 2 weeks or the mean value of DBP at least 90 mmHg or SBP at least 140 mmHg measured at the baseline.

Measurement of household fuel uses for heating and cooking
Information on household fuel uses for heating and cooking was collected using the same protocol published elsewhere [31]. In brief, the sources of household fuel use for cooking and heating during the cold or rainy season were collected according to a standard questionnaire at a medical institute and the acquired information was further validated by a trained site monitor staff who was responsible for quality control. As information on household fuel use was recorded at the household level, we assumed all participants from the same household adopted the same source of fuel for cooking and heating. We classified participants into the users of solid fuels for heating (coal open fire, wood open fire or agriculture/crop) and clean fuel (electricity, gas or central heating). Agriculture crop mainly referred to crop straws or residuals. As the socioeconomic background of people burning other types of fuel for heating or those who had never used heating differed from those using solid fuel or clean fuel, they were classified into separate groups. We defined the primary cooking fuel as solid fuel (charcoal, coal, wood, agriculture crop, animal dung and shrub) and clean fuel (electricity, gas or gobar gas) [31]. As some participants used mixed fuel for heating and cooking, we grouped them into clean or solid fuel groups based on their primary fuel use and then conducted sensitivity analysis by their secondary fuel use.
Measurement of ambient atmospheric particulate matter that have a diameter of less than 2.5 mm air pollution Detailed methodology of PM2.5 exposure assessment has been published elsewhere [32]. Briefly, annual PM2.5 concentrations were estimates from satellite and fixed monitoring data using geographically weighted regression method at 1 km Â 1 km resolution. Multiple satellite products were

Statistical analysis
First, we described basic characteristics of our study population and outcome measures. We presented the mean and SD for normally distributed continuous variables, the medium and interval of quantile range [interquartile range (IQR)] for skewed continuous data, and the frequency and proportion for categorical data respectively. Second, to assess the associations of household solid heating or cooking fuel with hypertension prevalence or BP, we used generalized linear mixed model, with community as a random intercept variable (to account for the PURE-China study design and clustering of individuals within communities) to calculate adjusted odds ratio (AOR) and 95% confidence intervals (CIs). Fully adjusted models include: age, sex (male/female), location (rural or urban), smoking status [never and ever (former or current)], drinking status [never/ever (current and former)], education level (primary or below/secondary/tertiary or above), household wealth index (by tertile), BMI status (normal/overweight/obesity), physical activity (low/middle/high), ambient PM2.5, energy intake and study center [20,33,34]. We included study center to control for the unmeasured differences between centers, which may also be correlated with solid fuel use for heating and cooking.
Subgroup analyses for the association of household solid heating or cooking fuel and their specific types with hypertension prevalence or BP were performed according to socio-demographic characteristics, lifestyle factors and ambient PM2.5 levels. Associations with hypertension and BP were also presented by study center using forest plot of meta-analysis approach in STATA (StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP) [35]. Sensitivity analysis was also conducted according to participant's secondary choice of household solid fuel use and status of receiving antihypertensive drug treatment. We used SAS software, version 9.4 (SAS Institute Inc., Cary, North Carolina, USA) to conduct all statistical tests using a two-sided significant level of 0.05.

RESULTS
Of the 44 007 participants included in this analysis, 18 427 (41.9%) were males and 25 580 were females (58.1%), with a mean age of 51.1 AE 9.6 years. 20 935 participants (47.6%) used household solid fuel for heating, 17 521 (39.8%) used clean fuel for heating and 5551 (12.6%) reported never using heating. A total of 26 946 (61.2%) participants used household solid fuel for cooking and 17 061 (38.8%) adopted clean fuels for cooking. 12 977 (33.7%) of individuals used household solid fuel for both heating and cooking and 16 136 (42.0%) of individuals using clean fuel for both heating and cooking. The annual median concentration of ambient PM2.5 was 45.8 mg/m 3 (IQR: 41.7 mg/m 3 ). Table 1 summarizes study participants' characteristics according to the status of household solid fuel uses for heating and cooking. Compared with those using clean heating fuel for heating, participants adopting solid fuel for heating were younger, more likely to be exposed to higher annual concentration of ambient PM2.5, had lower education and lower household wealth index, being ever smokers or ever alcohol drinkers, and taking hypertensive medicine. Regarding status of using solid fuel for cooking, a similar distribution of basic characteristics to that of using heating solid fuel for heating was observed except for ambient PM2.5. A higher ambient PM2.5 concentration was observed in the cooking group with clean fuel than that with solid fuel. Participants who did not use heating, tended to have slightly higher education attainment and household wealth index than those using solid fuels for heating, but they were exposed to a relatively lower level of ambient PM2.5 than other groups.
Associations between solid fuel use for heating, cooking and both heating and cooking are summarized in Table 2. Participants used solid fuel for heating via coal open fire (42.7%), wood open fire (4.2%) or burning straw of agriculture crop (0.6%). In the base model (Table S1, http:// links.lww.com/HJH/B486), positive associations were observed between household heating with solid fuel and its subtypes with hypertension, but the associations were only statistically significant among those burning agriculture crop for heating. Associations remained almost unchanged in the fully adjusted models for the association between household solid fuel for heating and hypertension prevalence (AOR ¼ 1.08, 95% CI: 0.98, 1.20), and the association with burning agriculture crop (AOR ¼ 1.49, 95% CI: 1.07, 2.06) and wood open fire (AOR ¼ 1.15, 95% CI: 0.96, 1.38) was relatively higher than that for coal open fire (AOR ¼ 1.06, 95% CI: 0.95, 1.18). For solid fuel use for cooking we observed no associations with hypertension.
Using solid fuel for household heating was positively but not significantly associated with SBP (elevated BP ¼ 1.41 mmHg, 95% CI ¼ À1.18, 4.01) and all subtypes in the fully adjusted model in Table 2. The associations between using solid fuel for heating and DBP were relatively weaker than SBP. Our results consistently showed that neither hypertension nor SBP or DBP was associated with household solid fuel use for cooking or never using heating, and the detailed results for base model could be seen in Table S1, http://links.lww.com/HJH/B486 which is consistent with the fully adjusted model.
Combined effects of using household solid fuel for both heating and cooking on hypertension prevalence or BP were inconsistent (Table 2). Compared with those using clean fuel for both cooking and heating, a positive association with hypertension or SBP or DBP was observed among participants using solid fuel for heating only, and there was no association with solid fuel use for cooking. In addition, individuals who used solid fuel for both cooking and heating did not have additional risk for hypertension or SBP or DBP, compared with those using solid fuel for heating only.
Associations between solid fuels for heating and hypertension prevalence or BP were further explored using metaanalysis according to the geographic location (i.e. center). As shown in Fig. 1, the associations between solid fuel for heating and hypertension prevalence was positive for all centers except for Qinghai and Inner-Mongolia, with a combined AOR of 1.07 (95% CI: 0.94, 1.20); however, the association between solid fuels for cooking and BP Clean household fuels for cooking include electricity, gas and gobar gas; Solid fuels for cooking include charcoal, coal, wood, agriculture, animal dung and shru.
among centers was inconsistent with a combined AOR of 0.98 (95% CI: 0.89, 1.06) (Fig. 2). Considering patients receiving antihypertensive drug treatment may have already changed their lifestyle behavior for heating and cooking, we conducted sensitivity analyses by removing those receiving antihypertensive drug treatment and the association between solid fuel for heating and hypertension prevalence (AOR ¼ 1.10, 95% CI: 0.99, 1.23), SBP (AOR ¼ 1.10, 95% CI: 0.21, 1.99) or DBP (AOR ¼ 0.30, 95% CI: À0.28, 0.88) remained consistent and robustness. Further sensitivity analysis was performed by removing the fixed effect for study center, as well as the random intercept for community, as these geographical variables may over adjust model results since cooking and heating are distributed regionally in China. These models showed a similar pattern to our main models, but indicated a positive association with solid fuel for heating for hypertension prevalence (AOR ¼ 1.23, 95% CI: 1.13, 1.33 vs. AOR ¼ 1.12, 95% CI: 1.01, 1.25), SBP and DBP (Tables S2, http://links.lww.com/HJH/B486 and S3, http://links.lww.com/HJH/ B486). Consistently, sensitivity analyses were repeated for solid fuel use for cooking and no associations were observed. Compared with using clean fuel for cooking and heating, combined use of solid heating fuel and solid cooking fuel did not introduce additional risk for hypertension or SBP or DBP.

DISCUSSION
This is the first population-based multicenter study to investigate the separate and combined associations of household solid fuel use for heating and cooking with hypertension and BP, after controlling for ambient PM2.5 air pollution and comprehensive individual and household factors. Association between solid fuel and hypertension for cooking fuel was neutral, and that for heating fuel was statistically insignificant increase. The combined exposure to solid fuel use for both heating and cooking was not associated with additional risk to hypertension and elevated BP.
Our finding of no consistent association between cooking and heating with solid fuels and hypertension and BP is different from previous studies conducted in China that identified a potentially increased risk with solid fuel use [9,16,17,20,36]. For example, one of them involved only 209 adults in China and reported a significantly increased SBP of 3.80% per IQR evaluation of indoor and outdoor PM2.5 level with a 2 days' lag; however, this small Chinese study did not examine the potential long-term effect on BP and the association with specific household heating source was not addressed [9]. Another hospital-based study analyzed 14 068 participants in Shanghai and found that using household solid fuel was positively associated with hypertension prevalence (OR ¼ 1.70, 95% CI: 1.40-2.07), but the analysis did not specify the effect of cooking from heating [16]. No separate effect of solid heating fuel and cooking fuel was reported in the Shanghai's study. In our study, we were able to separate the association of hypertension and BP with heating solid fuel from that with cooking solid fuel, and specially demonstrated the pooled effect of combined exposure to both heating and cooking fuel that had never been reported in the previous studies. In addition to Chinese studies, two multicountry studies which considered both indoor and outdoor air pollution found the similar results that the use of household solid fuel may haves insignificant association with hypertension prevalence [10,11]. One of them including 77 605 premenopausal women among 10 countries found that a positive pooled effect, but the statistically insignificant association between the use of solid fuel and hypertension prevalence (OR ¼ 1.07, 95% CI ¼ 0.99-1.16). They also found that the association between use of solid fuel and increment of BP only existed in SBP rather than DBP. Another study including 137 809 adults in 21 countries found a slightly negative association between household solid fuel use and hypertension prevalence (OR ¼ 0.93, 95% CI ¼ 0.88, 0.99) and BP. However, both of them only focused on cooking and the specific association with household solid fuel for heating has never been reported.
We also observed an increased association between hypertension prevalence and burning agriculture crop type of solid fuel for heating. This was more prominent for agriculture crop fuel use (hypertension AOR ¼ 1.49, 95% CI: 1.07, 2.06), which was the only fuel type showing a small and significant association with hypertension. The combustion of agriculture crop is ranked as one of the lowest energy ladder fuels, which emits more health-damaging pollutants than coal and wood [37]. In line with the finding reported in the global PURE study [10], our study with a large and diverse Chinese population also supports no association between solid fuel use for cooking and hypertension prevalence.
Results from sensitivity analysis by removing ambient PM2.5 from the multivariate models did not show an obvious change in the association between the use of heating solid fuel and hypertension prevalence or BP level (Table  [16,17]; however, these results may not reflect all of the random intercept caused by center. Including center in our models controlled for regional differences but restricted our comparisons with individuals living with the same regions of China. About 63.7% of our participants living in the north used solid fuel as their primary heating fuel (Table S5, http://links.lww.com/HJH/B486), which was remarkably higher than that in the south (6.8%). However, a similar pattern of solid fuel use for cooking was found between the north and south area (Table S6, http://links.lww.com/HJH/B486), which explains why the associations with cooking solid fuel remained stable regardless of the inclusion of center in the model. This makes us hard to isolate the true impact of solid fuel use for heating from other social contextual factors.
The strength of our study includes a standardized study design implemented in 12 regions in China by capturing diverse populations and household cooking and heating environments. We were also able to examine a large sample size from both the urban and rural communities, control for ambient PM2.5 concentration, and used robust modeling methods to control for potential unmeasured contextual factors. However, there are a few major limitations of this study. First, we had information only on primary heating fuels, and no information was available on fuel stacking. We also do not have any information about previous use of heating fuel and how long individuals used different fuels throughout their lifetime. We used geographic location north and south as a surrogate to represent the length of exposure to heating solid fuel, as it is expected that inhabitants in north area of China have longer duration of exposure to heating using different types of fuel than those living in the south area. As shown in Table 3, the association between solid fuel use for heating and hypertension or BP was consistently slightly higher in the north than those in the south, though there was a lack of statistical significance, this difference may partly reflect a positive gradient with longer duration of exposure of heating solid fuel in the north area. On the other hand, no obvious difference in daily hours for cooking between south and north may explain no variation for the association between solid fuel use for cooking and hypertension or BP being observed between the south and the north ( Table 4). As shown in Table S5, http://links.lww.com/HJH/B486, a majority of people (59.3%) in the northern regions of China used coal as their primary heating source and 87.8% of them lived in a rural community. Another study reported that the percentage of household coal use was nearly 40% in Chinese rural area in 1990 and it was unchanged until 2007 [39], which is in line with our survey results from 2005 to 2009 [39]. This indicates that most of the individuals living in northern communities kept their primary heating source stable during the 1980s to 2000s. In the southern part of China, most of our participants used clean fuels for heating (49.0%) or no heating (44.2%), and this is expected given the relatively warm temperature in the southern part of China [40]. Another limitation is that we only estimated the annual concentration of PM2.5 for study communities and did not have information on short-term exposures at the time of BP measurements. We also did not have information of the chemical components of PM2.5 of information for other air pollutants. Cao et al. [41] collected nationwide PM2.5 sample in 14 cities of China and showed that the main component of PM2.5 varied substantially. It is likely that the composition of PM2.5 varied between our study communities and future research is needed to include this information in the analyses with health outcomes. This limitation is suggested to be addressed in future follow-up study (i.e. PURE-AIR). Finally, we used the primary fuel type as an indicator of household and personal PM2.5 exposures, while adopting standard practice, it may incorporate substantial measurement error. To address this issue, we have already chosen several communities to collect indoor and outdoor air pollution samples to refine exposure estimates.
In conclusion, our study observed a statistically insignificant increase in the association between the use of solid fuel for heating and prevalence of hypertension or BP in this large and diverse Chinese population. A small effect of household heating fuel use with hypertension and elevated BP cannot be excluded but needs to be explored in future larger scale studies. Further research is needed using repeated BP measures as well as PM2.5 measures that address how heating versus cooking influences personal PM2.5 exposures.