The current study was based on the China Health and Retirement Longitudinal Study (CHARLS), aiming to collect high-quality data representative of middle-aged and elderly people in China [15]. All participants underwent an interview to collect data on sociodemographic characteristics, health status, and medical history. The CHARLS baseline survey was conducted in 2011, with regular follow-up visits every 2–3 years in 2013, 2015, and 2018, respectively. All participants provided written informed consent during the investigation. A detailed description is available on its website (http://charls.pku.edu.cn/en).

17,708 adults were recruited in 2011. We excluded participants who had been diagnosed of non-neoplastic digestive system diseases (NNDSD) in 2011 and before (n = 4717), participants with missing data on indoor fuel source (n = 1540), diagnosing NNDSD (n = 2789) and demographic characteristics (n = 778); then 7,884 participants were left for the baseline analysis of the use of cooking and heating fuel types and their combined effects on NNDSD. For the follow-up analysis, we additionally excluded 4922 participants with missing data on cooking fuel type, resulting in 2962 individuals for analysis (Figure 1). Since data on heating fuel in 2015 were not available, we analyzed the effect of switching fuel types on the risk of NNDSD only regarding cooking fuel data.

Physician-diagnosed NNDSD after 2011 was defined as the outcome of interest. Physician-diagnosed NNDSD was obtained from a self-reported questionnaire across follow-up surveys (“Have you been diagnosed with stomach or other non-neoplastic digestive system diseases by a doctor?”). If the subject’s answer to this question was “Yes”, he or she was defined as suffering from NNDSD. The survival time was defined as the time from the baseline date to the dates of the NNDSD diagnosis, loss to follow-up, or the end of follow-up (2018), whichever occurred first.

We dichotomized cooking fuels and heating fuels [clean fuels (natural gas, marsh gas, liquefied petroleum gas, solar, electric, or concentration heating) vs. solid fuels (coal, crop residues, wood-burning, or others)]. Notably, heating fuel information use was not available in 2015–2016. Therefore, we evaluated the effect of fuel type conversion on NNDSD only with the cooking fuel data in the follow-up analysis.

We divided the population into four groups in the follow-up analysis based on the type of fuel used at baseline and updated across follow-up visits: always solid fuel, always clean fuel, solid-to-clean fuel, and clean-to-solid fuel. The NNDSD emerged after fuel type conversion. For example, a subject who was diagnosed with NNDSD firstly in 2013, used solid energy in 2011 and clean energy in 2013, would be classified into the solid-to-clean fuel group; a subject who was a clean fuel user in 2011 while solid fuel user in 2015, diagnosed with NNDSD firstly in 2015, would be divided into clean-to-solid fuel group. Other groups were defined similarly.

Covariates adjusted in our analyses included sex, age (middle-aged adults, 45–65 years; old adults, >65 years), educational level (illiteracy or informal education; elementary school or above), marital status (married, unmarried), smoking status (smokers, non-smokers), residence region (city or town; village), self-reported economic level (relatively poor or poor, average, relatively high or high), hypertension comorbidity (yes, no), diabetes (yes, no), dyslipidemia (yes, no), cardiovascular disease (yes, no), stroke (yes, no), liver diseases (yes, no), lung diseases (yes, no), cancer or malignant tumor (yes, no), and kidney diseases (yes, no).

The educational level was obtained by question, “what’s the highest level of education you have attained now?” Then, the answers were classified into two groups (illiteracy or informal education, elementary school or above). Marital status was divided into married and unmarried. Non-smokers were defined as those who had never smoked, and smokers were defined as those who had smoked or were currently smoking. Systolic blood pressure (SBP) or diastolic blood pressure (DBP) over 140/90 mmHg was defined as hypertension. Diabetes was defined as [1]: had been diagnosed as diabetes by a clinical doctor [2], fasting blood glucose (FBG) >200 mg/dl (11.1 mmol/L). HDL-C, LDL-C, glucose, and triglyceride were obtained from blood examination. Cardiovascular disease, stroke, liver diseases, lung diseases, cancer or malignant tumors, and kidney diseases were defined as previously diagnosed with this disease by a clinical doctor. Cardiovascular diseases included heart attack, coronary heart disease, angina, congestive heart failure, and other problems.

Descriptive statistical analysis was used to compare population characteristics according to the indoor fuel types at baseline and fuel type change patterns during follow-up. The categorical variables were presented as frequency (proportions).

We performed Cox proportional hazards regression models to calculate adjusted hazard ratios (aHRs) with 95% confidence intervals (CIs) for the association between indoor solid fuel and NNDSD. We evaluated the proportional hazards assumption using the accumulated martingale residuals and found no obvious violation of the proportional hazards assumption [16]. In the baseline analysis, we estimate the relationship between cooking or heating fuel use, separately or simultaneously, and the risk of NNDSD among 7,884 participants. In the follow-up analysis, we evaluated the association between cooking fuel type conversion and the risk of NNDSD among 2962 participants. We conducted stratified analyses by baseline characteristics including age, sex, hypertension, and smoking status with the fully adjusted model.

Three models were fitted: Model 1 was a crude model; Model 2 adjusted for age, sex, and educational level; Model 3 was a fully adjusted model with additional adjustment for marital status, smoking status, residence region, self-reported economic level, hypertension comorbidity, cardiovascular disease, stroke, liver diseases, lung diseases, cancer or malignant tumor and kidney diseases.

We performed sensitivity analyses based on our primary Cox models. Firstly, we analyzed the association of household fuels conversion and NNDSD among non-smokers. Secondly, we performed stratified analyses by baseline characteristics to assess whether they could be effect modifiers of the studied association, including age, gender, smoking, and hypertension. Thirdly, we also performed sub-analyses with additional adjustment for population weighting concentrations of PM2.5 ( μ g / m 3 ) and geographical weighting concentrations of PM2.5. Fourthly, we examined the associations between household fuel exposure at baseline and NNDSD stratified by comorbidities using Cox models.

All analyses were performed by SAS version 9.4 (SAS Institute, Cary, NC, United States), and R 4.1.1. A two-sided test p < 0.05 was considered statistically significant.

The baseline characteristics of the participants in the baseline and follow-up analyses are shown according to the indoor fuel patterns (Table 1). We included 7,884 and 2962 enrolled participants in the baseline and follow-up analyses. Solid fuel was used for cooking by 3,709 (47.04%) participants and for heating by 3,732 (47.34%) participants. Solid cooking and heating fuel users were more inclined to be older, have families in poor financial conditions, be more likely to be smokers, and be illiterate or informally educated. 803 (27.11%) participants always used solid fuel during the follow-up period, 1,352 (45.6%) always used clean fuel, and 176 (5.94%) used fuel types that have switched from clean to solid fuel. Compared with participants who always used solid fuel, subjects who always used clean fuel and changed from solid to clean fuel tended to be females, aged from 45 to 65 years old, unmarried, and have a relatively high education level and high economic level at baseline.

In the fully adjusted models, compared to individuals who used clean fuel in the baseline, solid cooking fuel users showed a significantly increased risk of non-neoplastic digestive system diseases (NNDSD) (adjusted Hazard Ratio [aHR]: 1.17; 95% confidence interval [95% CI]: 1.02, 1.34). Similar results were found between solid (vs. clean) heating fuel use and NNDSD (aHR: 1.21; 95% CI: 1.07, 1.38). In addition, compared with those who were using clean fuel for both cooking and heating, the risk of NNDSD was found to be higher only in participants both cooking and heating with solid fuel (aHR: 1.34; 95% CI: 1.15, 1.55) (Table 2).

In the fully adjusted model of the follow-up analysis, we found that those who with always solid cooking fuel had a higher risk (aHR: 1.42; 95% CI: 1.09, 1.84) of NNDSD compared to those with always clean cooking fuel; and those who switched from solid fuel to clean fuel had a lower risk (aHR: 0.65; 95% CI: 0.49, 0.87) of NNDSD compared to those with always solid cooking fuel (Table 3).

Males (aHR: 1.66; 95% CI: 1.16, 2.35), over 65 years participants (aHR: 2.63; 95% CI: 1.65, 4.21) and people without hypertension (aHR: 1.48; 95% CI: 1.12, 1.96) who consistently used solid fuels for cooking had a higher risk of NNDSD than those who continued to use clean fuels for cooking. Compared with those who consistently used solid fuels for cooking, participants aged 65+ (aHR: 0.44, 95% CI: 0.26, 0.73), those without hypertension (aHR: 0.65, 95% CI: 0.47, 0.90), men (aHR: 0.65, 95% CI: 0.43, 0.99) and women (aHR: 0.63, 95% CI: 0.42, 0.94) switching cooking fuels from solid to clean had significantly lower risk of NNDSD (Table 4).

The subgroup analyses stratified by baseline characteristics showed that males (aHR: 1.50; 95%CI: 1.21, 1.87), people aged 45–65 years (aHR: 1.37; 95%CI: 1.16, 1.63), and subjects without hypertension history (aHR: 1.35; 95%CI: 1.14, 1.60) using solid fuel for both cooking and heating had a significantly greater risk of NNDSD than those with clean cooking and heating fuels (Supplementary Figure S2). Similar results to the primary analyses were observed in subgroup analyses stratified by comorbidities, restricted to non-smokers, and additionally adjusting for population weighting concentrations of PM2.5 ( μ g / m 3 ) and geographical weighting concentrations of PM2.5 (Supplementary Tables S2, S3, S5).

Several limitations should be noted. First, the use of self-reported diagnosis of NNDSD might have underestimated its prevalence, particularly among the older population and those with lower socioeconomic and educational backgrounds. Second, we obtained data on indoor fuel use by questionnaires, not by accurate external and internal exposure measurements. Moreover, we were unable to obtain data on concentrations of environmental air pollution to rule out potential impacts of specific pollutants on the digestive system. However, using measurements of global surface PM2.5 concentrations from the Atmospheric Composition Analysis Group at the University of Washington, we obtained the annual average PM2.5 concentration for each province of China [44]. We conducted baseline analysis of effect of solid fuel use for cooking or/and heating on NNDSD additionally adjusted for PM2.5. The results of the analysis after the inclusion of external confounding are similar to those of the main analysis, which indicates the stability of our results. Future studies might include objective assessment of external pollutants at the individual level. Third, NNDSD in the present study covers all non-neoplastic digestive system diseases, and there is no information about NNDSD subtypes (Peptic ulcer, inflammatory bowel disease, etc.); further studies may explore such associations in different NNDSD subtypes. Finally, the information on heating fuel during follow-up is unavailable. Future studies could incorporate heating fuel usage to fully explore the impact of fuel type conversion on adverse outcomes.

Indoor solid fuel use for cooking and/or heating was an independent risk factor for non-neoplastic digestive system diseases (NNDSD). Additionally, switching fuel types from solid to clean cooking fuels may help reduce the impact of indoor solid fuel use on NNDSD. The present study highlights the urgency of switching to clean cooking fuels, with major public health implications for developing countries.